Older adult 2 *(Knowledge)*

All care homes actively encouraged residents to keep hydrated. *"I would say actually physically passing the drink to them, so you would encourage them* to *drink and usually they say, oh you know, I've had a lot today. We say, oh well just a little bit more and try and just sort of encourage them."* Care home staff 3 *(Skills)* All general practice staff encouraged hydration as a preventative and self-care method. *"Hydration is what I focus on."* GP 1 *(Skills)*

Some care homes would decide to encourage drinking before concluding that the resident has a UTI. *"as harsh as it sounds we give them a drink and see if that perks them up and we see how far the confusion goes, we don't automatically think UTI, it could be dehydration."* Care home staff 4 *(Memory, attention and decision processes)*

One resident described drinking less in order to avoid urinating at night. *"because I keep going at night. Which isn't right* . . . *I'm not drinking more. I* hopefully *am drinking less."* Care home resident 2 *(Goals)*

Care staff believed that residents do not want to drink to avoid visiting the toilet regularly. *"they get worried about drinking too much because they don't want to keep going to the toilet."* Care home staff 7 *(Social influence)*

## **Antibiotics**

Many general practice staff were prescribing UTI antibiotics over the phone to care home residents. *"in the volume of work it's often, as you quite* rightly *say it's often over the phone."* GP 1 *(Environmental context and resources)*




## **Leaflet use**

Care staff that had used the leaflet used it as their guide for identifying and managing UTIs. *"It's our guide for how we appoint* (identify) *this UTI."* Care home staff 7 *(Environmental context and resources)*

moving forwards. *"We feel that if it's not required then it's one less thing that you have to try* and *get from people."* Care home staff 3 A few general practice staff reported prescribing antibiotics for UTI as a result of demanding patients. *"there is always still that pressure to prescribe. I came here because I've got a urine infection and you are going to prescribe me antibiotics no matter what you*

*think."* Nurse practitioner 3 *(Social influence)* Older adults do not mind taking antibiotics as long as it makes them well. *"I just want to feel well, and I don't care what I take to feel like me you know."* Care home residents 1 *(Goals)* Older adults aware of D-mannose were receptive to trying it as an antibiotic alternative. *"I went in and she immediately said I've been looking* something *up for you and she'd found them, they're expensive but if it's going to work then I'll pay the money."* Older adult 3 *(Social influence)*

A few general practice staff expressed interest in conducting a UTI antibiotic audit. *"Auditing the antibiotic use would be really interesting to do, if* we *could do that that would be good."* General practice staff 1 *(Intentions)*

One general practice mentioned auditing their UTI antibiotics. *"We've re-audited the antibiotic prescribing* . . . *it's kind of improved* . . . *my* trimethoprim *prescribing's halved."* General practice staff *(Behavioural regulation)* Care staff and general practice staff were confident in their ability to manage diagnosed UTI. *(Beliefs about capabilities)* All care staff reported changing soiled incontinence pads immediately, even if the resident has a limited pad allowance. *"So, the residents are restricted on how many day or night pads that they're assessed or allocated but if we find a resident that is soiled or their pad is wet we automatically change it." Care* home staff 5 *(Environmental context and resources)*


*Antibiotics* **2021**, *10*, 83

staff and relatives identify and manage UTIs. *"I think that might help the* relatives *understand a little bit more."* Care home staff 3

*(Environmental context and resources)*


*Antibiotics* **2021**, *10*, 83

#### 2.1.5. Leaflet Interaction with UTI Diagnostic Tools and Other Resources

GP staff reported that they mostly used the leaflet alongside either PHE or locally developed UTI diagnostic tools as part of quality improvement initiatives within general practice. Some reported that quality improvement was their overall goal.

One stakeholder was optimistic that their work in implementing UTI training, leaflet use and diagnostic guidance had reduced the number of unrequested urines being submitted by patients to several GP practices in their region.

#### 2.1.6. Indications of Behaviour Change Following Use or Implementation of the Leaflet

All commissioners intend to continue their promotion of the leaflet alongside the diagnostic flow charts within their local UTI or infection prevention campaigns.

Care staff are motivated by wanting the best for their residents and improving their wellbeing, and they reported consistent use of genital hygiene and hydration strategies as prevention methods with all residents. Barriers, reported by older adults and care staff, to implementing these prevention strategies for older adults included reluctance to drink in order to avoid regular toilet visits and limited incontinence pad allowance, although care homes also reported buying additional pads to supplement their allocations. One stated: *"because I keep going at night. Which isn't right* . . . *I'm not drinking more. I hopefully am drinking less."* Care home resident 2

Some care staff intended to cease use of urine dipsticks moving forwards. However, some care staff intended to continue using them and a few GPs were using urine culture to inform diagnostic decision making; dipsticking and culture were perpetuated by unrequested urine samples being dropped off at general practice receptions and perceived pressure from care staff to use urine dipsticks. One stakeholder had reduced unrequested urine submissions in several GP practices in their region through the combined use of training and implementation of the leaflet and diagnostic guidance.

Care staff reported being confident in their ability to identify changes in resident behaviour which indicated illness, but had some difficulty in distinguishing UTIs from other illnesses due to similar presentations. However, despite the leaflet providing signs and symptoms of UTI alongside other causes of confusion, clinicians reported that they had difficulty in diagnosing UTI in patients with dementia. Most clinicians reported that diagnosis of UTI was complicated by atypical presentations, vague symptoms reported by carers, incontinence and other conditions presenting like a UTI. Therefore, clinicians requested further information and resource to support diagnosis in this group.

#### 2.1.7. Key Findings and the Theoretical Domains Framework

To identify important behavioural determinants and using a deductive approach, key themes were placed into the corresponding 14 domains of the TDF listed in Box 1. The first key theme, 'Use and implementation of the UTI leaflet', addresses objectives 1–5 in order to determine the leaflet's acceptability, the feasibility of its use, perceived value, interaction with other resources, barriers to using the leaflet and potential further developments to the leaflet. The second and third key themes(Identifying/diagnosing a UTI' and 'Managing and/or treating a UTI) address objective 6, which explores potential indications of behaviour change by examining current behaviours around UTI diagnosis and management. This process is illustrated in Table 1, with each of the three themes used as column headings.

**Box 1.** The domains of the Theoretical Domains Framework.


#### **3. Discussion**

#### *3.1. Summary*

All participants including care staff, general practice staff, older adults and stakeholders reported that the leaflet is a valuable IPC tool, suitable for care homes and general practice, that can be used alongside diagnostic tools and antibiotic guidance, reinforcing messages to older adults while also providing a useful guide for care staff. Participants believed that younger adults would benefit from the leaflet and it should be provided in other health settings.

Participants provided valuable suggestions for dissemination, such as a provision to families and friends, and placement in clinical waiting rooms, reception areas and next to urine submission boxes.

Lack of awareness of the leaflet was the biggest barrier to its use, and implementation barriers prohibited commissioners from effective dissemination. Suggested changes to the leaflet included use of less alarmist terminology, the inclusion of three-day antibiotic courses, the inclusion of D-mannose, and mention of atrophic vaginitis.

Commissioners reported that they would continue to promote the leaflet locally and some care staff would cease use of urine dipsticks moving forwards. Barriers to preventing UTI which are not easily addressed by the leaflet include difficulties in diagnosing UTI in older adults with dementia, and reluctance by older adults to hydrate sufficiently to help reduce their toilet visits.

#### *3.2. Comparison with Existing Literature*

Patient information leaflets on antibiotic prescribing for a variety of conditions have led to reductions in antibiotic use [15,16]. However, process evaluation information, such as the acceptability and feasibility of the patient information leaflets, is rarely reported. In a systematic review of leaflet effectiveness [15], many studies only report clinical outcomes, and are therefore limited by not assessing patient acceptability or feasibility of use in real-world settings outside of controlled trial conditions.

To the best of our knowledge, there have not been any leaflets developed for older adults on the topic of UTI, and therefore drawing comparisons across different leaflets, audiences and conditions was inherently difficult. A qualitative study evaluating an interactive information booklet for parents of young children with respiratory symptoms 'When should I worry?' explored the views and opinions of parents and clinicians as part of a trial measuring the booklet's effectiveness [17]. Francis et al. found that parents and clinicians valued the leaflet and many parents had kept the booklet for future references. They concluded that the role of leaflets and other information resources can help facilitate effective communication, and indeed, the link between effective communication of health information with clinical outcomes is well documented [18]. Despite some similarities

to the present study, these findings must be accepted with caution as leaflets can vary in content and quality, and therefore perceived value will vary.

A qualitative study exploring patient views of medication information leaflets found that leaflet font size, paper quality, writing style and size of the paper are important factors for enhancing readability [19]. We addressed these design parameters iteratively during the development of the UTI leaflet, which may explain why there were few criticisms relating to its aesthetics [20].

A similar study by Fleming, et al. [21] explored antibiotic prescribing in long-term care facilities using the TDF and Behaviour Change Wheel (BCW), and recommended the provision of education on the topic of antibiotics, prescribing guidelines and AMR, with the provision of management guidelines and supporting evidence. Even though Fleming et al. [21] explored general antibiotic prescribing across conditions without a UTI focus, a similar recommendation from this study included the provision of education for care homes on UTI diagnosis, management and ASB. The current study further supports the need for additional resources in care homes, specifically around UTI education and the provision of guidance on ASB, diagnosis of UTI and urine dipstick use. Using urine dipsticks to diagnose UTI in older adults is not accurate due to high rates of ASB [22–24]. GP staff reported using the leaflet as part of quality improvement programmes. The leaflet compliments the To Dip or Not To Dip (TDONTD) [8] quality improvement programmes that aim to improve the diagnosis and management of UTI in older patients. Regional evaluations of TDONTD have found significant reductions in UTI antibiotics prescribed in care home residents, unplanned hospital admissions, urosepsis and acute kidney injury [8].

#### *3.3. Strengths and Limitations*

This is the first study to explore the acceptability and feasibility of a patient-facing evidence-based UTI leaflet for older adults and their carers in the community, including in care home settings. This study offers valuable insight into implementation and attitudes towards the leaflet as well as current diagnostic and management practices across both settings. To the best of our knowledge, in the UK, there is currently no nationally available patient-facing information leaflets for older adults on the topic of UTI, and there are no evaluation studies assessing acceptability and feasibility, or implementation of patientfacing resources on UTIs for any age group.

The present study included a large number of participants from a range of backgrounds. However, care staff and general practice staff may have had greater interest in UTIs, leading to some selection bias. Selection bias was reduced by inviting care homes and GP practices from two regions, approaching potential participants in random order [25] and by providing an incentive to participate. Our findings indicate a wide variation in management and use of the tools, which indicates that selection bias was minimised. However, many participants had not used or seen the leaflet before and were viewing it for the first time as part of this study. This suggests that either implementation strategies by the CCGs could be improved, or that more time was needed between initial leaflet implementation and participant recruitment [26].

A further sample limitation is that only adults with full capacity were approached to take part, and therefore only older adults who could read and understand the leaflet took part, and no data were captured for those older adults deemed unsuitable to receive the leaflet due to the content being 'overwhelming'. Insights from this group may have proved useful for informing improvements to the leaflet or informing the development of a new leaflet.

Qualitative methodology was employed to gain detailed insight into leaflet use and how it contributes to identifying, managing and preventing UTIs. Using both interviews and focus groups facilitated recruitment as participants could choose the format to suit their preference, and use of both provided breadth of exploration across many individuals, and in-depth exploration with individual experiences and attitudes [27].

#### *3.4. Implications*

As the older adult UTI leaflet was reported as "invaluable" by patients and carers, GP staff and CCGs reported that they would continue to implement the leaflet "as a guide for patients and carers to help identify and manage UTIs". We suggest that it should be made available in both electronic and hard-copy formats to suit users' preferences, as part of a quality improvement program to advance the management of UTIs. However, due to reported implementation barriers, commissioners may want to consider electronic dissemination as an inexpensive and potentially easier method of promotion to care homes and general practices. This could include use of QR codes or integration into GP clinical systems for ease of access including use of computer prompts as reminders.

Following minor changes to terminology, and the inclusion of information about a three-day antibiotic course, D-mannose and vaginal atrophy, the leaflet will correspond to the current PHE UTI diagnostic flowcharts and NICE/PHE UTI guidance information [28,29] and should therefore be disseminated in care homes as an educational guide to staff. The leaflets should also be disseminated to older adult care home residents of any age. However, where residents lack capacity or may find the leaflet overwhelming, the leaflet could be given to families and friends of residents to provide education and to reinforce health behaviour messages from staff around hydration, self-care and prevention.

General practices should consider the provision of the leaflet as a waiting room resource or to be given/emailed to patients during or following consultations, or to be given at reception to educate patients bringing in urine samples. Primary care clinicians using the PHE diagnostic flow charts may also want to consider the leaflet as a complementary triaging resource to reinforce and communicate their diagnostic decisions with patients.

As weak working relationships between care homes and general practices could be a contributing factor to lack of implementation, commissioners should consider promoting the leaflet during training sessions for both care homes and general practices as an infection prevention and control resource. Regional strategies must include plans for widespread dissemination to care homes including monitoring of attendees and non-attenders to training sessions, monitoring of leaflet use with TARGET UTI audits. A greater implementation may be needed in OOH and community pharmacy settings.

National promotional strategies through the RCGP and CAA should be considered to ensure national dissemination. Health Education England has a short video explaining the value of the leaflet and diagnostic flowchart [30].

Currently, there is a separate non-pictorial UTI leaflet for younger adults that has been used in GP and pharmacy settings [31]. As participants reported that the information in the older adult leaflet is relevant to people of all ages and some patients did not relate to the label 'older adults', a combined leaflet may be useful. A combined leaflet has been developed, but further work is needed to evaluate this in primary care settings.

#### **4. Materials and Methods**

Study design: This is a cross-sectional qualitative study using interviews and focus groups informed by the TDF [13].

Leaflet implementation: To understand implementation and usage in a real-world setting, PHE researchers were not involved in the implementation of the leaflet. As such, those unfamiliar with the leaflet were still eligible to participate to understand their management of UTIs, their reasons for not having seen/used the leaflet, and their assessment of the leaflet's value. All participants were sent the leaflet alongside the study information form to allow reflection prior to the discussions.

Gloucestershire CCGs' plan for implementing the UTI leaflet included posting the leaflet to all general practices; workshops and educational training on the UTI guidance offered to all care homes and GP practices; a hydration campaign in care homes, promoted using merchandise and a touring marketing bus. The East Kent CCG disseminated the leaflet to all practices and care homes electronically or in hard copy depending on preference. Links practitioners were established in every care home and general practice in East

Kent, who were then offered training on the UTI guidance with the view of disseminating it in their respective settings.

Data collection started after the CCGs had implemented the leaflet in each region for a minimum of four months. Each region aimed to saturate their regions with the leaflet but did not monitor uptake to determine whether this had occurred.

#### *4.1. Participant Selection and Eligibility*

General practice staff and care staff were invited from two CCG regions in the UK, Gloucestershire CCG and East Kent CCG, which were selected due to their intentions to disseminate the leaflet and willingness to support the study. To avoid recruiting only AMR enthusiasts, regions were selected based on antibiotic prescribing at a primary care level [32], and individual facility prescribing data were not explored. Region size and regional demographic variation were taken into account as these can impact implementation and health literacy [33]. Stratification by rural/urban allowed for a variance in participant demographics.

Lists were formed of all care homes and general practices in each region. All care homes and general practices were contacted with an introductory letter describing the study, and the vast majority of facilities did not respond to the initial letter—only a minority responded expressing an interest. Each list was then randomised using Excel's RAND function, and two weeks following receipt of the letter, care homes and general practices were contacted in random order with a follow-up telephone call.

In accordance with the Enabling Research In Care Homes (ENRICH) guidelines, care homes considered 'inadequate' in the CQC inspection rating were not selected for this study [34]. This only equated to 3 care homes across both regions.

Managers or the point of contact were asked to disseminate the study information to recruit staff and older adults. Managers/contacts were requested to approach older adults with experience of UTI and who were able to provide informed consent.

Stakeholders were identified using known contacts through PHE and previous engagement with professional societies, with the aim to recruit national representatives of primary care clinicians and care staff, as well as commissioners of primary care services to discuss their implementation and regional strategies.

All participants gave written and verbal consent and were offered £20 in vouchers; staff were offered certificates of participation as well as vouchers.

#### *4.2. Data Collection*

General practice staff and care staff were offered interviews or focus groups depending on their preference. Focus groups were conducted in a quiet room provided by the facility and were heterogenous i.e., all job roles were permitted to attend. Interviews were either face to face or via telephone, depending on participant preference. Older adults were offered interviews rather than focus groups, as discussing experiences of UTI could be considered personal. However, three older adults from one care home requested a focus group.

Seventeen interviews were conducted and lasted 13–47 min, and 12 focus groups containing between 3 and 10 individuals lasted for 24–57 min. After each, discussion field notes were made of important topics and non-verbal data. All interviews and focus groups were conducted by one researcher (LJ).

#### *4.3. Interview Schedules*

Questions were informed by the TDF [13] and the qualitative findings from the needs assessment to develop the leaflet [6].

The schedules were semi-structured and used flexibly (see Supplementary Material 3). Interviews and focus groups with general practice staff and care staff covered their leaflet use, and barriers and facilitators to usage. Discussions with older adults explored their experiences of having UTIs, their attitudes and opinions of receiving the leaflet, its content

and its perceived usefulness. Interviews with stakeholders focused on organisational barriers or facilitators to implementation. The interview schedules were piloted with 1–2 people from each group, and pilot data were included in the results as no major amendments were made.

The care home managers viewed the older adult interview schedule in order to ensure they were aware of the questions being asked to their residents and identified older adults with sufficient understanding to participate. However, managers were asked to keep schedules confidential to prevent potential priming.

#### *4.4. Data Analysis*

Transcripts were analysed by one researcher (LJ) in Nvivo 11 [35] using Inductive Thematic Analysis (ITA). Following ITA, a deductive approach was adopted by placing key themes into the domains of the TDF to identify important behavioural determinants. A double coder (PD) coded 10% (3) of the transcripts.

#### *4.5. Researcher Context*

The primary investigator, LJ, has previous experience of using the TDF, conducting research in this area with care staff, GP staff, and older adults as part of the leaflet development work. Researcher bias in this study has been mitigated by utilising patient input into the interview schedule development, use of a double coder and by presenting the results to both regions and receiving their feedback.

#### **5. Conclusions**

This novel study has provided insights into the acceptability and feasibility of using the UTI leaflet for older adults and their carers in general practice and care home settings, including current diagnostic and management practices, variation in implementation, and barriers and facilitators. Consequently, this study highlights the ways in which the leaflet has influenced recognition and treatment behaviours, and also ways to improve the leaflet, implications for successful implementation, and suggestions for ways in which new interventions could overcome the barriers to appropriate UTI diagnosis and management. A combination of new complementary interventions, and improvements to the leaflet and its implementation will be needed in order to further influence behaviour change in this context.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2079-6 382/10/1/83/s1, Supplementary Material 1: The leaflet 'urinary tract infections; a leaflet for older adults and carers'; Supplementary Material 2: Final recruitment strategy and figures; Supplementary Material 3: Interview schedules

**Author Contributions:** Conceptualisation, L.F.J. and C.M.; methodology, L.F.J.; formal analysis, L.F.J. and P.D.; investigation, L.F.J.; writing—original draft preparation, L.F.J.; writing—review and editing, H.W., D.H., D.M.L. and C.M.; supervision, H.W., D.H., D.M.L. and C.M.; project administration, L.F.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Ethics Committee of the University of the West of England (UWE REC REF No: HAS.18.10.042 Jones).

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The anonymised data presented in this study are available on request from the corresponding author. The data are not publicly available due to the sensitive nature of the topic.

**Acknowledgments:** Thank you to Julie Brooke who helped with recruitment and to the local commissioning teams in Gloucestershire and East Kent, especially Leslie MacLeod-Downes and Esther

Taborn who supported this work, and the care staff, general practice staff, older adults and stakeholders who contributed valuable insights.

**Conflicts of Interest:** L.F.J., P.D., D.M.L. and C.M. work on Public Health England's TARGET Antibiotics programme of work, developing and evaluating antimicrobial stewardship interventions.

#### **References**


#### **Risk Factors for the Acquisition of** *Enterococcus faecium* **Infection and Mortality in Patients with Enterococcal Bacteremia: A 5-Year Retrospective Analysis in a Tertiary Care University Hospital**

	- <sup>3</sup> Division of Infectious Diseases, Department of Public Health, Kobe University Graduate School of Health Sciences, Kobe 654-0142, Japan; ko1.kitgwa@gmail.com
	- <sup>4</sup> Division of Urology, Kobe University Graduate School of Medicine, Kobe 650-0017, Japan; yym1112@gmail.com (Y.Y.); masato@med.kobe-u.ac.jp (M.F.)
	- <sup>5</sup> Division of Advanced Medical Science, Kobe University Graduate School of Science, Technology and Innovation, Kobe 657-8501, Japan
	- <sup>6</sup> Department of Medical Technology, Kobe Tokiwa University, Kobe 653-0838, Japan; osawak@kobe-u.ac.jp

**Abstract:** The incidence of bacteremia caused by *Enterococcus faecium*, which is highly resistant to multiple antibiotics, is increasing in Japan. However, risk factors for the acquisition of *E. faecium* infection and mortality due to enterococcal bacteremia are not well known. We compared demographic, microbiological, and clinical characteristics using a Cox regression model and univariate analysis. We performed a multivariate analysis to identify risk factors for patients treated between 2014 and 2018. Among 186 patients with enterococcal bacteremia, two groups included in the Kaplan–Meier analysis (*E. faecalis* (n = 88) and *E. faecium* (n = 94)) showed poor overall survival in the *E. faecium* group (HR: 1.92; 95% confidence interval: 1.01–3.66; *p* = 0.048). The median daily antibiotic cost per patient in the *E. faecium* group was significantly higher than that in the *E. faecalis* group (\$23 (\$13–\$34) vs. \$34 (\$22–\$58), *p* < 0.001). *E. faecium* strains were more frequently identified with previous use of antipseudomonal penicillins (OR = 4.04, *p* < 0.001) and carbapenems (OR = 3.33, *p* = 0.003). Bacteremia from an unknown source (OR = 2.79, *p* = 0.025) and acute kidney injury (OR = 4.51, *p* = 0.004) were associated with higher risks of 30-day mortality in patients with enterococcal bacteremia. Therefore, clinicians should provide improved medical management, with support from specialized teams such as those assisting antimicrobial stewardship programs.

**Keywords:** enterococcal; bacteremia; epidemiology; risk factors; mortality; antimicrobial stewardship

#### **1. Introduction**

*Enterococcus* species are Gram-positive, facultative anaerobic cocci that constitute the normal bacterial flora in human and animal intestines. Enterococcal bacteremia is associated with a high mortality rate and prolonged hospitalization [1–3]. *Enterococcus faecalis*, followed by *E. faecium* are the most frequent *Enterococcal* species that cause bacteremia [4]. *Enterococcus* species are intrinsically cephalosporins-resistant, which inhibit bacterial cell wall synthesis. Primarily, infections caused by vancomycin-resistant enterococci are associated with higher mortality and are a major problem in the United States and Western countries [3,5,6]. However, the prevalence of vancomycin-resistant *Enterococcus* infections

**Citation:** Uda, A.; Shigemura, K.; Kitagawa, K.; Osawa, K.; Onuma, K.; Yan, Y.; Nishioka, T.; Fujisawa, M.; Yano, I.; Miyara, T. Risk Factors for the Acquisition of *Enterococcus faecium* Infection and Mortality in Patients with Enterococcal Bacteremia: A 5-Year Retrospective Analysis in a Tertiary Care University Hospital. *Antibiotics* **2021**, *10*, 64. https://doi.org/10.3390/ antibiotics10010064

Received: 12 December 2020 Accepted: 7 January 2021 Published: 11 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

in Japan is markedly lower than that in other countries [4,7,8]. In contrast, most *E. faecalis* are susceptible to penicillins, although *E. faecium* tends to have resistance to some antimicrobial agents, including penicillins, aminoglycosides, and carbapenems [4]. The hospital cost and mortality for patients with multidrug-resistant pathogens were higher than those with antimicrobial-susceptible pathogens [9–11]. Many studies have reported that previous use of broad-spectrum antibiotics is a risk factor for acquiring the multidrug-resistant pathogens [12–14]; however, few studies have been done to identify the correlation between the previous antibiotic exposure and the acquisition of *E. faecium* strains [15]. In the United States and Western countries, the enterococcal isolates of *E. faecalis* (80–90%) and *E. faecium* (5–20%) [8,16] are considerably different than those in Japan (*E. faecium* strains account for 40%) [4]. Although there are many studies on enterococcal bacteremia, the clinical outcomes, epidemiological features, and risk factors for nosocomial infection produced different results depending on the country, hospitalization ward, or patient characteristics. Only few studies have described the situation in Japan [17,18], and the conclusions are inconsistent. Therefore, we aimed to investigate the clinico-epidemiological features and risk factors predisposing to the acquisition of *E. faecium* strains and mortality due to nosocomial enterococcal bacteremia.

#### **2. Results**

#### *2.1. Patient Data*

During the 5-year study period (2014–2018), 186 patients had bacteremia caused by *Enterococcus* species. The most common *Enterococcus* species were *E. faecium* (n = 94, 51%), followed by *E. faecalis* (n = 88, 47%), *E. avium* (n = 1, 0.5%), *E. casseliflavus* (n = 1, 0.5%), *E. raffinosus* (n = 1, 0.5%), and *E. gallinarum* (n = 1, 0.5%). As few patients had bacteremia caused by *E. avium*, *E. casseliflavus*, *E. raffinosus*, and *E. gallinarum*, only the clinical characteristics of bacteremia with *E. faecalis* and *E. faecium* were further investigated among two subgroups. After excluding four cases of bacteremia caused by strains other than *E. faecalis* or *E. faecium*, 182 patients were eligible for inclusion in this study. The survival rates for both study groups are shown in the Kaplan–Meier curves in Figure 1, showing a significant decrease in overall survival rates among the patients with *E. faecium* bacteremia (hazard ratio (HR): 1.92; 95% confidence interval (CI): 1.01–3.66; *p* = 0.048).

**Figure 1.** Kaplan–Meier survival curves of patients with *E. faecalis* and *E. faecium* bacteremia.

*2.2. Demographic and Clinical Characteristics*

The demographic and clinical characteristics of the study participants are shown in Table 1. There was no significant difference between groups regarding sex, age, hospitalization ward, length of hospitalization before the onset of bacteremia, quick Sequential Organ Failure Assessment (qSOFA) score ≥ 2, use of invasive devices, or surgical history. The coexistence of hepatobiliary (*p* = 0.005) and hematologic (*p* = 0.027) tumors were more

frequently observed in patients with *E. faecium* bacteremia. The most common primary source of infection arose from an insertion of a central venous catheter (n = 41, 23%), followed by cholecystocholangitis (n = 38, 21%), urinary tract infection (n = 18, 9.9%), and intra-abdominal infection (n = 12, 6.6%). *E. faecium* bacteremia originated more frequently from cholecystocholangitis (*p* < 0.001) and febrile neutropenia (*p* = 0.015) than in the *E. faecalis* group. However, the *E. faecalis* group had an unknown source of infection more frequently (*p* = 0.041). The incidence of urinary tract infections was lower in *the E. faecium* group than in the *E. faecalis* group (*p* = 0.017). A history of antibiotic therapy with antipseudomonal penicillins (*p* < 0.001) and carbapenems (*p* < 0.001) was more frequently observed in the *E. faecium* group. Based on the analysis by a logistic regression model, preexisting hematologic tumors (adjusted OR = 7.85, *p* = 0.004), cholecystocholangitis (adjusted OR = 5.21, *p* = 0.001), and previous use of both antipseudomonal penicillins (adjusted OR = 4.04, *p* < 0.001) and carbapenems (adjusted OR = 3.33, *p* = 0.003) were independent risk factors for the acquisition of an *E. faecium* infection.

**Table 1.** Intergroup comparison of demographic and clinical characteristics of patients with enterococcal bacteremia and risk factors for the acquisition of *E. faecalis* and *E. faecium.*



**Table 1.** *Cont.*

IQR: interquartile range, qSOFA: quick Sequential Organ Failure Assessment, MRSA: methicillin-resistant *Staphylococcus aureus*, VCM: vancomycin, DAP: daptomycin, LZD: linezolid.

#### *2.3. Microbiological Data*

Table 2 shows the microbiological characteristics of patients with enterococcal bacteremia. There was no significant intergroup difference in polymicrobial cultures; all isolates of *E. faecalis* and 15% (14/94) strains of *E. faecium* were susceptible to ampicillin. We found no vancomycin-resistant isolates among the enterococci. No imipenem-resistant *E. faecalis* strain was isolated; however, all *E. faecium* isolates were imipenem-resistant. Susceptibility to levofloxacin was detected in 91% (80/88) and 12% (11/94) of *E. faecalis* and *E. faecium* isolates, respectively.

#### *2.4. Clinical Management and Outcomes*

Table 3 shows the clinical management and outcomes of patients with bacteremia caused by *E. faecalis* and *E. faecium* infections. The rate of source control with drainage in the *E. faecium* group was ~1.7 times higher than that in the *E. faecalis* group, although the difference was not significant (*p* = 0.15). Non-antipseudomonal penicillins (n = 55, 63%) was the most common antibiotic used for *E. faecalis* bacteremia, whereas vancomycin (n = 74, 79%) was most frequently prescribed in *E. faecium* bacteremia. Compared with *the E. faecium* group, the *E. faecalis* group more frequently received penicillins (*p* < 0.001) and aminoglycosides (*p* = 0.031). Vancomycin (*p* < 0.001) and other anti-methicillin-resistant *Staphylococcus aureus* (MRSA) agents (*p* = 0.028) were more frequently administered in the *E. faecium* group. Non-antipseudomonal penicillins (*p* = 0.022) and quinolones (*p* = 0.022) were prescribed for longer durations in the *E. faecalis* group, whereas the median duration of vancomycin use was longer in the *E. faecium* group (*p* < 0.001). When vancomycin treatment took more than three days, all patients underwent therapeutic drug monitoring. The *E. faecalis* group showed a shorter median duration to the commencement of initial antibiotic therapy against enterococci (*p* = 0.049), although there was no significant intergroup difference in the total duration of antibiotic treatment (*p* = 0.99). The median daily antibiotic cost per patient in the *E. faecium* group was significantly higher than that in the *E. faecalis* group (\$23 [\$13–\$34] vs. \$34 [\$22–\$58], *p* < 0.001). Patients in the *E. faecium* group more frequently attained a vancomycin median serum trough concentration ≥20 mg/L (*p* = 0.007) than in the *E. faecalis* group. Acute kidney injury (AKI) was observed in both groups but was more frequent in *the E. faecium* group (*p* = 0.02). The clinical outcome for patients with enterococcal bacteremia was analyzed based on the length of hospitalization; however, no significant between-group difference (*p* = 0.34) was observed.

**Table 2.** Microbiological characteristics of enterococcal bacteremia.


**Table 3.** Clinical treatments and outcomes of patients with enterococcal bacteremia.


IQR: interquartile range, MRSA: methicillin-resistant *Staphylococcus aureus*, VCM: vancomycin, DAP: daptomycin, LZD: linezolid, AKI: acute kidney injury.

In this study cohort, the overall 30-day mortality rate was 23% (41/182). Table 4 shows the risk factors associated with the overall 30-day mortality in patients with enterococcal bacteremia. Patient groups with variables such as admission to an intensive care unit, an unknown source of infection, qSOFA score ≥ 2, previous immunosuppressive and corticosteroid treatment, or encountered AKI were associated more frequently with 30-day death. These variables were included in the multivariate logistic regression analysis revealing the following independent risk factors for mortality: unknown source of infection (OR = 2.79, *p* = 0.025), qSOFA score ≥ 2 (OR = 2.96, *p* = 0.024), previous corticosteroid treatment (OR = 2.84, *p* = 0.034), and AKI (OR = 4.51, *p* = 0.004).


**Table 4.** Risk factors that were associated with 30-day mortality due to enterococcal bacteremia.

IQR: interquartile range, qSOFA: quick Sequential Organ Failure Assessment.

#### **3. Discussion**

This observational retrospective study analyzed the epidemiological and clinical outcomes of enterococcal bacteremia and evaluated the risk factors for the acquisition of *E. faecium* and *E. faecalis* infection and mortality in enterococcal bacteremia. We found a significant increase in *E. faecium* bacteremia among patients with enterococcal bacteremia, especially those who were previously treated with antipseudomonal penicillins and carbapenems. We also found that severely ill patients, those with an unknown source of infection, and AKI during treatment conferred higher risks of mortality in enterococcal bacteremia. The results of our analyses suggest the need for greater efforts to provide accurate medical treatment, including appropriate antimicrobial use in patients with enterococcal bacteremia.

As a leading cause of nosocomial bacteremia, enterococci have become more prevalent worldwide. In particular, the spread of vancomycin-resistant *Enterococcus* has become a major public health problem in the United States and in Western Europe [3,5,6]. However, we found no vancomycin-resistant strains of enterococci in our hospital during the 5-year study period. *E. faecalis* and *E. faecium* are two major *Enterococcus* species that can cause various complicated infectious diseases [3,4,16] and were the most common strains in our cohort. In agreement with the results of an earlier study [19], we observed significantly lower survival rates with *E. faecium* than with *E. faecalis*, as determined using the Kaplan– Meier survival curves. Furthermore, the *E. faecium* group showed a higher daily antibiotic cost than the *E. faecalis* group. Infections with drug-resistant pathogens are typically associ-

ated with increased hospitalization costs [9]. De-escalation therapy to narrower spectrum antibiotics is a cost-saving strategy [20,21]. After bacterial identification, antibiotics should be changed, if needed, to administer appropriate targeted antibiotic therapy in accordance with bacterial culture and susceptibility data. In general, patients infected with enterococcal bacteremia are older and more likely to develop renal failure; thus, less toxic regimens such as penicillins may be preferred. In this study, because all the *E. faecalis* remained susceptible to penicillins, the *E. faecalis* group received narrower spectrum antibiotics, such as non-antipseudomonal penicillins. Whereas the *E. faecium* group, which had a higher resistance to penicillins, was most frequently prescribed vancomycin. Non-antipseudomonal penicillins, which constitute the most prescribed antibiotics in the *E. faecalis* group, were of a lower price in Japan than anti-MRSA agents [22], which resulted in a lower median daily antibiotic cost in the *E. faecalis* group. We demonstrated that *E. faecium* bacteremia caused more serious problems regarding therapeutic outcomes.

Previous studies have focused on risk factors for the acquisition of *E. faecium* infections among various ethnic groups [15,23], but little is known about the risk factors in Japan [17]. In this study, multivariate analysis revealed that the risk factors for bacteremia due to *E. faecium* included preexisting illness (hematologic tumor), source of infection (cholecystocholangitis), and previous use of broad-spectrum antibiotics (antipseudomonal penicillins and carbapenems). For patients with nosocomial or intra-abdominal infections, broad-spectrum antibiotics were often prescribed to treat anaerobic bacteria and Gram-negative bacteria such as *Pseudomonas aeruginosa*. In a previous report, carbapenems were reported as the only independent risk factor associated with *E. faecium* bloodstream infections [15]. However, this study is the first to characterize antipseudomonal penicillins as a predictive risk factor for the acquisition of *E. faecium*. In Japanese national and public university hospitals, the consumption of antipseudomonal penicillins increased five times from 2008 to 2015 [24], and the rates of *E. faecium* among enterococcal bacteremia patients increased from 36% to 43% in the same period [4]. The higher rates of *E. faecium* isolation might correlate with the increased use of antipseudomonal penicillins in Japan. Broad-spectrum antibiotics might destroy the normal anaerobic flora of the gastrointestinal tract by selective elimination of enterococci due to the bactericidal activity against these organisms, which might subsequently induce infectious diseases. A case-control study revealed that broad-spectrum antibiotic therapy, including antipseudomonal penicillins, was a risk factor for *P. aeruginosa* resistance among hospitalized patients [25]. The Infectious Diseases Society of America guidelines recommend the implementation of antibiotic stewardship programs to restrict the prevalence of antimicrobial-resistant pathogens [26]. Overuse of broad-spectrum antibiotics leads to the selective growth of resistant bacteria; thus, one of the aims of antimicrobial stewardship is to promote appropriate antibiotic use. Our hospital has practiced antimicrobial stewardship since 2010 to optimize antibiotic usage [27–29]; however, the consumption of antipseudomonal penicillins significantly increased between 2009 and 2016 [27]. Thus, narrower spectrum antibiotics should be prescribed to avoid the development and prevalence of bacterial resistance as much as possible with consideration to preexisting infectious diseases and patient conditions even before the onset of bacteremia.

Empirical antibiotic therapy is often commenced before pathogen identification and without susceptibility data. *E. faecium* is typically resistant to penicillins [4]; thus, when enterococcal infection is suspected, initial empiric therapy often requires the prescription of anti-MRSA agents, especially vancomycin. We found that the *E. faecium* group showed a longer duration of vancomycin use, higher vancomycin trough levels, and higher rates of AKI. Higher vancomycin trough levels have been previously identified as a risk factor for nephrotoxicity [30], and thus, the longer duration of vancomycin use in the *E. faecium* group might have elevated the vancomycin trough levels leading to renal injury. Furthermore, glycopeptide use is associated with higher mortality in patients with *E. faecalis* bacteremia [31]. These findings suggest that vancomycin doses need to be considered carefully for patients with enterococcal bacteremia to prevent AKI.

Risk factors for mortality due to enterococcal bloodstream infections may include malignancy, admission to the intensive care unit, severity of illness, and high-level resistance to ampicillin and ciprofloxacin [1,15,18,32,33]. In this study, we identified AKI, unknown source of infection, previous corticosteroid treatment, and qSOFA score ≥ 2 as independent risk factors associated with mortality due to enterococcal bacteremia. These findings may be clinically plausible because impairment of kidney function and consequent multiorgan dysfunction syndrome, which likely leads to mortality, occurs in critically ill patients [34,35]. Moreover, severely ill patients, such as those with collagenase disease, nephrotic syndrome, and advanced cancer, receive corticosteroid therapy. However, corticosteroids not only decrease inflammation but also have side effects, including a reduction in the activity of the immune system, as well as hyperglycemia. A previous study reported that an unknown focus of bacteremia was associated with inappropriate antibiotic therapy and poor clinical outcome [36]. These findings suggest that the risk factors associated with mortality provide useful information for clinicians to avoid treatment failure, thereby enabling appropriate medical therapy after the onset of bacteremia. Specialized personnel, such as antimicrobial stewardship teams, can support clinical management through appropriate antibiotic use and early diagnosis.

This study has some limitations. First, we conducted a retrospective study in a single university hospital, and the data were gathered by reviewing electronic medical records, relying on other investigators for data collection; hence, a measurement bias could not be ignored. Second, this study investigated only *E. faecalis* and *E. faecium* bacteremia. To evaluate the characteristics of enterococci, we must assess the clinical outcomes of patients with bacteremia caused by *Enterococcus* species, including *E. avium*, *E. casseliflavus*, and *E. raffinosus*, other than *E. faecalis* and *E. faecium*.

#### **4. Materials and Methods**

#### *4.1. Setting and Patients*

We conducted an observational retrospective study between 1 January 2014, and 31 December 2018, at Kobe University Hospital. We defined nosocomial enterococcal bacteremia as positive blood cultures obtained after 48 h of hospitalization. We investigated all adult patients (age > 18 years) with only the first episode of at least one positive blood culture for *Enterococcus* species. In patients with two or more blood cultures for the same organism, only one was included in the analysis. Patient data were obtained from electronic medical records.

#### *4.2. Definitions*

The demographic information included age and sex. The clinical and microbiological data for each case were carefully reviewed. The hospitalization ward, comorbidities, the source of infection, qSOFA score, and use of invasive devices were evaluated on the day of bacteremia onset. Data pertaining to recent surgery, immunosuppressive treatment, and previous antibiotic therapy within 30 days prior to the first positive blood culture were collected. The polymicrobial culture was defined as the isolation of more than one organism, excluding contaminated pathogens, which were defined if the following pathogens (coagulase-negative staphylococci, *Bacillus* species, *Corynebacterium* species, *Propionibacterium* species, Viridans-group *streptococci*, and *Micrococcus* species) were detected in one of two or more blood culture sets on the same day. Antibiotic therapy against enterococci was defined as the prescription of antibiotics to which the isolated enterococci were susceptible to bacteremia. Source control with surgical drainage, time to the initiation of antibiotic therapy against enterococci, and total duration of antibiotic treatment was reviewed from the onset of bacteremia. Daily antimicrobial cost during antibiotic therapy for enterococcal bacteremia was calculated by multiplying the drug prices per dose by the total number of given doses and dividing the product by the total number of days of antibiotic therapy. All costs are shown in US dollars (\$; exchange rate, 1 \$ = 104.30 yen in 1 December 2020). We counted the number of patients whose vancomycin median serum trough concentration

was ≥20 mg/L. AKI was defined as an absolute increase in the serum creatinine level to >0.3 mg/dL or a >1.5-fold increase from the baseline value within 7 days. We defined the 30-day mortality as death due to any cause within 30 days after the onset of bacteremia.

#### *4.3. Identification and Antibiotic Susceptibility Testing*

The BACTEC FX system (Becton and Dickinson, Tokyo, Japan) was used to process the blood cultures. Enterococci were isolated according to standard microbiological procedures. The isolates were identified using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS; Bruker, Tokyo, Japan). The minimum inhibitory concentrations of ampicillin, vancomycin, imipenem, and levofloxacin were determined by a broth microdilution method using Microscan Walk Away 96plus (Beckman Coulter, Tokyo, Japan) and interpreted according to the breakpoints proposed by the Clinical and Laboratory Standards Institute guidelines. *Staphylococcus aureus* (ATCC 29213, VA, USA) was used as the quality control strain.

#### *4.4. Statistical Analysis*

Continuous variables are expressed as medians with interquartile ranges (IQRs) and categorical variables as frequency counts with percentages. Continuous variables were compared using the Mann–Whitney *U* test. Categorical variables were compared using the Chi-square test. Multivariate conditional logistic regression analysis of factors that were potentially associated with *E. faecium* acquisition and mortality included clinically important variables of the statistically significant variables in the univariate analysis. We assessed the in-hospital mortality by using Kaplan–Meier analysis and estimated HR and 95% CI using multivariate Cox proportional hazard regression models. All statistical analyses were performed using the statistical software EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan).

#### *4.5. Ethics Approval*

This study was approved by the Kobe University Graduate School of Health Sciences Institutional Review Board (approval no. 472-6) and was performed in accordance with the ethical standards of the Institutional Research Committee and the tenets of the Declaration of Helsinki (1964).

#### **5. Conclusions**

In summary, we conducted an observational retrospective study to compare the clinical features and outcomes of bacteremia caused by *E. faecalis* and *E. faecium.* Multivariate analysis was used to identify the risk factors for the acquisition of *E. faecium* bacteremia and mortality due to enterococcal bacteremia. This study demonstrated that the *E. faecium* group had a shorter survival period and higher antimicrobial cost than the *E. faecalis* group. *E. faecium* bacteremia occurred more frequently among patients treated with broad-spectrum antibiotics, especially patients with hematologic tumors and cholecystocholangitis. Furthermore, we identified that severe illness, which tends to be associated with a worse renal function without an infectious focus, was an independent risk factor for mortality due to enterococcal bacteremia. These findings suggest that clinicians should provide a rational treatment strategy supported by specialized teams, such as those assisting antimicrobial stewardship programs, even before the onset of bacteremia in hospitalized patients.

**Author Contributions:** Conceptualization, A.U.; methodology, A.U., K.O. (Kayo Osawa) and K.O. (Kenichiro Onuma); software, A.U.; validation, A.U.; formal analysis, A.U. and K.O. (Kayo Osawa); investigation, A.U.; resources, A.U. and K.O. (Kenichiro Onuma); data curation, A.U. and K.O. (Kayo Osawa); writing—original draft preparation, A.U. and K.S.; writing—review and editing, A.U., K.S., and I.Y.; visualization, A.U.; supervision, K.K., Y.Y., T.N., M.F., and I.Y.; project administration, T.M.; funding acquisition, T.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Informed Consent Statement:** The requirement of patient consent was waived owing to the retrospective nature of the study.

**Data Availability Statement:** Data sharing is not applicable to this article.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


#### *Article* **Antimicrobial Stewardship: Development and Pilot of an Organisational Peer-to-Peer Review Tool to Improve Service Provision in Line with National Guidance**

**Olaolu Oloyede <sup>1</sup> , Emma Cramp <sup>2</sup> and Diane Ashiru-Oredope 1,\***


**\*** Correspondence: diane.ashiru-oredope@phe.gov.uk

**Abstract:** Antimicrobial resistance continues to be a considerable threat to global public health due to the persistent inappropriate use of antibiotics. Antimicrobial stewardship (AMS) programs are essential in reducing the growth and spread of antibiotic resistance, in an environment which lacks incentives for the development of new antibiotics. Over the years, a variety of resources have been developed to strengthen antimicrobial stewardship. However, the differences in resources available present a challenge for organisations/teams to establish the best resources to utilise for service provision. A peer review tool was formulated using four national documents on AMS and tested through three phases with feedback. A survey method was used to collect feedback on the validity, feasibility, and impact of the AMS peer review tool. Feedback received was positive from the earlier pilots. The tool was found to be useful at identifying areas of good practice and gaps in antimicrobial stewardship across various pilot sites. Feedback suggests the tool is useful for promoting improvements to AMS programs and highlights that the content and features of the tool are appropriate for evaluating stewardship.

**Keywords:** AMS; antimicrobial resistance; antimicrobial stewardship intervention; PDSA cycle

#### **1. Introduction**

Antimicrobial stewardship (AMS) is an organisational or healthcare system-wide approach to improving and optimising antimicrobial therapy through the promotion and monitoring of the appropriate use of antimicrobials to prevent the development of resistance. Evidence suggests that a coordinated and comprehensive AMS programme is vital in tackling the emergence of antimicrobial resistance [1].

Several national AMS guidance and toolkits have been developed to support and encourage best practice in acute National Health Service (NHS) hospitals in England, as well as the goals outlined in the UK 5-year National Action Plan 2019–2024. These collections of resources, produced by different expert groups and at different time points, present a challenge for organisations/teams to establish the best resources to improve service provision. In response to this, the consolidation of recommendations from these national resources into one complete AMS tool can provide clarity to enable adherence to national guidance, thus consistent and better stewardship. The Health Foundation defines peer review as the professional assessment against standards of the organisation on healthcare processes and quality of work, to foster improvement [2]. The peer review tool developed aims to support hospitals to systematically review their processes for appropriate antimicrobial prescribing, stewardship and improving patient outcomes. The tool is intended for use at the host site with an external peer reviewer, to allow for an impartial assessment of AMS practices and development of an improvement plan [3].

Organisational peer-to-peer reviews offer an objective assessment to drive internal improvement through the evaluation of a provider by another organisation without the

**Citation:** Oloyede, O.; Cramp, E.; Ashiru-Oredope, D. Antimicrobial Stewardship: Development and Pilot of an Organisational Peer-to-Peer Review Tool to Improve Service Provision in Line with National Guidance. *Antibiotics* **2021**, *10*, 44. https://doi.org/10.3390/antibiotics 10010044

Received: 26 November 2020 Accepted: 25 December 2020 Published: 5 January 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

need for formal regulatory authority involvement. Cases of organisational peer-to-peer review are rare in healthcare; an example of this approach includes the UK National Chronic Obstructive Pulmonary Disease (COPD) Resources and Outcomes Project and the regional intervention to improve the hospital mortality associated with coronary artery bypass graft surgery (The Northern New England Cardiovascular Disease Study Group) [4].

To prevent the growing issue of antimicrobial resistance (AMR), NHS England (now NHS England & Improvement) launched the world's largest healthcare incentive scheme for hospitals and other health service providers. The programmed offered NHS Trusts incentive funding valued up to £150 million to support expert clinicians and pharmacist's assessment and reduction of inappropriate antimicrobial prescription [5]. The development of the AMS peer review tool with a plan-do-study-act (PDSA) cycle approach aims to support national investment in tackling AMR issues through organisational peer-to-peer reviews.

#### **2. Results**

#### PDSA Cycle 1:

In 2016, positive feedback was received from all participants from the East of England pilot using the first version of the tool [6]. The tool was found to be beneficial at identifying areas of good practice and gaps in antimicrobial stewardship at each pilot site, as well as presenting opportunities to learn from peers. The participants found that the tool was relatively easy to use and indicated peer review visits annually would be adequate.

The average length of time to undertake the peer review was five hours in total. These five hours where made up of approximately 2 h for reviewing necessary documents prior to visiting the host organisation and 3 h to conduct the site visit which included attendance at the AMS committee meeting and visiting a ward area to interview healthcare workers.

#### PDSA Cycle 2:

In 2018, following the presentation of the peer review tool to the national multidisciplinary group on antimicrobial resistance and utilisation, the English surveillance programme for antimicrobial utilisation and resistance (ESPAUR) group; the number of indicators in the tool was reduced from 101 to 37 following a two stage process and updated to include indicators from the current antimicrobial stewardship guidance and toolkits. Similar indicators were merged so that repetition was minimised, and themes were grouped together. Table 1 summarises the number of indicators from each stage of the toolkit development during cycle 2.

#### PDSA Cycle 3:

In 2019, feedback on the revised shortened tool from another pilot of five participating acute hospitals (three teaching and two non-teaching trusts) in two regions of England suggested a two-week lead time for submission of the hospitals' documented evidence of the AMS programme was appropriate and the documents shared were found to be "mostly relevant". It was also viewed that the tool could be beneficial in "promoting shared learning across the hospital stewardship programmes" and the peer review should be repeated every three years. One of the challenges highlighted was arranging a suitable date for the peer review on-site visit for both parties.

Despite the revision of the tool and reduction in the number of questions, the time taken to complete the peer review did not decrease from the initial pilot, with an average of 5 h (1.6 h to review the document before the site visit and 3.4 h for the on-site visit). The reason for this was because although the indicators were merged to form fewer questions for the reviewer, the themes required an in-depth review which did not change the overall amount of time needed for the review. In circumstances where the NHS Trust's AMS programme is satisfactory, the tool was considered time-consuming, but the outcome of the visit provided assurance for the AMS team. All participants agreed that the benefit from the tool included its application to reinforce good practice and benchmarking against peers.

The majority of respondents highlighted that all domains assessing the NHS Trust's antimicrobial stewardship programme were either "very relevant" or "relevant" with

the exception of the "Patient and Carers" domain in which some responded "neutral". The consensus was that the tool would be best used by healthcare professionals with an infection specialist background, as well as an excellent resource to promote shared learning across the hospital stewardship programme. Some of the planned actions by the host organisations following the peer review were:



**Table 1.** Number of indicators at each stage of the toolkit development.

\* Moved to the AMS management team/ASC.

#### **3. Discussion**

The development of the AMS peer review tool focussed on establishing a resource that amalgamates the variety of national guidance and tools on good antimicrobial stewardship into a single resource. Thus, providing a comprehensive and structured instrument to strengthen an NHS Trust antimicrobial stewardship programme. However, selecting a sample of acute NHS Trusts conveniently located within the same geography to pilot the tool at various stages of development proved to be quite challenging. With the variation across NHS Trusts with different processes, cultures, capacity, and attitudes on tackling antimicrobial resistance, there is a need to broaden the geographical spread sample that will provide a more detailed insight to the feasibility of the tool. Thus, it was intended that the tool would be piloted across at least two hospital trusts in all the NHS regions to have a representative sample in the development of the tool. However, having a broad and large sample of secondary care institutions (National Health Service Trusts) willing to participate in the pilot proved to be a significant challenge as coordinating the on-site visit between both organisations was dependent on the dates the antimicrobial stewardship committee meetings were due to take place, as well as the pharmacist's clinical commitments, and annual leave. During the pilot, it was highlighted that having a central team to coordinate organisation of on-site visit across the hospitals is important and beneficial; however, this strategy will require additional resources within each regional health system. To support widespread uptake of such measuring instruments and opportunity for regional or national perspective on the variation of stewardship programmes, implementation of future organisational peer-to-peer review tools could benefit from a central resource to coordinate dates and visits to reduce burden on individual teams.

The feedback from the pilot showed that participants found the tool to be a useful resource that encourages shared learning between peers and identifying gaps within their antimicrobial stewardship programme. However, the results of the pilots may be biased towards individuals with a keen specialist interest in improvement measuresand actively

involved in promoting and implementing good stewardship practices across the hospital. Furthermore, there is the potential for selection bias from the pilots, as there was an overrepresentation of pharmacists conducting the peer review (all reviews completed by pharmacists). Although it is hoped that through the feedback from the ESPAUR oversight group (which includes members from a wide range of backgrounds such as microbiology consultants, members of the dental profession, and nurses) who provided critical review of the tool, some of the selection bias may have been addressed. The use of the tool is not intended to be exclusive to a specific profession, but rather to be used by any healthcare professional with an antimicrobial stewarship/infection management background.

In addition, the evidence across various medical areas suggests that following a peer review, improvements to services can sometimes occur slowly with inconsistent outcomes. It has been suggested, to achieve better results that will deliver change and improved services, that a multidisciplinary peer review visit may be a more attractive mechanism as a collective and agreed strategy to implement the recommendations is likely to breed success [7]. Organisation-wide support on antimicrobial stewardship is considered crucial to addressing AMR, thus without this, minimal impact may be expected following visits.

The iterations of developing the tool demonstrated that the content and elements of the assessment tool are suitable for evaluating stewardship, thus providing a robust and systematic approach. The challenge from the outset was coordinating a system-wide simultaneous uptake within regions of the AMS peer review tool, with a proposed annual visit to reassess and support consistent improvement in stewardship programmes.

#### **4. Materials and Methods**

The AMS peer review tool was originally developed and piloted by East of England Antimicrobial Pharmacists Network in 2016 across eight NHS acute Trusts. The tool includes consolidated recommendations from a number of national AMS guidance and toolkits [8–11] into one easy-to-use document, assessing an organisation's stewardship programme on the following domains:


The intention was to develop a voluntary tool (available to download as a Supplementary Material) to strengthen antimicrobial stewardship in acute hospitals through the facilitation of organisational peer reviews within the regional health systems in England. The tool was tested through three phases with the feedback and outcome shared with the English surveillance programme for antimicrobial utilisation and resistance (ESPAUR) group, consisting of experts from various backgrounds (microbiologist, paediatricians, infection disease specialist, and pharmacists). Through the phases of developing the AMS peer review tool, a survey method was used to collect feedback on the validity, feasibility, and impact.

The findings of the pilot were first presented to ESPAUR oversight group in January 2017. The average length of time to perform a peer review was five hours (two hours to review key documents before the site visit and three hours to carry out the site visit which included attending an AMS committee meeting). Feedback from the ESPAUR members was to simplify the tool by reducing the number of indicators before the next piloting stage. The tool was further updated in January 2019 and validated using the checklist outlined by Pulcini et al. [12] in their publication on developing a global checklist for hospital AMS programmes. The updated version of the tool was shared via the AMR network leads with antimicrobial pharmacists across some acute hospitals in England to request their participation to pilot. Incorporated into the updated peer-review process was a pre-visit

stage, to encourage more focussed and productive on-site visits. Volunteering hospitals were paired within their region to avoid long-distance travel to conduct the peer-review.

The three iterations of testing and evaluating feedback were used to validate and refine the tool, with time to completion measured, and participants' experience of utilising the tool collected (Table 2).


**Table 2.** Summary of the plan-do-study-act (PDSA) cycles.

#### *Using the AMS Peer Review Tool*

It was recommended that those wishing to pilot the tool (available as supplementary material) consider choosing peer hospitals based on geographical location to minimise the amount of travel time, and that the whole review process may take one full working day to complete and that peer review process can be considered every two to three years (as this allows enough time to lapse to accrue the benefits of the peer review and appropriate time for reassessment), or more frequently if an improvement plan is implemented. The peer review may be carried out by an individual or team from an external organisation, which is not limited to the list below, and may include one or more of the following:


The AMS peer review tool outlines key aspects across the six AMS domains mentioned earlier to be critically reviewed and assessed by the host organisation and peer reviewer. The recommended process for conducting the peer review is outlined below:

Step 1

Plan and schedule the onsite visit to occur ideally on the day the Antimicrobial Stewardship Committee (ASC) is held. This would allow the reviewer to witness firsthand the attendance, management, and leadership at the meetings. The host organisation should schedule an opportunity for the peer review team to meet with senior clinicians and managers. The date and time of visit should be scheduled during less busy periods and where possible consider staff availability due to annual leave.

Step 2

Host organisation to prepare documents listed in the tool for submission to peer reviewer two weeks ahead of scheduled visit and self-assess AMS prior to peer reviewer visit.

Peer reviewer to review documents submitted by host organisation and prepare approach prior to visit.

Step 3

During the onsite visit, the reviewers should speak with clinical staff on the ward. In addition, where it is considered necessary or additional benefit for the review process, reviewers may also consider having discussions with senior clinicians, and managers including the medical director, lead for infection prevention and control (e.g., the director for IPC in the UK (DIPC)), director of nursing, microbiologist lead for AMS and chief pharmacist within the NHS Trust, the director of nursing and quality, AMR/AMS lead pharmacist and chief pharmacist within the CCG, and the system AMR lead.

Step 4

Peer review report to be submitted within the agreed time frame at the onsite visit. The report should outline the areas of success and opportunities for improvement.

#### **5. Conclusions**

The pilots were important in assessing the feasibility of the tool and outlining the barriers for use. The feedback from participants and expert group suggest that the tool is best used where a gap or issue has been identified within a hospital stewardship programme, with the tool providing a comprehensive review to help develop strategies for improvements. The tool also presents an opportunity for regional antimicrobial groups to voluntarily lead on coordinating the visits to share best practices amongst hospital within the region. Overall, this quality improvement project showed there was a need for tools that support organisational peer-to-peer review. However, future work on developing such a quality improvement model need to build in considerations that will reduce the time burden and pressure for the reviewers.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2079-6382/ 10/1/44/s1, Antimicrobial Stewardship (AMS) Peer Review Tool as PDF and editable excel document.

**Author Contributions:** Conceptualization, E.C. and, D.A.-O.; data curation, O.O.; formal analysis, O.O.; methodology, O.O., E.C., and, D.A.-O.; project administration, O.O.; resources, D.A.-O.; supervision, D.A.-O.; writing—original draft, O.O.; writing—review and editing, E.C. and, D.A.-O. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project received no external funding.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data is contained within the article or supplementary material.

**Acknowledgments:** All participating hospitals and infection leads are acknowledged for their support in piloting the tool. In particular, the authors would like to acknowledge the East of England Antimicrobial Pharmacy Network for their initial support with shaping the original version of the peer review tool and piloting it across the region.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


*Article*

#### **Evaluation of an Antimicrobial Stewardship Program for Wound and Burn Care in Three Hospitals in Nepal**

**Varidhi Nauriyal 1,\* , Shankar Man Rai <sup>2</sup> , Rajesh Dhoj Joshi <sup>3</sup> , Buddhi Bahadur Thapa <sup>4</sup> , Linda Kaljee 5,\*, Tyler Prentiss <sup>5</sup> , Gina Maki <sup>1</sup> , Basudha Shrestha <sup>3</sup> , Deepak C. Bajracharya <sup>6</sup> , Kshitij Karki <sup>6</sup> , Nilesh Joshi <sup>6</sup> , Arjun Acharya <sup>4</sup> , Laxman Banstola <sup>4</sup> , Suresh Raj Poudel <sup>4</sup> , Anip Joshi <sup>4</sup> , Abhinav Dahal <sup>3</sup> , Niranjan Palikhe <sup>3</sup> , Sachin Khadka <sup>3</sup> , Piyush Giri <sup>2</sup> , Apar Lamichhane <sup>2</sup> and Marcus Zervos <sup>1</sup>**


Received: 9 November 2020; Accepted: 12 December 2020; Published: 16 December 2020

**Abstract:** Antimicrobial stewardship (AMS) programs can decrease non-optimal use of antibiotics in hospital settings. There are limited data on AMS programs in burn and chronic wound centers in low- and middle-income countries (LMIC). A post-prescription review and feedback (PPRF) program was implemented in three hospitals in Nepal with a focus on wound and burn care. A total of 241 baseline and 236 post-intervention patient chart data were collected from three hospitals. There was a significant decrease in utilizing days of therapy per 1000 patient days (DOT/1000 PD) of penicillin (*p* = 0.02), aminoglycoside (*p* < 0.001), and cephalosporin (*p* = 0.04). Increases in DOT/1000 PD at post-intervention were significant for metronidazole (*p* < 0.001), quinolone (*p* = 0.01), and other antibiotics (*p* < 0.001). Changes in use of antibiotics varied across hospitals, e.g., cephalosporin use decreased significantly at Kirtipur Hospital (*p* < 0.001) and Pokhara Academy of Health Sciences (*p* = 0.02), but not at Kathmandu Model Hospital (*p* = 0.59). An independent review conducted by infectious disease specialists at the Henry Ford Health System revealed significant changes in antibiotic prescribing practices both overall and by hospital. There was a decrease in mean number of intravenous antibiotic days between baseline (10.1 (SD 8.8)) and post-intervention (8.8 (SD 6.5)) (*t* = 3.56; *p* < 0.001), but no difference for oral antibiotics. Compared to baseline, over the 6-month post-intervention period, we found an increase in justified use of antibiotics (*p* < 0.001), de-escalation (*p* < 0.001), accurate documentation (*p* < 0.001), and adherence to the study antibiotic prescribing guidelines at 72 h (*p* < 0.001) and after diagnoses (*p* < 0.001). The evaluation data presented provide evidence that PPRF training and program implementation can contribute to hospital-based antibiotic stewardship for wound and burn care in Nepal.

**Keywords:** antibiotic resistance; stewardship; wound care; burn care; Nepal

#### **1. Introduction**

Antimicrobial resistance (AMR) has been recognized as a complex global health challenge lacking a universal solution. In 2015, the World Health Organization (WHO) released a global action plan (GAP) on AMR which provides a framework for developing national action plans on a country-by-country basis. The burden of health care associated infection is higher in low- and middle-income countries (LMIC) compared to higher income countries [1]. Inappropriate antimicrobial prescribing practices, lack of adequate antibiotic tracking systems, and limited healthcare funding to facilitate surveillance and laboratory infrastructure are some of the factors that have contributed to rising antimicrobial resistance in LMIC. There is evidence that antimicrobial stewardship (AMS) interventions are effective in increasing compliance with antibiotic policy, reducing duration of antibiotic treatment and potentially reducing hospital length of stay [2].

Post-prescription review and feedback (PPRF) programs include expert review of antibiotic prescribing decisions and feedback to the attending physician [3]. PPRF programs have been shown to be effective in U.S. hospitals, and on the basis of four studies reviewed by Dijck et al., there was a decrease in antibiotic days noted with audit and feedback in LMIC [4]. In addition, comparison of baseline and post-intervention data of a PPRF program in medical, obstetrics and gynecology, and general surgery wards at Kathmandu Model Hospital indicated decreased days of therapy per 1000 patient-days for courses of aminoglycoside and cephalosporin, increased justified use of antibiotics, de-escalation, and rational use of antibiotics [5].

To date, there are very limited data on the potential impact of AMS programs on antibiotic prescribing practices in burn and wound care centers in LMIC. Annually, an estimated 5 million deaths occur in LMIC due to injuries, with 10 to 50 times more individuals living with associated permanent disabilities [6]. Fire-related burns alone account for about 300,000 deaths annually, 95% percent of those occurring in LMIC. The highest rates of burns occur in Asia. In rural Nepal, burns are the second most common injury, accounting for 5% of disabilities. Overall, burns are the third most common injury after fall and road traffic accidents in the country [7]. Lack of a national burn registry makes estimating burden of disease difficult, however, a recent systematic review suggested the average hospital stay among burn victims ranged from 13 to 60 days in Nepal, with mortality estimates of 4.5 to 23.5% [8].

Studies have estimated infection-related mortality in burn victims to range from 40 to 60% [9–11]. Within LMIC, chronic wounds are commonly colonized and infected with antibiotic-resistant bacteria both during hospitalization and after discharge [12,13]. Inadequate hospital and clinic infection control protocols, delay in treatment, and use of self-treatment are some of the factors contributing to prevalence of infection in this population [14,15]. A primary concern is the increasing prevalence of antibiotic resistance in LMIC [16–18].

High morbidity and mortality associated with wounds and burns in Nepal and other LMIC has incentivized an urgent need to develop infection prevention practices and improve treatment. The aim of this study was to implement and evaluate the role of a hospital-based AMS program to support optimal antibiotic use for wounds and burns and in the longer-term decrease risks of infection from resistant pathogens.

#### **2. Results**

#### *2.1. Patient Chart Data: Demographics*

A total of 241 baseline and 236 post-intervention patient chart data were collected from the three study hospitals. At both baseline and post-intervention, a majority of patients were male and the mean age was less than 40 years. Number of study patients were evenly distributed across the three hospitals at both baseline and post-intervention. At post-intervention, there were more patient charts from the burn and less from the plastic and reconstructive surgery wards compared to baseline (*p* < 0.001). Length of stay decreased significantly within the burn unit (*p* < 0.001), as well as overall (*p* = 0.006) (see Table 1). There were no reported deaths among study patients.


**Table 1.** Baseline and post-intervention patient characteristics, length of hospital stay, and distribution across hospital sites and wards.

#### *2.2. Patient Chart Data: Antibiotic Use at Baseline and Post-Intervention*

Overall, there was a decrease in mean number of intravenous (IV) antibiotic days between baseline (10.1 (SD 8.8)) and post-intervention (8.8 (SD 6.5)) (*t* = 3.56; *p* < 0.001). There was no significant change for mean number of oral (PO) antibiotic days between baseline (4.2 (SD 3.3)) and post-intervention (3.7 (SD 3.5)) (*t* = 0.66; *p* = 0.510).

Across the three sites, there was a significant decrease in mean days of therapy between baseline and post-intervention for both aminoglycoside (6.1 (SD 4.3) vs 4.6 (SD 2.1)) (*t* = 2.08, *p* = 0.04) and cephalosporin (6.0 (SD 6.1) vs. 4.1 (SD 3.4)) (*t* = 3.54, *p* < 0.001). There were no significant changes in mean days of therapy for quinolines, penicillin, metronidazole, and other prescribed antibiotics.

Utilizing days of therapy per 1000 patient days (DOT/1000 patient-days (PD)) for data across the three study sites, we found no change in administering antibiotics either IV (*p* = 0.67) or PO (*p* = 0.09). There was a significant decrease in use of penicillin (*p* = 0.02), aminoglycoside (*p* < 0.001), and cephalosporin (*p* = 0.04). Increases in DOT/1000 PD at post-intervention were significant for metronidazole (*p* < 0.001), quinolone (*p* = 0.01), and other antibiotics (*p* < 0.001) (Figure 1 and Table 2).

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 4 of 12

**Figure 1.** Changes in days of therapy (DOT) per 1000 patient-days (PD) between baseline and postintervention in three hospitals in Nepal (Kirtipur, Kathmandu Model, and Pokhara hospitals). PCN = penicillin; MTZ = metronidazole; FQ: fluoroquinolone; AG: aminoglycosides; CPN: cephalosporin. **Figure 1.** Changes in days of therapy (DOT) per 1000 patient-days (PD) between baseline and post-intervention in three hospitals in Nepal (Kirtipur, Kathmandu Model, and Pokhara hospitals). PCN = penicillin; MTZ = metronidazole; FQ: fluoroquinolone; AG: aminoglycosides; CPN: cephalosporin.

**Table 2.** Days of therapy (DOT) per 1000 patient-days (PD) of prescribed antibiotics at baseline and post-intervention periods total and by study sites (Kirtipur, Kathmandu Model, and Pokhara

hospitals).

TOTAL SITES

KIRTIPUR

KATHMANDU MODEL

POKHARA

**Site Antibiotic Delivery &** 

**Class** 

**Baseline** 

Oral antibiotics 101 (46) 75 (31) 0.09 Penicillin 301 (91) 241 (70) 0.02 Cephalosporin 525 (167) 454 (167) 0.04 Metronidazole 75 (30) 160 (56) <0.001 Quinolone 46 (17) 72 (24) 0.01 Aminoglycoside 266 (84) 117 (39) <0.001 Other course 57 (16) 177 (53) <0.001

Intravenous antibiotics 292 (70) 304 (77) 0.56 Oral antibiotics 64 (27) 37 (10) 0.004 Penicillin 49 (15) 61 (19) 0.30 Cephalosporin 264 (71) 228 (71) <0.001 Metronidazole 10 (4) 30 (10) 0.002 Quinolone 9 (30) 14 (5) 0.40 Aminoglycoside 8 (40 16 (4) 0.15 Other course 18 (5) 16 (4) 0.73

Intravenous antibiotics 289 (70) 354 (72) 0.80 Oral antibiotics 29 (17) 32 (18) 0.80 Penicillin 63 (18) 54 (17) 0.04 Cephalosporin 125 (63) 144 (61) 0.59 Metronidazole 42 (18) 112 (42) <0.001 Quinolone 37 (14) 24 (9) 0.02 Aminoglycoside 34 (21) 40 (16) 0.90 Other course 20 (6) 16 (9) 0.23

Intravenous antibiotics 584 (82) 436 (77) 0.12 Oral antibiotics 8 (2) 5 (3) 0.37 Penicillin 189 (58) 127 (34) 0.22

**Post-Intervention DOT/1000 PD (N)** 

*p***-Value** 


**Table 2.** Days of therapy (DOT) per 1000 patient-days (PD) of prescribed antibiotics at baseline and post-intervention periods total and by study sites (Kirtipur, Kathmandu Model, and Pokhara hospitals).

Looking at these data by site, we found no change in IV administration of antibiotics, but there was a decrease in administering PO antibiotics at Kirtipur Hospital (*p* = 0.004). Moreover, at Kirtipur Hospital, there was a significant decrease in use of cephalosporin (*p* < 0.001) but an increase in use of metronidazole (*p* = 0.002). At Kathmandu Model Hospital, there were significant decreases in use of penicillin (*p* = 0.04) and quinolones (*p* = 0.02), but a significant increase in use of metronidazole (*p* < 0.001). At Pokhara Academy of Health Science, there was a decrease in both cephalosporin (*p* = 0.02) and aminoglycoside (*p* < 0.001), but increases in quinolones (*p* < 0.001) and other antibiotics (*p* < 0.001). (Table 2).

*Antibiotics* **2020**, *9*, 914

An independent review conducted by infectious disease specialists at the Henry Ford Health System revealed significant changes in antibiotic prescribing practices both overall and by hospital. Over the 6-month post-intervention period, there was a noted increase in justified use of antibiotics, de-escalation, accurate documentation, and adherence to the study antibiotic prescribing guidelines at 72 h and after diagnoses (definitive) (Table 3).


**Table 3.** Justification, de-escalation, treatment rationale, and fidelity to guidelines at baseline and post-intervention by total sites and individual hospitals (Kirtipur, Kathmandu Model, and Pokhara).

Physician champions recorded information on recommendations made during the post-intervention period. Across the three study sites, there were a total of 249 logbook entries with 71 recommendations (28.5%). Among the recommendations, there were 53 cases (74.6%) in which the physician champion recommended a change in the antibiotic and 18 cases (25.4%) in which the recommendation was to stop antibiotics. Overall, 41/71 (57.7%) recommendations were followed by the prescribing physician. Among 47 entries with the reason listed for the recommendation, 25 cases (53.2%) were related to obtaining data on resistance/sensitivity patterns, 6 cases (12.8%) due to no definitive evidence of infection, and 5 cases (10.6%) were related to extended duration of antibiotic use. Other reasons included use of multiple antibiotics, IV to oral conversion, patient symptoms, and change from a broader to a narrower spectrum antibiotic.

#### **3. Discussion**

The World Health Organization Global Action Plan (GAP) for AMR includes five strategic objectives that must be addressed to decrease pathogen resistance to available pharmaceutical therapeutics. These objectives include to (1) increase awareness and understanding of AMR, (2) strengthen knowledge through surveillance and research, (3) reduce the incidence of infection, (4) optimize the use of antimicrobial medicines, and (5) ensure sustainable investment in countering antimicrobial resistance [18]. Over the past 5 years, the partnership between the Henry Ford Health System, Nepali private, public, and non-profit hospital systems, and the Group for Technical Assistance in Kathmandu has supported AMR stewardship education and programs. The data presented in this paper represent an important step for implementation of hospital-based stewardship programs with an emphasis on the urgent need in LMIC to reduce risks of infection in wound and burn care. The hospitals selected included the non-profit and government health sectors. Both Kathmandu Model and Kirtipur hospitals were part of a previous post-prescription review and feedback (PPRF) project, and all three hospitals remain connected to ongoing education and training through a new web-based program Global Learning in Antimicrobial Resistance (GLAMR).

The data presented provide evidence that PPRF training and program implementation can contribute to hospital-based stewardship in Nepal. Across all three hospitals, there is a clear indication that prescribing practices at post-intervention were more likely justified and followed antibiotic prescribing guidelines. Utilizing DOT/1000 PD analytics, we found that across the three hospitals there were significant decreases in the use of penicillin, cephalosporins, and aminoglycosides. Within each individual hospital, there was some variation in prescribing practices, however, there was evidence of decreased use of penicillin, aminoglycosides, quinolones, and cephalosporins. Variations may be attributable to differences in prescribing practices at baseline within the various study wards. A recent study published on bacteriological profile of burn wound infections in Nepal suggested predominance of resistant Gram-negative organisms such as *Acinetobacter* spp., *Pseudomonas* spp., and *Enterobacter* spp. [19]. The decrease in use of penicillin, cephalosporins, and quinolones is hence an encouraging signal towards recognition of bacterial epidemiology and appropriate use of antimicrobials. While there was an increase in metronidazole use at post-intervention, the overall increase in "justified use" of antibiotics in our post-intervention group indicates an overall improvement in prescribing practices. The use of metronidazole with another agent would have been deemed "unjustified".

The changes across three hospitals within two separate locations in Nepal (Kathmandu and Pokhara) indicate that PPRF programs can be successfully implemented under different hospital administration and supports further dissemination as a part of hospital-based stewardship elsewhere in Nepal. Our evaluation of the previous implementation of the PPRF program at Kathmandu Model Hospital indicated decreases in use of cephalosporins and aminoglycosides. While there was no further decrease in these two antibiotics in the current study at Kathmandu Model Hospital, this may have been due to less use of these antibiotics at baseline.

The review of the prescribing practices shows significant improvements across all three study sites in terms of justified use, following guidelines, and documentation. Therefore, even as use of some antibiotics increased, these data suggest that prescribing practices at post-intervention were more likely appropriate to the diagnosis in terms of type and duration. Furthermore, the logbook data indicate that physician champions were actively reviewing patient charts within their wards and making recommendations. More than half of those recommendations were followed by the prescribing physician, with a majority of those recommendations due to information on pathogen resistance/sensitivity, lack of evidence of infection, and long duration of use of a single antibiotic.

The study strengths include the potential for introducing antimicrobial stewardship programs in low-resource hospital settings with comprehensive training and locally salient antibiotic prescribing guidelines. In addition, non-infectious disease physicians were successfully trained as physician champions and supported prescribing changes within their wards. Utilizing the existing training materials and guidelines reflective of potential regional differences in resistance and availability of antibiotics, the program can be duplicated elsewhere in Nepal. There are some limitations to the study. Data collection was dependent on manual review of handwritten notes given the lack of electronic medical record in Nepal. "Other courses" is an un-identified pool of antibiotics since the focus was on obtaining usage data on the most commonly prescribed antibiotics. Physician champions had to manually fill out log-books, which added to their workload, and hence not all data points were consistently entered. Out of a total of 249 log-book entries, only 47 outlined rationale for recommendations. The study length did not provide time to determine if there were any changes in resistance levels among pathogens.

#### **4. Materials and Methods**

The study was part of a larger AMR and AMS collaboration between the Henry Ford Health System Division of Infectious Diseases and Global Health Initiative (Detroit, MI, USA); the Group for Technical Assistance (Kathmandu, Nepal); and various non-profit, public, and private hospitals in Nepal. PPRF programs include expert review of antibiotic prescribing decisions and feedback to the prescribing physician. One strength of the current study is the focus of the PPRF program on wound and burn care. In many LMIC, there are inadequate numbers of infectious disease specialists, and therefore a key element of the adapted PPRF program in Nepal includes training "physician champions" in AMR and AMS. More details about the adaptation of the PPRF intervention for use in Nepal is described in detail elsewhere [5]. Variations for the current study included revisions to the antibiotic prescribing guidebook to include information on wound and burn care and a training-of-trainers approach, whereby the AMR "physician champions" were both trained in the PPRF program and provided with information and tools to train other healthcare providers within their wards. The antibiotic prescribing guidebook included 5 sections: (1) empiric guidelines, (2) suggested definitive guidelines with options depending on susceptabilities, (3) suggested duration of antibiotic therapy based on indications, (4) intravenous to oral conversions, and (5) renal dosing (Table 4). A total of 52 healthcare providers and hospital administrators including 6 physician champions from 3 study sites (Kathmandu Model, Kirtipur, and Pokhara hospitals) were trained over 2 days (11–12 July 2018).


**Table 4.** Examples of empiric, definitive, and duration antibiotic prescribing guidelines.

#### *4.1. Study Sites Antibiotics* **2020**, *9*, x FOR PEER REVIEW 9 of 12

The research took place in 2 hospitals in Kathmandu (Kathmandu Model Hospital and Kirtipur Hospital) and 1 hospital in Pokhara (Pokhara Academy of Health Science). Kathmandu is located in central Nepal and Pokhara is further toward the western region (Figure 2). The research took place in 2 hospitals in Kathmandu (Kathmandu Model Hospital and Kirtipur Hospital) and 1 hospital in Pokhara (Pokhara Academy of Health Science). Kathmandu is located in central Nepal and Pokhara is further toward the western region (Figure 2).

**Figure 2.** Map of major cities in Nepal. **Figure 2.** Map of major cities in Nepal.

Both Kathmandu Model and Kirtipur hospitals are a part of a larger non-profit health organization, the Public Health Concern Trust (Phect, Nepal). Kathmandu Model is a 125-bed hospital that opened in 1993. Kathmandu Model Hospital provides a range of in-patient and outpatient services. The current study was focused in the general and specialized surgical wards. Kirtipur Hospital is a 100-bed hospital with additional specialized services including a 24 h emergency department. Kirtipur Hospital has a reconstructive surgery ward and the only burn intensive care unit in Nepal, which were the study sites for the current project. Kirtipur Hospital receives burn patients from throughout the country, and is part of the Resurge International Surgical Outreach Program that provides training and support to local hospitals and surgeons engaged in reconstructive and burn-related surgeries. Pokhara Academy of Health Science is a government facility and the second largest hospital in Nepal. The 500-bed hospital offers a broad range of services Both Kathmandu Model and Kirtipur hospitals are a part of a larger non-profit health organization, the Public Health Concern Trust (Phect, Nepal). Kathmandu Model is a 125-bed hospital that opened in 1993. Kathmandu Model Hospital provides a range of in-patient and out-patient services. The current study was focused in the general and specialized surgical wards. Kirtipur Hospital is a 100-bed hospital with additional specialized services including a 24 h emergency department. Kirtipur Hospital has a reconstructive surgery ward and the only burn intensive care unit in Nepal, which were the study sites for the current project. Kirtipur Hospital receives burn patients from throughout the country, and is part of the Resurge International Surgical Outreach Program that provides training and support to local hospitals and surgeons engaged in reconstructive and burn-related surgeries. Pokhara Academy of Health Science is a government facility and the second largest hospital in Nepal. The 500-bed hospital offers a broad range of services including a trauma center, burn unit, and surgical ward. These 3 wards were the study sites for the current project.

#### current project. *4.2. Study Population*

*4.2. Study Population*  The 3 study hospitals provide services to a range of socio-economic groups from both urban and rural areas in Nepal. Eligibility criteria for the patient chart review evaluation data included (1) The 3 study hospitals provide services to a range of socio-economic groups from both urban and rural areas in Nepal. Eligibility criteria for the patient chart review evaluation data included (1) inpatient within the study wards, (2) aged 15 + years, and (3) prescribed antibiotics for at least 72 h within the hospital.

including a trauma center, burn unit, and surgical ward. These 3 wards were the study sites for the

#### inpatient within the study wards, (2) aged 15 + years, and (3) prescribed antibiotics for at least 72 h within the hospital. *4.3. Data Collection*

#### *4.3. Data Collection*  4.3.1. Patient Chart Data

located in Kathmandu.

4.3.1. Patient Chart Data Eligible patient chart data were collected for 6 months baseline (pre-PPRF training) from January 2018 to June 2018 and 6 months post-intervention between August 2018 and January 2019. The gap month (July 2018) was the implementation of the intervention. Data collection was coordinated and Eligible patient chart data were collected for 6 months baseline (pre-PPRF training) from January 2018 to June 2018 and 6 months post-intervention between August 2018 and January 2019. The gap month (July 2018) was the implementation of the intervention. Data collection was coordinated and conducted by trained staff at a local nongovernmental agency (Group for Technical Assistance) located in Kathmandu.

Sample size was based on published data on duration of injectable drug use from Nepali

conducted by trained staff at a local nongovernmental agency (Group for Technical Assistance)

Sample size was based on published data on duration of injectable drug use from Nepali hospitals [20]. Using a two-sided comparison of a continuous variable (days of therapy per 1000 study patient days, DOT/1000PD), we calculated that a sample size of 211 patients per group (baseline and post-intervention) was necessary to detect a difference of 20% between time periods (90% CI, α = 05).

Patient chart data included: (1) demographics (gender and age), (2) hospital/ward, (3) length of stay, (4) source of infection, (5) patient height/weight, (6) conditions present at study enrollment, (7) systemic antibiotic use during prior 72 h, (8) origin of onset of infection, (9) working and final diagnosis, (10) systemic antibiotic use throughout the hospital stay period, (11) therapy prescribed at discharge, (12) infection-related complications, (13) factors associated with persistent infection, and (14) disposition at end of hospital stay (if deceased, date and cause).

After data collection, patient chart data were reviewed by infectious disease specialists at the Henry Ford Health System to determine whether prescribed antibiotics were justified. Justification was determined by diagnosis, pathogen (when available), duration, and route (IV or PO) as described within the antibiotic prescribing guidelines (Table 4). Reviews included both initial therapy and therapy changes after recommendations by the physician champions.

#### 4.3.2. Physician Logbook Data

Physician champions were provided with antibiotic prescribing guidelines, which included a logbook. Through the logbooks, physicians documented chart reviews, recommendations made, and acceptance of recommendations by the prescribing physician. Recommendations were made verbally to the prescribing physician and/or as written notes. Logbooks were collected on a monthly basis to ensure that they were completed as required by the evaluation protocol.

#### *4.4. Data Management and Analysis*

Patient chart data were entered into REDCap (Research Electronic Data Capture) [21] by trained staff at the Group for Technical Assistance in Nepal. REDCap is a secure web application for building and managing online databases. REDCap allowed immediate access to the data both at the Henry Ford Health System and the project data team in Nepal. Data were reviewed and cleaned, which included deleting 2 cases that were under 15 years and 4 cases that were collected after the end of the 6-month post-intervention period. Continuous data were described using means and standard deviations, and univariate two group comparisons used independent two-group *t*-tests to assess significance. Categorical data were described using counts and percentages, and chi-squared tests were used to assess significance. Days of therapy DOT/1000 PD was calculated at baseline and intervention periods for IV and PO delivery and specific antibiotics. Days of therapy was calculated as 1921 at baseline (*N* = 241) and 1520 (*N* = 236) at post-intervention across the 3 sites. Days of therapy at baseline and post-intervention were calculated by site at 321 and 390 (Kathmandu Model), 358 and 365 (Kirtipur), and 592 and 449 (Pokhara), respectively. The proportion of DOT/1000 PD was compared between baseline and intervention time points using tests of proportion and Fisher's exact test to determine significance. Statistical significance was set at *p* < 0.05.

Logbooks were collected by the project data team on a monthly basis and scanned. Scanned logbook data were sent to the evaluation team at the Henry Ford Health System for review. Scanned data were entered into Excel and analyzed using descriptive statistics. All statistical analysis was performed using SPSS 25.0 (Version 25.0. Armonk, NY, USA).

#### *4.5. Ethical Review*

The study was approved by the Institutional Review Board at the Henry Ford Health System, Detroit, MI (#11732), and the Nepal Health Research Council, Kathmandu, Nepal (#1523).

#### **5. Conclusions**

This study demonstrates successful implementation of PPRF as an antimicrobial stewardship tool at burn and wound in-patient centers in Nepal. There is an encouraging trend towards change in antimicrobial prescribing practice with more thoughtful and justified use.

**Author Contributions:** Conceptualization, S.M.R., L.K., and M.Z.; data curation, V.N., T.P., and B.S.; formal analysis, V.N. and L.K.; funding acquisition, L.K. and T.P.; investigation, T.P., G.M., D.C.B., K.K., L.B., S.R.P., A.J., A.D., N.P., S.K., P.G., and A.L.; methodology, L.K.; project administration, V.N., S.M.R., R.D.J., T.P., D.C.B., K.K., N.J., and A.A.; supervision, S.M.R., R.D.J., B.B.T., G.M., A.A., and M.Z.; writing—original draft, V.N. and L.K.; writing—review and editing, G.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the MERCK INVESTIGATOR STUDIES PROGRAM (MISP), grant number 57055.

**Acknowledgments:** We would like to thank the physician champions from Kathmandu Model and Kirtipur hospitals and the Pokhara Academy of Health Science for their commitment to the intervention and the research. We would also like to thank the support staff at the Group for Technical Assistance (Kathmandu) who contributed significantly to the organization of the physician training program logistics and day-to-day tasks associated with the program.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

#### **Feasibility and Validity of a Framework for Antimicrobial Stewardship in General Practice: Key Stakeholder Interviews**

#### **Lesley A. Hawes 1,2,\* , Jaclyn Bishop 2,3,4 , Kirsty Buising 2,5,6 and Danielle Mazza 1,2**


Received: 26 October 2020; Accepted: 10 December 2020; Published: 13 December 2020

**Abstract:** There is little guidance about developing systems for antimicrobial stewardship (AMS) for general practice. A literature review identified six key components: governance, monitoring of antibiotic prescribing and resistance with feedback to prescribers, consultation support, education of the public and general practitioners, pharmacist and nurse involvement, and research, which were incorporated into a potential framework for the general practice context. Objectives: to determine the feasibility and validity of the proposed AMS framework. A secondary objective was to identify likely bodies responsible for implementation in Australia. We undertook interviews with 12 key stakeholders from government, research, and professional groups. Data were analysed with a thematic approach. The framework was considered valid and feasible. No clear organisation was identified to lead AMS implementation in general practice. The current volume-based antibiotic prescription monitoring system was considered insufficient. AMS education for the public, further development of GP education, and improved consultation support were strongly recommended. The role of community-based pharmacists and nurses is largely unexplored, but their involvement was recommended. A clear leader to drive AMS in general practice is essential for an action framework to gain traction. Monitoring and feedback of antibiotic prescribing require urgent development to include monitoring of prescribing appropriateness and patient outcomes.

**Keywords:** antimicrobial stewardship; general practice; family practice; antibiotic; health policy; quality of health care; antibiotics; public health; pharmacist; nurse

#### **1. Introduction**

Antimicrobial resistance is a global problem with a major impact on health care and associated costs [1]. Exposure of microbes to antimicrobials contributes to the problem [2,3]; unnecessary use of antimicrobials must be minimised. The consumption of antibiotics in the Australian community is high in comparison with similar countries [4], with most antibiotics prescribed by general practitioners [4]. There is a high rate of prescribing of moderate- (66% of use) and broad-spectrum antibiotics (25%) [5], and inappropriate use is still common for conditions such as upper respiratory tract infections [4]. For these conditions, antibiotics are prescribed at rates 4–9 times that recommended by the Australian national antibiotic prescribing guidelines Therapeutic Guidelines—Antibiotic [6]. Australia's National Antimicrobial Resistance Strategy calls for the introduction of antimicrobial stewardship (AMS) to address inappropriate antibiotic prescribing [7]. However, there is little guidance for how to implement AMS across Australian general practice.

Through a review of international health system approaches to AMS in general practice [8–23], a potential framework to guide AMS in general practice was formulated. This framework contains six key components: governance, monitoring of antimicrobial resistance and prescribing with feedback to GPs, education for general practitioners (GPs) and the public, consultation support, the involvement of community-based pharmacists and nurses, and research [24]. Details of the framework are provided in Appendix A.

The aim of this study was to interview key stakeholders to determine the likely feasibility and validity of the proposed AMS framework and a secondary aim was to identify any existing organisations who may take on responsibility for implementation in Australia.

#### **2. Results**

Of the 24 invited stakeholders, 13 accepted. Two declined, another was on extended leave, and eight did not respond to two emails. One of those who declined—despite being invited to participate as an expert, not as a representative—replied, "[name of organisation] is not in the best place to help with your query regarding AMS in general practice and we recommend you contact [another named organisation]." We already had stakeholders from the organisation recommended. One respondent accepted but could not be interviewed in the timeframe. The 12 interviewed stakeholders' background, relevant expertise, and location are outlined in Table 1. The COREQ checklist is available in Supplementary Table S1.

**Table 1.** The professional background, antimicrobial stewardship (AMS) involvement, and location of the 12 interviewed stakeholders.


Overall, stakeholders reported that the proposed AMS framework for general practice and its components were feasible and valid; and that it provided a link between the objectives of Australia's National AMR Strategy and action. However, most stakeholders highlighted that it would require leadership and prioritisation for implementation to have the desired impact. Importantly, the stakeholders had difficulty nominating the best organisation to oversee this implementation. (Representative quotes are supplied; additional quotes are available in Supplementary Table S2).

*It seems very comprehensive to me* . . . *able to be implemented* . . . *I think we need to have an agreed upon governance structure and agreed upon priorities* . . . *I don't think there is one clear person or group who is responsible for the whole caboodle of this.* (Participant (P) 6)

Asked how they would define success, stakeholders nominated short- and long-term goals. Short-term goals were increased adherence to prescribing guidelines and improved patient outcomes with no increase in harm. Stakeholders also commented that increased professional support provided by such a framework may lead to improved professional satisfaction for GPs. The long-term goals that they stated were a decrease, or at least, no increase in antimicrobial resistance (AMR).

Governance was reported by the stakeholders to be important to set strategic priorities and harmonize approaches. The importance of aligning work in primary care with work in other sectors was highlighted. A national action plan for AMS in general practice was regarded as a Commonwealth responsibility, with the Office of Health Protection (within the Department of Health) suggested to lead stakeholder engagement.

*I think within the implementation plan the O*ffi*ce of Health Protection has an important role* . . . *I mean they have the remit of the strategy. In terms of the organisations that will have a responsibility some of them are probably clear, and some of them just need coordination. The important part of that is to work in a collaborative way, coordinated way* . . . *We shouldn't be* . . . *isolating sectors such as hospital, aged care* . . . *primary care.* (P5)

There were calls to make practice accreditation mandatory and to include AMS activities such as antibiotic monitoring or education in this. Suggestions were made for financial incentives to encourage AMS activity in general practice.

Stakeholders also generally supported greater regulatory controls on prescriptions, the removal of automatic repeats, and promotion of unit dispensing (dispensed quantities match antibiotic guideline recommendations, not pack sizes).

#### *People you can educate as much as you like, but until you actually restrict the antibiotics people aren't going to stop using them* . . . (P6)

Monitoring and feedback on antibiotic prescribing was perceived as effective for changing behaviour, but the current process was viewed as problematic. Unresolved practical considerations included that complete datasets are not available, the possible defensiveness of GPs about their data being reviewed, questions about who would analyse and provide feedback to GPs, and whether collection should be mandatory or incentivised. The government was regarded as responsible for obtaining complete datasets. Stakeholders saw potential for the Practice Incentives Program—Quality Improvement Incentive [25] (GP data collected by the Primary Health Networks (PHNs) for process measures) to include antibiotic monitoring. Stakeholders said that feedback should include peer comparisons, and ideally link in with education and consultation support. The potential use of positive variance was described, that is, investigating the strategies used by those who prescribe fewer antibiotics than their peers.

*Government needs to incentivize, to capture [antibiotic prescribing] information. You know organisations like the PHNs are really well suited to that.* (P11)

*In terms of investigating what works, one thing that we do poorly is to look for positive variance.* (P7)

Community education in the form of ongoing tailored public health campaigns was considered important and viewed as a government responsibility. There were suggestions that health literacy education for antibiotic awareness should start at school.

*We do need the consumer to come on board to* . . . *not have that expectation [for antibiotics], which then does make the consultation very di*ffi*cult.* (P8)

GP education endorsed by the Royal Australian College of General Practitioners (RACGP) or supplied by PHNs or medical specialists was well regarded and trusted. NPS MedicineWise (an independent organisation supporting quality use of medicines) was acknowledged as an existing channel for GP education, but it was questioned as to whether what was currently provided was at the depth necessary to have the largest impact. There were concerns that pharmaceutical marketing may undermine AMS messages.

*What type of education do GPs trust? And often that'll be one that comes from kind of RACGP-branded things, or PHNs, and sometimes specialist.* (P4)

Stakeholders wanted improved clinical software that integrated prescribing guidelines, patient information resources, and alerts. There was a suggestion that some GPs are using product information rather than guidelines to inform decision making because unlike guidelines, product information is integrated into the clinical software. Government-funded health services (e.g., NPS MedicineWise, PHNs) were suggested as potential developers of patient information resources with PHN Health Pathways as another potential host to make the resources widely available. Keeping the resources current was identified as a challenge.

*I think electronic decision support can work well if it's in real time* . . . *. the first line choices of antibiotics are* . . . *if you couple that with patient information that will be* . . . *made available to the patient, that's helpful.* (P7)

Rapid and point-of-care tests elicited mixed comments. Some thought these could be useful if subsidised. Others thought they should only be available if it would change the decision to prescribe an antibiotic. Selective reporting of antibiotic susceptibilities was suggested as a priority along with standardised information for GPs about the use of microbiology testing, particularly around specimen collection and interpretation of results. It was suggested that the Royal College of Pathologists of Australasia (RCPA) should oversee this.

*Not all labs do selective reporting of antibiotics; it should be implemented* . . . *we need one o*ffi*cial form rather than lots of di*ff*erent ones—they are not as strong as one consistent message.* (P12)

Expert advice sought from hospital specialists (including infectious diseases consultants) was often based on relationships developed during training. There were calls for a central advice line, or lines of communication to enable consistent messages or access to the local hospital specialist's guidance.

*Expert advice for me is very dependent on relationships that I built when I was in the hospital system. So if you've got a good network of experts you can call on but you know from an infection perspective it's* . . . *reliant on the goodness of* . . . *them giving you their time* . . . (P6)

. . . *whether or not the government would be interested in having access lines for antibiotic resistance* . . . *if someone could ring them up* . . . *and get advice, probably wouldn't be a bad thing.* (P8)

Respondents suggested that adding the reason-for-prescription (subject to privacy requirements) and providing an exact duration of antibiotic therapy to the prescription would help community pharmacists be more engaged in AMS. It was perceived that to successfully implement delayed prescriptions (where the patient is told when and under what conditions antibiotics should be dispensed), better communication between GPs and community pharmacists is needed. Pharmacists employed by the general practice were identified as an opportunity for practice-level AMS support.

[pharmacists] *put a sticker on the box of antibiotics that says finish the course* . . . *we should change the stickers to 'take as long as prescribed'* . . . (P8)

Allergy testing was regarded as beneficial for individual patients but not a system issue. Handover of antimicrobial prescribing on patient transfer was considered part of the larger issue of handover of all information.

Stakeholders thought that nurses may have a role in AMS, e.g., patient triage and education in the community and in the practice, but there was a perceived lack of funding.

*I think [nurse triage is] fantastic in an ideal world, but we don't have the funding.* (P8)

Stakeholders agreed that research into general practice AMS with translation of the evidence into practice was required. Research areas suggested included understanding the potentially negative effects of antibiotics on the gut microbiome, and better understanding of the use of delayed antibiotic prescription "*whether an illness that's been present for more days is more likely to respond to antibiotics*" (P7). Stakeholders also suggested more research to understand low prescribing GPs:

*Those who seem to manage to preserve this resource [antibiotics] really well and apparently not with any problems in terms of the health of their patients. Yeah. How does it work for them? What helps them, what supports them? What can we put in place to enable others to not prescribe?* (P9)

#### **3. Discussion**

Stakeholders agreed that the proposed framework was valid and feasible, and provided a suitable action framework for the introduction of AMS into Australian general practice. Central coordination was identified as a priority, but the lack of clarity around who would provide this leadership was surprising, particularly given the seniority of the participating stakeholders. The Office of Health Protection (OHP) was suggested to lead and coordinate the introduction of AMS into Australian general practice. Whether the OHP has the capacity for this was not investigated. Sweden's Strama program offers an example of leadership at county and national levels [17,18,26].

Monitoring of and feedback on antibiotic prescribing will enable targeting and evaluation of AMS interventions. Several issues were highlighted including GP trust in a transparent external audit process [27] and a need to obtain complete datasets (including the reason-for-prescription in a standardised format). Inclusion of information on any adverse patient outcomes, e.g., hospitalisations, would require linkage of datasets [27]. There was a view that monitoring and feedback needs urgent development beyond the current volume-based feedback so that it better meets clinical need. No current monitoring system was identified that could provide the information required. An example of monitoring and reporting are the annual reports published by the English Surveillance Programme for Antimicrobial Utilisation and Resistance (ESPAUR) [28].

Regulatory changes were supported. Manufacturer's pack sizes rarely match the recommended duration for common conditions [29], and when antibiotics were supplied by the pack, patients were thought to be likely to save leftovers for future use [30]. Restrictions on repeat prescriptions for five of the most commonly prescribed antibiotics in Australia were introduced on 1 April 2020 [31], illustrating that regulatory changes are achievable.

Electronic decision support was strongly supported and should be further examined in Australia. It has been used to guide prescribing in hospitals, and has been effective at reducing antibiotic prescribing when combined with other AMS interventions [32]. Work is required to develop and pilot suitable electronic decision support to ensure that the tools meet prescriber needs in Australian primary care, are usable, fit in with workflow [33], and have the desired impact.

Stakeholders were unanimous that community education is required to support general practice AMS. Evidence suggests that campaigns may work best when developed in partnership with consumer organisations, are coordinated with health professionals, and promoted at local and national levels [34]. Community awareness of a common colds campaign reflected changes in the frequency of the

campaign [35], suggesting that community education should be ongoing. School-based programs, such as Europe's eBug [36] and Canada's Do Bugs need Drugs? [37], have introduced AMS to children. Alongside community education, the provision of written patient information was widely supported by stakeholders and has been associated with reduced antibiotic prescribing in common infections [38]. However, the issues of updating the information, which languages and cultural information are required, and the most appropriate place to host these have not yet been well addressed in the literature.

Ongoing work on selective reporting of antibiotic susceptibilities by microbiology services, which has been shown to be effective in influencing prescribing behaviour [39], should be pursued as a priority in Australia [40].

Increased access to expert advice has been utilized internationally as a method to influence antibiotic prescribing choices. Telephone advice has been provided to GPs in France for patient management [22] and in Sweden, experts provide advice on interpretation of audit results [18]. While stakeholders supported the provision of centralised expert advice, there was no clarity on who should provide it beyond the suggestion that local hospital specialists might participate.

Internationally, pharmacists have participated effectively in activities to help reduce antibiotic prescribing and increase prescribing guideline concordance [41], but Australian community pharmacists may require additional support for this expanded role [42]. Non-dispensing pharmacists in general practice may be suitable for an AMS role. Research to explore the role of pharmacists in general practice AMS is recommended. The role in AMS of practice nurses and that of nurses in the community (e.g., phone triage lines) and their need for formal AMS education remains largely unexplored.

Allergy testing and handover of antimicrobial prescribing on patient transfer will be removed from the framework as the former is an individual issue and the latter part a broader issue. No other changes were recommended.

There are limitations to this research: the recruited practice nurse stakeholder was unavailable for interview in the timeframe, so there may be additional insights to be gained regarding the involvement of practice-based nurses. There were only 12 interviews conducted and stakeholder identification was partly reliant on the authors' networks. Areas covered in less detail were the roles for specific organisations in implementation. The RCPA and the Office of Health Protection were specifically named by one stakeholder for each. However, other stakeholders referred more generally to the "professional colleges" and "Department of Health", respectively. Components in which only three stakeholders commented were: planning for new antibiotics, the role of allergy testing, handover of patient information, unit prescribing, and knowledge about other AMS models. Components discussed by four stakeholders included: pharmaceutical company marketing, nurse involvement, monitoring of AMR. All other components were discussed with at least five stakeholders.

The views of the expert stakeholders may not reflect those of the wider GP community. Experts are likely to be early adopters or innovators in a field [43], whereas the wider community will include those who fear the consequences of not having antibiotics and those who may not perceive that AMR affects them. The stakeholders were speaking as experts, not as representatives of their organisations, thus it is unknown if the organisations have the current capacity to implement the framework.

#### *I should just say* . . . *I'm not doing this from [a named organisation] policy view.* (P3)

This is a health system-wide framework developed from a review of the international literature [24] which identified components that may play an interdependent role affecting GP antimicrobial prescribing. A systematic review of interventions found that "No single intervention can be recommended for all behaviours in any setting" and that "local barriers should be removed before implementation" [44]. Examination of these components may help to explain why an intervention may be successful in one setting but fail in another [44,45]. This research highlights that AMS in general practice needs a health system leader, the involvement of health departments, especially One Health AMS committees, with input from professional colleges and health professional representatives. Implementation science and behaviour change principles [46–48] with GP and relevant professional input are recommended to pilot, implement, and evaluate changes.

#### **4. Materials and Methods**

#### *4.1. Study Design and Participants*

A qualitative approach was used. Australian-based senior expert stakeholders in AMS in general practice were identified through the authors' AMS networks (8), relevant organisations' websites (3), and via contact with government and professional organisations (2). Stakeholders were provided with a study information sheet and purposively invited to participate in a telephone interview. Gift cards to the value of AUD 150 were offered as compensation for their time.

#### *4.2. Data Collection and Qualitative Analysis*

Consented participants received an outline of the proposed AMS framework prior to the interview (Appendix A).

In-depth telephone interviews using a semi-structured interview guide (Appendix B) were conducted and recorded between September and December 2019. Stakeholders were purposively invited until key components had been discussed with at least one stakeholder. Feasibility and validity were assessed by asking participants the extent to which components and subcomponents were being done or if plausible, what needed to be done to make them implementable; their priorities; and if they could identify any gaps. Data collection was completed before analysis commenced. Interview recordings were transcribed and returned to stakeholders with a 10–14-day window for amendments. Transcripts underwent thematic analysis using deductive coding targeting comments about the proposed framework and its components, and by open coding for other comments [49]. Two transcripts were independently coded by two authors and an agreed coding framework was developed. Three more interviews were dual coded using the agreed framework and adjustments made. Seven transcripts were coded by one author. NVivo 12 qualitative data analysis software (QSR International Pty Ltd. Chadstone, Australia) was used to manage the transcripts and coding.

Ethics approval was granted by the Monash University Human Research Ethics Committee, number 20721.

#### **5. Conclusions**

The stakeholders regarded this AMS framework as feasible and valid for Australian general practice. The individual subcomponents were viewed as providing a link between the objectives of Australia's National AMR Strategy and action. However, stakeholders considered that the framework required an implementation process with priorities and an integrated approach. The identification of a clear leader to drive AMS in general practice is essential for AMS to gain traction. Monitoring and feedback of antibiotic prescribing require urgent development beyond the current volume-based system and should include monitoring of appropriateness of the prescriptions and patient outcomes. AMS education for the public, further development of GP education, and improved consultation support were strongly recommended. The role of community-based pharmacists and nurses is largely unexplored but their involvement, particularly for patient education, was recommended. Several areas for research were suggested.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6382/9/12/900/s1, Table S1: The COREQ checklist, Table S2 Representative quotes for AMS components.

**Author Contributions:** Conceptualization, L.A.H.; methodology, L.A.H.; formal analysis, L.A.H., J.B., K.B., D.M.; resources, D.M.; data curation, L.A.H.; writing—original draft preparation, L.A.H.; writing—review and editing, L.A.H., J.B., K.B., D.M.; supervision, K.B., D.M.; project administration, L.A.H.; funding acquisition, K.B., D.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by a National Health and Medical Research Council Centre for Research Excellence Grant for the National Centre for Antimicrobial Stewardship (APP1079625), and a PhD stipend from the National Centre for Antimicrobial Stewardship for L.A.H. and J.B.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Appendix A. Component List Used during the Interviews**

The detailed list of the subcomponents for antimicrobial stewardship in general practice. This list was sent to each Stakeholder before interview and referred to during the interview.

	- a. National action plan;
	- b. Antimicrobial resistance included on national risk register;
	- c. Multi-level and/or multi-disciplinary response;
	- d. Regulations around antimicrobial stewardship and antibiotic prescribing;
	- e. Accreditation of prescribers;
	- f. Funding for antimicrobial resistance and stewardship activities;
	- g. Planning for release of new antibiotics;
	- h. Practice level antimicrobial stewardship policy/program/activities;
	- i. Handover of antibiotic information.
	- a. Community and patient education;
	- b. GP continuing education in antimicrobial stewardship;
	- c. GP education on communication skills, patient-centred approaches and shared decision making;
	- d. GP education on non-antibiotic management of self-limiting infection;
	- e. GP education on delayed prescribing;
	- f. General practice team member education;
	- g. Independent education (restrict pharma marketing).
	- a. Prescribing guidelines;
	- b. Point of care tests;
	- c. Microbiology testing and reporting;
	- d. Allergy testing;
	- e. Electronic decision support for prescribers;
	- f. Expert advice;
	- g. Decision support for use with patients.
	- a. Unit dispensing;
	- b. Supply and timely access to antibiotics;
	- c. Pharmacy review and advice;
	- d. Appropriate disposal of leftover antibiotics;
	- e. Nurse triage, patient assessment and education.
	- a. Monitoring of antibiotic prescriptions;
	- b. Monitoring of antimicrobial resistance;
	- c. Feedback to prescribers and reporting.
	- a. Research into AMR/AMS gaps, translation into practice.

#### **Appendix B. The Semi-Structured Interview Guide**


Now I will take 2–3 min to explain the model framework and then I will ask you for your comments on it.

	- a. How do they differ from this model?
	- a. What needs to be done to make it happen?

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Review*

#### **Are Follow-Up Blood Cultures Useful in the Antimicrobial Management of Gram Negative Bacteremia? A Reappraisal of Their Role Based on Current Knowledge**

**Francesco Cogliati Dezza** † **, Ambrogio Curtolo** † **, Lorenzo Volpicelli** † **, Giancarlo Ceccarelli , Alessandra Oliva and Mario Venditti \***

Department of Public Health and Infectious Diseases, University "Sapienza" of Rome, 00185 Rome, Italy; francesco.cogliatidezza@uniroma1.it (F.C.D.); ambrogio.curt@uniroma1.it (A.C.); lorenzo.volpicelli@uniroma1.it (L.V.); giancarlo.ceccarelli@uniroma1.it (G.C.); alessandra.oliva@uniroma1.it (A.O.)


Received: 6 November 2020; Accepted: 9 December 2020; Published: 11 December 2020

**Abstract:** Bloodstream infections still constitute an outstanding cause of in-hospital morbidity and mortality, especially among critically ill patients. Follow up blood cultures (FUBCs) are widely recommended for proper management of *Staphylococcus aureus* and *Candida* spp. infections. On the other hand, their role is still a matter of controversy as far as Gram negative bacteremias are concerned. We revised, analyzed, and commented on the literature addressing this issue, to define the clinical settings in which the application of FUBCs could better reveal its value. The results of this review show that critically ill patients, endovascular and/or non-eradicable source of infection, isolation of a multi-drug resistant pathogen, end-stage renal disease, and immunodeficiencies are some factors that may predispose patients to persistent Gram negative bacteremia. An analysis of the different burdens that each of these factors have in this clinical setting allowed us to suggest which patients' FUBCs have the potential to modify treatment choices, prompt an early source control, and finally, improve clinical outcome.

**Keywords:** follow-up blood cultures; Gram negative bacteremia; critically ill patients; antibiotic therapy

#### **1. Introduction**

Bloodstream infections (BSIs) represent a leading cause of death in industrialized countries, with an estimate of two million episodes and 250,000 deaths annually in North America and Europe, despite the availability of new potent antimicrobial therapies and advances in supportive care. In particular, hospital-acquired BSIs are a major cause of morbidity and mortality in intensive care units (ICU), and septic shock still represents the first cause of ICU total mortality [1]. This burden is likely to grow over the next few decades due to the increase in life-expectancy and in median number of patient comorbidity [2]. Unlike Gram positive (GP) BSIs, whose incidence rate has declined over the last few decades, Gram negative (GN) BSIs have markedly increased overtime and nowadays account for up to half of BSIs, with a mortality rate of 20–40% [3–5].

When considering GP-BSIs, international guidelines and consolidated evidence-based procedure bundles are available for the management of the leading pathogen species, *Staphylococcus aureus*. In this setting, follow-up blood cultures (FUBCs) are regarded as essential to document clearance of bacteremia after treatment initiation and exclude seeding [6–8]. On the other hand, FUBCs are

mandatory in the case of *Candida* spp. BSIs in order to determine the end of candidemia and optimize treatment duration [9].

As for GN-BSIs, relevant advances in management strategies have been made in the last few years, such as the non-inferiority demonstration of 7 vs 14 day antibiotic courses [10] and of oral step-down vs continued parental therapy [11] in uncomplicated GN-BSIs. Recently, a combined approach of rapid diagnostic testing with a bundle of antimicrobial stewardship found a decrease in readmission rate and in cost per case [12]. Anyway, the management of GN-BSIs remains poorly codified and thus prone to personal clinician judgement, as compared to Gram positive settings. In a recently published scoping review, Fabre et al. proposed an algorithm for bacterial blood culture (BC) recommendations. They found that in bacteremias due to *Enterobacterales*, FUBCs are unlikely to grow unless the source of infection is endovascular or there is inadequate source control. Although with small numbers, similar results were found in *Pseudomonas* infections. The authors suggest clinical judgment to evaluate the need of FUBCs for GN. Of note, this review considered studies published from 1 January 2004 to 1 June 2019 [13], but the majority of studies addressing the topic of FUBCs in GN infections are actually subsequent to this time frame. At the time of writing, the role of FUBCs in Gram negative bacteremia (GNB) still represents an important matter of debate, with controversial results [14].

The early studies conducted have focused on the disadvantages of FUBCs, mainly represented by the risk of false positive results, prolonged hospitalization, inappropriate antibiotic use and increased cost [15,16]. Recently, the issue of FUBCs came out on top through the availability of new evidence that may have tipped the balance in favor. In this review, we aimed to examine and summarize the current knowledge on the usefulness of FUBCs in GNB, especially in the light of the reassessment of this management tool by recent studies. Moreover, we propose two guidance tools (clinical and microbiological) that summarize and graduate the recommendations for FUBCs.

#### **2. Results: Review of the Literature**

In 2004, Tabriz et al. published the first study where FUBCs in GNB were suggested. The authors conducted a retrospective single center study of 96 patients with at least one FUBC over 1 month (199 BC episodes without differences between GP and GN). Most FUBCs were performed within 4 days from first positive blood culture (FPBC) and during antimicrobial therapy (AT) (both 158, 79.4%). The common reasons to repeat BCs were fever, follow-up of positive BCs, and persistent leukocytosis. Positive FUBCs after FPBC were 21 (25.9%) and after a negative BC were 1.7% of cases. The conclusions were that persistent leukocytosis and fever are poor predictors of bacteremia. Then, the authors gave some general indications about when to perform FUBCs correctly, even if no definition of FUBCs was provided. They suggested FUBCs in these circumstances: 1. new septic episode, 2. suspected endocarditis, 3. follow-up of a positive BC in certain conditions that may have diagnostic and therapeutic implications, such as *S. aureus*, GNB and candidemia, 4. confirmation of response to therapy for endocarditis or other endovascular infections caused by *S. aureus*, *Enterococcus* spp., GN or other difficult-to-treat organisms because the only use of clinical data may not be reliable, 5. confirmation of diagnosis of intravascular catheter-associated bacteremia. Even if the data presented do not seem to support the above cited conclusions, this was the first time that the question of FUBCs in GNB was dealt with [17].

A summary of the main subsequent studies on the role of FUBCs in the management of GNB is reported in Table 1.


**Table 1.** Summary of the main studies on follow up blood cultures (FUBCs) in Gram negative bacteremia.


**Table 1.** *Cont.*

*Antibiotics* **2020**, *9*, 895

*Antibiotics* **2020**, *9*, 895



**Abbreviations**: AT, antimicrobial therapy; BC, blood culture; BSI, blood stream infection; CCI, Charlson comorbidity index; CRP, C-reactive protein; CVC, central venous catheter; ESBL, extended spectrum β-lactamase; ESRD, end stage renal disease; FUBC, follow-up blood cultures; GN, Gram negative; GNB, Gram negative bacteremia; GP, Gram positive; ICU, intensive care unit; KpB, *Klebsiella pneumoniae* bacteremia; MDR, multidrug resistant; NA, not available/not applicable; PB, persistent bacteremia; RF, risk factor; SOT, solid organ transplant; UTI, urinary tract infection.

In a retrospective multicenter case-control study of 2013 [18], the authors analyzed 1068 individuals with *Klebsiella pneumoniae* bacteremia (KpB) and showed a wide prescription of FUBCs as these were performed in 80.7% of cases, while only 7.2% were found to be positive. Moreover, 53.2% of patients with non-persistent KpB underwent more than two consecutive BCs. The routine use of FUBCs was considered not justified because of the low incidence of persistent bacteremia (PB) detected. Unfavorable treatment response on the second day after the initial BCs, intra-abdominal infection, high weighted Charlson comorbidity index, and prior solid organ transplantation (SOT) were recognized as independent risk factors for persistent KpB. The authors stated that the retrospective analysis, the small sample size, and the lack of a multivariable analysis of mortality-related factors represented possible limitations of their study. Furthermore, they focused only on a specific pathogen (*K. pneumoniae*) and, although transfer to ICU was considered as an outcome, no data were available concerning the original patient allocation (ICU vs non-ICU ward) [18].

In 2016, Wiggers et al. conducted a retrospective monocentric cohort SCRIBE study on a mixed population of 1801 patients with a first episode of bacteremia caused by GP, GN or anaerobic bacteria. FUBCs were executed in 701 patients (38.9%) and PB was demonstrated in only 118 (6.6%) of the whole population. As expected, an endovascular source of infection, *S. aureus* and the inability to achieve source control in 48 h were associated with higher risk of PB. Analyzing the data provided, there were 901 GNBs (50% of the whole cohort) and 247 of them had FUBCs taken (27.4%), of which 27 (10.9%) tested positive, compared to GP bacteremias (GPBs) where BCs were repeated in 457 out of 882 patients (51.8%), with a positive yield in 90 (19.7%). Male sex, admission to a medical service, *S. aureus* bacteremia and endovascular or epidural focus were identified as risk factors for PB, but unfortunately this multivariate analysis was conducted on the whole population, rather than only on patients with GNB. Authors concluded that bacteremias caused by GNs, viridans group or beta-hemolytic streptococci are common situations in which repeat BCs offer low yield, with the related inappropriate expense. When possible, a revision of the charts regarding physician's impression on clinical status was made and authors inferred that only 30.3% of FUBCs were drawn because of patients' instability, a situation in which FUBCs could be suitable. Regardless of the result, 30 day mortality was significantly higher (27%) among patients undergoing repeated BCs [19].

Until 2020, the most relevant and influential article addressing the topic of FUBCs in GNBs was that conducted by Canzoneri et al. in 2017. They retrospectively analyzed 500 episodes of bacteremia, of which 383 (77%) had at least one FUBC taken. Among these 383, 206 (54%) had initial bacteremia caused by a GP organism and 140 (37%) by a GN, with an average of 2.37 FUBCs per patient. The FUBCs yielded positive in 55 (14%) of the overall population, 43 (78%) of those with GP cocci and eight (15%) with GN bacilli. The incidence of PB, defined as positive FUBC for the same original organism, was 21% in GPBs, 10% in polymicrobial bacteremias and 6% only in GNBs. Fever on the day of FUBC sampling, presence of an intravenous (IV) central catheter and end-stage renal disease (ESRD) were associated with a higher probability of PB in the whole cohort. Subgroup analysis confirmed these factors, with the adjunct of diabetes mellitus, as predictors of positive FUBC, only among subjects with GPB, while fever was the only factor associated with PB in GNB. No impact of positive FUBC on ICU admission or mortality was detectable. No clue concerning the clinical reasons for drawing FUBCs was available. The authors concluded that FUBCs may have little utility in patients with GNB, as compared to the serious negative implications of unrestrained use, represented by false positive results, longer hospital stays and increased healthcare costs [15].

In 2019, Shi et al. reported the results of a monocentric case-control study in 333 patients with bacteremic urinary tract infection (UTI): 306 (91.9%) of them had FUBCs drawn, of which 55 (18%) tested positive. Among all those that underwent FUBCs, 264 (86.3%) had a GN-related UTI with positivity in 39 (14.8%), compared to 14 (4.6%) with a GP UTI that yielded positive FUBCs in six (42.9%). Of note, four out of six of this latter group were caused by *S. aureus*. PB, defined as more than seven days of positive BCs, occurred in only six (3.3%) out of 306 patients. Several clinical and biochemical factors were associated with higher probability of PB. Eventually, four factors were selected and

confirmed through multivariate analysis as independent predictors: malignancy, initial ICU admission, high c-reactive protein (CRP) level and longer time to defervescence. Among the subgroup of patients without any of these risk factors, no one had a positive FUBC. The authors concluded that, due to the low positivity rate, liberally prescribed FUBCs have little utility in the management of bacteremic UTI [20]. However, the results of this study might be limited by the design focused on only one clinical syndrome (UTI), the mixed causative agents considered (including a not negligible proportion of *S. aureus*) and the lack of data regarding the effect of pathogen antibiotic susceptibilities on FUBC results, clinical reasons for drawing FUBCs and original admission service of the patients.

In the same year, a study conducted in a pediatric hospital of Tokyo firstly questioned the conception of the usefulness of FUBC in the setting of GNB. Uehara et al. enrolled 99 children with GNB, with a median age of two years. The most frequent underlying diseases were SOT (21.2%), malignant neoplasm (17.2%) and kidney/urinary tract malformation (15.2%); a central venous catheter (CVC) was in place in 57% of patients. Twenty-one patients (21.2%) had positive FUBCs, with *Klebsiella* spp. and *Escherichia coli* being the most represented pathogens. Interestingly, no cases of positive FUBCs emerged among patients with UTI. Multivariate analysis revealed the presence of CVC and resistance to empirical therapy as significantly associated with PB. More importantly, the authors reviewed clinical charts and reported that the positive yield of FUBCs promoted a treatment modification in 57% of patients, which included optimization of antibiotic therapy and/or removal of medical devices [21].

In 2020, the prior view of FUBCs as a tool of little utility in patients with GNB, counterpoised to many clinical and economical drawbacks, underwent a systematic reassessment. Completely different results from those obtained by Canzoneri et al. [15] were in fact reported by Giannella et al. in a single center, retrospective cohort analysis of 1576 patients with GNB [22]. As in previous studies, FUBCs were prescribed based on personal clinical judgement rather than systematically. Nevertheless, FUBCs were performed in only 278 (17.6%) patients but demonstrated a high rate of PB: 107 (38.5%). Patients that underwent FUBCs were younger, with a lower Charlson comorbidity index, but more frequently immunocompromised, admitted to ICU, with a hospital-acquired GNB, and with a non-urinary source of infection, compared to those without FUBCs performed. Furthermore, patients with FUBC taken had higher initial severity of GNB clinical pictures according to SOFA (Sequential Organ Failure Assessment) score and septic shock criteria, higher frequency of carbapenem-resistant enterobacteriaceae (CRE) isolation and of inappropriate empirical therapy. Thus, as a matter of fact, the patient complexity seemed to progressively rise from those without FUBCs drawn, those with FUBCs, to those with positive FUBCs. That is to say that, for the first time since the topic of FUBCs in the setting of GNB has been debated, the authors provided elements to interpret the mechanism by which physicians currently use FUBCs. Interestingly, taking into account the higher complexity of patients that underwent FUBCs, the execution of FUBCs was followed by increased rate of source control, infectious disease consultation and longer treatment duration. Thus, performance of FUBCs appeared to act for physicians as an incitement to more careful management. At the same time, Giannella et al. demonstrated that FUBCs had a favorable impact on patient outcome, an effect probably linked to prompt source control. In fact, through multivariate analysis, FUBCs, along with UTI origin of BSI, source control and active empiric therapy, resulted as independent factors protective from all-cause 30 day mortality. The authors concluded that future prospective studies with a systematic use of FUBCs in GNB are necessary in order to better identify the settings where FUBCs could be cost-effective [22].

Similar favorable results of FUBC use were obtained by Maskarinec et al. in an observational study of 1702 prospectively enrolled inpatients with monomicrobial GNB [23]. FUBCs were drawn in 1164 patients (68%) and more commonly in patients with *Pseudomonas aeruginosa* and *Serratia* spp. (80%). PB was detected in 228 (20%). Patients with PB had a lower probability of having under effective antibiotic treatment (with higher rates of fluoroquinolone and/or carbapenem-resistant isolates in FUBCs) and higher probability of being a transplant recipient, hemodialysis dependent, having a cardiac device, recent corticosteroid use, a malignancy, or an endovascular source of infection. Bacteremias caused by *Serratia* (32%, 95% CI 24–44%) and *Stenotrophomonas maltophilia* (52%, 95% CI

32–72%) had the higher rate of persistence. Patients' clinical outcomes were also evaluated. Relevantly, the regression model showed that obtaining FUBCs was associated with decreased rates of both all-cause and attributable mortality. This result was also confirmed in a species-specific analysis performed for *E. coli* and *K. pneumoniae* and in a sensitivity analysis that excluded all deaths occurring in the first 48 h. On the contrary, PB implied a nearly double all-cause and attributable mortality relative to those with negative FUBCs, and similar to those without FUBCs drawn. The probability of PB was estimated through a risk scoring system and finally, an endovascular source of infection was identified as the only breakpoint separating high and low rates of FUBC positivity. The authors concluded that FUBCs have clinical utility in detecting patients with increased risk of poor outcome, and that could benefit from additional diagnostic and therapeutic interventions. Considering the low rate (2%) of false positivity FUBCs and the little difference in duration of antibiotic treatment (2 days), this study also scaled back the traditional concerns of increased healthcare costs and antimicrobial prescription usually attributed to FUBCs [23].

On the other hand, Jung et al. from South Korea recently presented a retrospective observational cohort study conducted on 1481 cases of GNB. FUBCs were widely performed (86.2%), while positivity resulted in 122 (9.6%) [24]. Comparing the clinical characteristics of those that underwent FUBCs and those that did not, female gender, neutropenia, hematologic malignancy, presence of an intravascular device and of an extended spectrum β-lactamase (ESBL)-producing organism were more represented in the first group, while a biliary source was more common in the latter. No difference was detected in terms of incident mortality between FUBCs drawn and not. The comparison between patients with positive and negative FUBC yield was made by sub-stratification according to eradicable and non-eradicable source of infection. Several factors were identified and included in a predictive scoring model if independently associated with FUBC positivity through a multivariate logistic regression. Results indicated that, in the case of a removable source of infection, if there is appropriate management (early source control and appropriate therapy) followed by a favorable clinical response (quick SOFA score <2), performing FUBCs adds little value. Furthermore, even in a non-eradicable setting, the administration of effective treatment corresponded to 95% probability of negative conversion, regardless of the underlying disease, offending pathogen, or treatment response. Author conclusions were that FUBCs can be avoided in most uncomplicated cases of GNB and could be considered selectively in high-risk patients. In addition to the retrospective nature of the study, it should be underlined that neither the patients' outcome nor the relative impact of FUBCs' execution/results were evaluated. In addition, mortality was not taken into consideration in the comparison between positive and negative results of FUBCs and no data concerning ICU vs non-ICU ward allocation were provided [24].

Mitaka et al. conducted a retrospective multicenter observational study in all adults with at least one BC positive for GNs admitted between January 2017 and December 2018 [25]. A total of 463 patients were included; of these, 306 (66%) had FUBCs performed at least once. The results showed positive FUBCs in only 10% of patients. The authors found a correlation between positive FUBCs and the following risk factors: ESRD, presence of intravascular devices, and bacteremia due to ESBL-producing organism or CRE. The yield of positive FUBCs in patients without the risk factors was 1.6%, compared to 14.8% in the presence of ≥1 risk factor. The authors concluded that FUBCs may not be necessary for all GNBs, but only in the presence of risk factors [25]. In addition to the retrospective design and the lack of standardization of decision making regarding FUBCs, other limitations of the study included that authors only analyzed the positivity or negativity of the first FUBC without checking the possibility of intermittent bacteremia and no information about clinical outcomes and therapeutic change based on the results of FUBCs was provided.

In 2020, Spaziante et al. conducted a retrospective single center observational study on 307 patients admitted to a multidisciplinary ICU in 2017 [26]. Sixty-nine patients (22.4%) presenting with at least one GNB episode for a total of 107 episodes were included in the study. Exclusion criteria were the occurrence of fungemia, GP or mixed GP/GN bacteremic episodes. FUBCs were defined as BC performed within 48 h from the beginning of antimicrobial therapy (AT) and then every 24–72 h after FPBCs. PB was defined as repeatedly positive BCs for GNs after ≥96 h of appropriate AT and ≥48 h after removal of all potentially infected endovascular indwelling devices. Twenty-nine GNB episodes (27.1%) were excluded from the study because no FUBCs were performed. Eventually, 28 (35.9%) out of 78 GNB episodes were diagnosed as PB. Under these circumstances, septic thrombosis (ST) was the hematogenous source of infection in approximately half of the cases, resulting in a significant association with positive FUBCs (*p* < 0.001). On the other hand, negative FUBCs were associated with primary bacteremia (*p* < 0.001). As part of the retrospective design, this is the only paper entirely focused on critically ill patients. In particular, the study was conducted in an ICU that is a reference center for polytrauma and, based on the aforementioned results, the authors hypothesized that frequent deep venous thrombosis occurring near to bone fractures may provide a suitable medium for microbial seeding for GNB originating from other body sites [26].

#### **3. Discussion**

GNB still represents an extremely relevant cause of morbidity and mortality in hospitalized patients, especially in ICU settings. Therefore, it is crucial to achieve an optimization of patient care, which, at the same time, could reduce mortality rates and meet the growing demands of antimicrobial stewardship and cost control. In this regard, the use of FUBCs in patients with GNB has represented a contentious topic in the past few years and especially in the very last period.

Indeed, unnecessary FUBCs may cause patient discomfort and carry the risk of false positive results. According to previous reports, as many as 90% of all BCs grow no organisms and of the approximate 10% that do grow organisms, almost half are considered contaminants (false positives) [27]. Thus, given a constant rate of contamination, performing more FUBCs may result in a higher chance of encountering contaminant organisms, and consequently, in increased costs and patients discomfort, longer hospital stays, unnecessary consultations, and inappropriate antimicrobial therapy [15,16]. Of note, this reasoning includes an inherent fallacy. Although theoretically acceptable in cases of GP yield, the growth of GN bacteria in BCs should always be regarded as relevant and never, or just anecdotally, be considered as a contamination [28].

On the other side, performing FUBCs may have a relevant impact on patient management and outcome, reducing mortality rates. When FUBCs are performed in more severe patients with comorbidities and without adequate infection source control, in cases of bacteremia due to multidrug resistant (MDR) GNs and without an appropriate empiric therapy, the positive or negative results may guide the clinician to the correct decision about type and duration of antibiotic therapy [22,23].

Not surprisingly, the evidence concerning the usefulness of FUBCs underwent a progressive shift in the last few years, going from a restrictive to a selective approach. Taking a look at Table 1, it appears that papers with a higher rate of FUBCs performed, more frequently found no evidence of benefit in contrast to those that applied FUBCs more selectively. In fact, Kang et al. [18] and Jung et al. [24] performed FUBCs in GNB in 81% and 86% of the cases, respectively, as compared to Maskarinec and Giannella that used this tool only in 68% and 17% of cases, respectively. Likely, Spaziante et al. [26] found a high rate of PB by only performing FUBCs in high-risk ICU patients with GNB. Furthermore, the inclusion in the analysis of a mixed GP and GN population was also associated with pessimistic results. Therefore, it seems that the more refined the selection of patients in which to draw FUBCs, the more evident the benefits they bring. Indeed, the results of our literature review show that, while in some subgroups of patients the use of FUBCs may not translate into a clear benefit, it is possible to identify several situations where the application of this tool may steer the clinical decision making in the correct way. Thus, selection is a sticking point in this topic.

For the purposes of performing rational FUBCs, physicians well trained in infectious diseases should be available in all settings where risk factors for positive FUBCs are present. For this reason, ICU patients might be evaluated from dedicated infectious disease consultants in order to avoid unnecessary BCs [22,26]. The study written by Ceccarelli et al. shows an example of this management [29]. In fact, they reported some cases of GN-related septic thrombosis (ST) with indolent clinical course and long-term positive BCs despite adequate antibiotic treatment as one possible exception to the restricted use of FUBCs. In these cases, FUBCs allowed the determination of the correct treatment duration, representing a tool of critical importance for patient management [29]. The same group of authors further stressed this concept in a case series of 13 critical care patients with ST caused by GN bacilli (Figure 1): this disease was characterized by PB despite prompt source control and appropriate antibiotic treatment, an indolent clinical course and, even more important, a rapid defervescence with normalization of procalcitonin (PCT) values preceding bacteremia clearance. This phenomenon was interpreted as a mechanism of immune tolerance [30]. [29]. In fact, they reported some cases of GN-related septic thrombosis (ST) with indolent clinical course and long-term positive BCs despite adequate antibiotic treatment as one possible exception to the restricted use of FUBCs. In these cases, FUBCs allowed the determination of the correct treatment duration, representing a tool of critical importance for patient management [29]. The same group of authors further stressed this concept in a case series of 13 critical care patients with ST caused by GN bacilli (Figure 1): this disease was characterized by PB despite prompt source control and appropriate antibiotic treatment, an indolent clinical course and, even more important, a rapid defervescence with normalization of procalcitonin (PCT) values preceding bacteremia clearance. This phenomenon was interpreted as a mechanism of immune tolerance [30].

unnecessary BCs [22,26]. The study written by Ceccarelli et al. shows an example of this management

For the purposes of performing rational FUBCs, physicians well trained in infectious diseases

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 9 of 15

chance of encountering contaminant organisms, and consequently, in increased costs and patients discomfort, longer hospital stays, unnecessary consultations, and inappropriate antimicrobial therapy [15,16]. Of note, this reasoning includes an inherent fallacy. Although theoretically acceptable in cases of GP yield, the growth of GN bacteria in BCs should always be regarded as relevant and

On the other side, performing FUBCs may have a relevant impact on patient management and outcome, reducing mortality rates. When FUBCs are performed in more severe patients with comorbidities and without adequate infection source control, in cases of bacteremia due to multidrug resistant (MDR) GNs and without an appropriate empiric therapy, the positive or negative results may guide the clinician to the correct decision about type and duration of antibiotic therapy [22,23]. Not surprisingly, the evidence concerning the usefulness of FUBCs underwent a progressive shift in the last few years, going from a restrictive to a selective approach. Taking a look at Table 1, it appears that papers with a higher rate of FUBCs performed, more frequently found no evidence of benefit in contrast to those that applied FUBCs more selectively. In fact, Kang et al. [18] and Jung et al. [24] performed FUBCs in GNB in 81% and 86% of the cases, respectively, as compared to Maskarinec and Giannella that used this tool only in 68% and 17% of cases, respectively. Likely, Spaziante et al. [26] found a high rate of PB by only performing FUBCs in high-risk ICU patients with GNB. Furthermore, the inclusion in the analysis of a mixed GP and GN population was also associated with pessimistic results. Therefore, it seems that the more refined the selection of patients in which to draw FUBCs, the more evident the benefits they bring. Indeed, the results of our literature review show that, while in some subgroups of patients the use of FUBCs may not translate into a clear benefit, it is possible to identify several situations where the application of this tool may steer the clinical decision making in the correct way. Thus, selection is a sticking point in this topic.

never, or just anecdotally, be considered as a contamination [28].

**Figure 1.** Duration of bacteremia and clinical course features of Gram negative septic thrombosis in critically ill patients, modified from Spaziante et al. [30]. This figure shows that bacteremia may persist despite clinical improvement (fever disappearance, negative procalcitonin (PCT) values and no vasopressor support); under these circumstances, FUBCs may remain the only driver of antibiotic therapy.

Based on the fascinating hypothesis that FUBCs in BSIs due to GNs could be part of clinical practice, we tried to investigate which conditions make FUBCs either necessary or unfounded. Figures 2 and 3 show clinical and microbiological risk factors for PB, respectively. Rows and columns of each figure intersect in a colored square and every color means a risk threshold of PB, from green (FUBCs highly recommended), light green (moderate recommendation), yellow (weak recommendation) to red (avoid FUBCs), of various combinations of all risk factors recognized in this review.

In order to point out the right setting where FUBCs should be prescribed (otherwise when they are not warranted), we created Figures 2 and 3 analyzing the risk factors for persistent bacteremia found in the revised articles. On the other hand, we identified settings where FUBCs are moderate or even weakly recommended based on risk factors cited by less authors/articles and our judgment. For these reasons, a clinical decision on a case-by-case basis is needed to judge when to perform FUBCs.

In general, FUBCs should be always considered in critically ill patients because they frequently present multiple risk factors that may account for resistant or persistent GNB: intravascular catheter, antibiotic resistant pathogen or an occult source of infection that requires control. As an example, FUBCs are warranted in patients with an endovascular source of GNB that represents the single most important indication to this procedure, even in some instances where a biomarker of active infection, such as PCT, is decreased to negative values. In fact, Spaziante et al. stressed this point by analyzing Gram negative ST where positive FUBCs played a crucial role in patient outcomes (Figure 1). Along the same line, the presence of a non-eradicable infection source appears to be a condition in which FUBCs should be performed, especially in patients that require ICU, with persistent fever or as a moderate recommendation (light green) in individuals afflicted with ESRD on hemodialysis [23–25]. FUBCs

might also be indicated in persistently febrile patients with primary bacteremia, long term intravascular catheter or urinary tract infection, to a lower degree of evidence. Additionally, we suggest that even in some instances of a lower positivity rate of FUBCs, a clinical decision might be reached in cases of light green and yellow squares both in Figures 2 and 3. For this reason, in the case of patients without a clear clinical indication to perform FUBCs as shown in Figure 2, they should be checked for microbiological risk factors, as shown in Figure 3. To this end, in our opinion, FUBCs might be performed even in instances of microbiological risk factors alone. In fact, taking clinically stable patients with UTI as an example (yellow to red squares in Figure 2), instances where the column of MDR microorganisms' etiology and the row of ineffective therapy intersect in a green square might represent a recommendation to perform FUBCs [ *Antibiotics* **2020**, *9*, x FOR PEER REVIEW 20,21,24,25]. 11 of 15

**Figure 2.** Analysis of recommendations for FUBC sampling in the setting of Gram negative bacteremia on the basis of clinical features and source of infection**. Note***:* infection source: \* when apparently there is not an infection source; \*\* if another infection source is presented; \*\*\* fever for more than 72– 96 h; \*\*\*\* malignancy, solid organ transplantation; \*\*\*\*\* Charlson comorbidity index ≥ 6. References: critical ill patient [20,22,23,25,26]; persistent fever [15,20,26]; ESRD, immuno-suppression [15,18,20,23–25]; prolonged corticosteroid (CCS) therapy [23]; persistent PCT/C-reactive protein (CRP) positivity [20]; Charlson comorbidity index [18]; PCT/CRP negative [20]; endovascular source [19,23,24,26]; no eradicable source [22,24,25]; primary bacteremia [26]; cardiac device/intravenous (IV) catheter [15,21,23–25]; intra-abdominal source [18]; UTI [20,23,25]; adequate source control [24]. **Abbreviations***:* UTI, urinary tract infections; ICU, intensive care unit; ESRD, end stage renal disease; CCS, corticosteroids; PCT, procalcitonin; CRP, C-reactive protein. **Figure 2.** Analysis of recommendations for FUBC sampling in the setting of Gram negative bacteremia on the basis of clinical features and source of infection. **Note:** infection source: \* when apparently there is not an infection source; \*\* if another infection source is presented; \*\*\* fever for more than 72–96 h; \*\*\*\* malignancy, solid organ transplantation; \*\*\*\*\* Charlson comorbidity index ≥ 6. References: critical ill patient [20,22,23,25,26]; persistent fever [15,20,26]; ESRD, immuno-suppression [15,18,20,23–25]; prolonged corticosteroid (CCS) therapy [23]; persistent PCT/C-reactive protein (CRP) positivity [20]; Charlson comorbidity index [18]; PCT/CRP negative [20]; endovascular source [19,23,24,26]; no eradicable source [22,24,25]; primary bacteremia [26]; cardiac device/intravenous (IV) catheter [15,21, 23–25]; intra-abdominal source [18]; UTI [20,23,25]; adequate source control [24]. **Abbreviations:** UTI, urinary tract infections; ICU, intensive care unit; ESRD, end stage renal disease; CCS, corticosteroids; PCT, procalcitonin; CRP, C-reactive protein.

Getting to the point, if Figure 2 recommends FUBCs, it deals with a "green light" and they should be performed in any case. Additionally, when FUBCs are not indicated by clinical risk factors, clinicians should check the presence of microbiological risk factors, as shown in Figure 3. In the case of neither clinical nor microbiological risk factors, we are in front of a "red light" and FUBCs are not warranted.

A possible limitation of Figures 2 and 3 is that the definitions of risk factors are often different between studies or even not provided at all. For instance, the dosage and duration of corticosteroid treatment with a significant immunosuppressive effect are not clearly defined in the literature as they are considered to depend on the characteristics of the patient and underlying disease [31].

Finally, progressive acquisition of resistance during antimicrobial therapy through the selection of a hidden resistant subpopulation (named hetero-resistance) is a growing concern in the case of persistent bacteremia [32]. Future studies should elucidate the possible role of FUBCs in early detection *Antibiotics*  and management of this increasingly appreciated mechanism of resistance. **2020**, *9*, x FOR PEER REVIEW 12 of 15

**Figure 3.** Analysis of recommendations for FUBC sampling in the setting of Gram negative bacteremia on the basis of microbiological factors and efficacy of antibiotic therapy. **Note***:* \* MDR definition by Magiorakos AP et al., 2012 [33]; ESBL, extended-spectrum beta-lactamases. References: ineffective therapy in the preceding 24 h [18,21–23]; effective therapy in the preceding 24 h [18,21–24]; preceding positive FUBC [26]; MDR [22,23,25,26]; ESBL+, Serratia spp. and P. aeruginosa [23–25]; antibiotic susceptible pathogen [22,23,25,26]. **Figure 3.** Analysis of recommendations for FUBC sampling in the setting of Gram negative bacteremia on the basis of microbiological factors and efficacy of antibiotic therapy. **Note:** \* MDR definition by Magiorakos AP et al., 2012 [33]; ESBL, extended-spectrum beta-lactamases. References: ineffective therapy in the preceding 24 h [18,21–23]; effective therapy in the preceding 24 h [18,21–24]; preceding positive FUBC [26]; MDR [22,23,25,26]; ESBL+, Serratia spp. and P. aeruginosa [23–25]; antibiotic susceptible pathogen [22,23,25,26].

#### **4. Materials and Methods 4. Materials and Methods**

We searched in the Pubmed database for articles addressing the use of FUBCs in patients with GNB. The following search strategy was adopted: "((FUBCs) or (Follow-up blood cultures) or (follow-up blood culture) or (repeat blood cultures)) and ((gram-negative) or (gram negative rod)) and ((bacteremia) or (BSI) or (bloodstream infection))". The research yielded 102 results. Two reviewers independently assessed the titles and abstracts to identify papers that fulfilled the inclusion criteria: (1) clinical studies; (2) studies that included human subjects; and (3) studies that evaluated the utility of FUBCs in patients with GNB. Full texts of studies assessed as relevant or unclear were evaluated. Studies that only discussed either GNB or FUBCs were excluded. We also examined the bibliographic references of articles to identify any relevant studies that were not identified in the initial literature search. Eleven articles were selected, compared, and critically evaluated (Figure 4). We searched in the Pubmed database for articles addressing the use of FUBCs in patients with GNB. The following search strategy was adopted: "((FUBCs) or (Follow-up blood cultures) or (follow-up blood culture) or (repeat blood cultures)) and ((gram-negative) or (gram negative rod)) and ((bacteremia) or (BSI) or (bloodstream infection))". The research yielded 102 results. Two reviewers independently assessed the titles and abstracts to identify papers that fulfilled the inclusion criteria: (1) clinical studies; (2) studies that included human subjects; and (3) studies that evaluated the utility of FUBCs in patients with GNB. Full texts of studies assessed as relevant or unclear were evaluated. Studies that only discussed either GNB or FUBCs were excluded. We also examined the bibliographic references of articles to identify any relevant studies that were not identified in the initial literature search. Eleven articles were selected, compared, and critically evaluated (Figure 4).

**Figure 4.** Flow chart of article selection process. References of eleven selected articles: [15,17–26].

susceptible pathogen [22,23,25,26].

**4. Materials and Methods** 

initial literature search. Eleven articles were selected, compared, and critically evaluated (Figure 4).

**Figure 3.** Analysis of recommendations for FUBC sampling in the setting of Gram negative bacteremia on the basis of microbiological factors and efficacy of antibiotic therapy. **Note***:* \* MDR definition by Magiorakos AP et al., 2012 [33]; ESBL, extended-spectrum beta-lactamases. References: ineffective therapy in the preceding 24 h [18,21–23]; effective therapy in the preceding 24 h [18,21–24]; preceding positive FUBC [26]; MDR [22,23,25,26]; ESBL+, Serratia spp. and P. aeruginosa [23–25]; antibiotic

We searched in the Pubmed database for articles addressing the use of FUBCs in patients with GNB. The following search strategy was adopted: "((FUBCs) or (Follow-up blood cultures) or (follow-up blood culture) or (repeat blood cultures)) and ((gram-negative) or (gram negative rod)) and ((bacteremia) or (BSI) or (bloodstream infection))". The research yielded 102 results. Two reviewers independently assessed the titles and abstracts to identify papers that fulfilled the inclusion criteria: (1) clinical studies; (2) studies that included human subjects; and (3) studies that evaluated the utility of FUBCs in patients with GNB. Full texts of studies assessed as relevant or unclear were evaluated. Studies that only discussed either GNB or FUBCs were excluded. We also examined the

**Figure 4.** Flow chart of article selection process. References of eleven selected articles: [15,17–26]. **Figure 4.** Flow chart of article selection process. References of eleven selected articles: [15,17–26].

#### **5. Conclusions**

The usefulness and the drawbacks of FUBCs in GNBs have been largely investigated in the last few years. Of course, some clinical and microbiological factors define the settings where FUBCs exert their maximum capacity to detect PB. Expert clinicians and a correct selection of high-risk patients could make the difference in terms of the efficiency of this diagnostic tool. Furthermore, a targeted and optimized selection of the occasions where to draw FUBCs may also provide a positive impact on patients' management and outcomes. Of note, valuable insights about outcomes of patients where FUBCs were performed remain poor and should be further investigated.

**Author Contributions:** Conceptualization, M.V.; methodology, M.V., F.C.D., A.C., L.V.; formal analysis, F.C.D., A.C., L.V., A.O.; investigation, F.C.D., A.C., L.V.; writing—original draft preparation, F.C.D., A.C, L.V.; writing—review and editing, G.C., A.O., M.V.; supervision, A.O., M.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

#### **Impact and Sustainability of Antibiotic Stewardship in Pediatric Emergency Departments: Why Persistence Is the Key to Success**

**Elisa Barbieri 1,\* , Maia De Luca <sup>2</sup> , Marta Minute <sup>3</sup> , Carmen D'Amore <sup>4</sup> , Marta Luisa Ciofi Degli Atti <sup>4</sup> , Stefano Martelossi <sup>3</sup> , Carlo Giaquinto <sup>1</sup> , Liviana Da Dalt <sup>5</sup> , Theoklis Zaoutis <sup>6</sup> and Daniele Dona <sup>1</sup>**


Received: 15 October 2020; Accepted: 30 November 2020; Published: 4 December 2020

**Abstract:** Antibiotic stewardship programs proved to be effective in improving prescribing appropriateness. This multicenter quasi-experimental study, aimed to assesses the stewardship impact on antibiotics prescribing in different semesters from 2014 to 2019 in three pediatric emergency departments (Center A, B, and C) in Italy. All consecutive patients diagnosed with acute otitis media or pharyngitis were evaluated for inclusion. Two different stewardship were adopted: for Center A and B, clinical pathways were implemented and disseminated, and yearly lectures were held, for Center C, only pathways were implemented. Broad-spectrum prescription rates decreased significantly by 80% for pharyngitis and 29.5 to 55.2% for otitis after the implementation. In Center C, rates gradually increased from the year after the implementation. Amoxicillin dosage adjusted to pharyngitis recommendations in Center C (53.7 vs. 51.6 mg/kg/die; *p* = 0.011) and otitis recommendations in Center A increasing from 50.0 to 75.0 mg/kg/die (*p* < 0.001). Days of therapy in children < 24 months with otitis increased from 8.0 to 10.0 in Center A, while in older children decreased in Center A (8.0 vs. 7.0; *p* < 0.001) and Center B (10.0 vs. 8.0; *p* < 0.001). Clinical pathways combined with educational lectures is a feasible and sustainable program in reducing broad-spectrum antibiotic prescribing with stable rates over time.

**Keywords:** antibiotic stewardship; pharyngitis; acute otitis media; clinical pathways; children; emergency departments; antibiotic use; prescribing appropriateness

#### **1. Introduction**

Antibiotics remain the most commonly prescribed drugs in the pediatric population [1], with pharyngitis and acute otitis media (AOM) accounting for more than half of the prescriptions in the emergency departments and primary care practices [2,3], with an overprescribing of broad-spectrum antibiotics.

Both conditions have a viral and a bacterial etiology: AOM is mostly caused by Streptococcus pneumoniae, non-typeable *Haemophilus influenzae*, and *Moraxella catarrhalis* [4], while around 20% of pharyngitis are caused by Group A β-hemolytic streptococcus [5].

Although most of the pathogens remain sensible to first line amoxicillin, co-amoxiclav and III-gen cephalosporins are, respectively, prescribed in around 30% and 15% of AOM primary care cases, and in more than 24% and 15% of Group A streptococcus pharyngitis cases [3].

To decrease or reverse this trend, various antibiotic stewardship programs (ASPs) have been implemented worldwide, focusing on different approaches [6].

Clinical pathways have proven to be a feasible and efficient first step in improving prescribing appropriateness, especially in settings where funding is limited [7–10]. A clinical pathway is a task-oriented plan designed to support the implementation of clinical guidelines and protocols in primary care and inpatient settings.

In October 2015, an ASP based on clinical pathways was implemented in the pediatric emergency department of Padova University Hospital. Preliminary results reported an increase in "wait-and-see" approach rate for AOM (21.7% vs. 33.1%) and an increase in narrow-spectrum antibiotics treatments for both AOM (32.0% vs. 51.6%) and pharyngitis (53.6% vs. 93.4%), with no variation in treatment failures [10].

While there is not yet a consensus on the most effective ASP—especially in terms of settings and costs—we aimed to evaluate the efficacy and sustainability over time of ASPs based on clinical pathways with and without yearly educational lectures in three pediatric emergency departments.

#### **2. Results**

#### *2.1. Pharyngitis*

During the study, 4534 pharyngitis episodes were evaluated, accounting for around 3% of total pediatric emergency department visits. In total, 3249 episodes were included; the demographic characteristics of children included were similar with respect to sex, with a higher prevalence among older children in both Center B and C (Supplementary Materials, Table S2).

In Center A and B amoxicillin prescriptions rate increased (Center A: from 53.6% to 98.5%; *p* < 0.001, Center B: from 69.7% to 96.0%, *p* < 0.001) with a consequent decrease in broad-spectrum-antibiotic prescription rates (broad-spectrum antibiotic prescriptions) (Table 1).

The interrupted time series model strongly suggest a broad-spectrum antibiotic prescriptions reduction following the introduction of the clinical pathways by 81.6% (relative risk (RR) 0.184 (95% CI: 0.072–0.471); *p* = 0.002) for Center A and by 88.6% (RR 0.114 (95% CI: 0.016–0.816); *p* = 0.0471) for Center B, as illustrated in Figure 1.

The interrupted time series model shows a 77% reduction (RR 0.230 (95% CI: 0.167–0.316); *p* < 0.001) in broad-spectrum antibiotic prescriptions rates after the intervention with rates increasing monthly by 3.5% (RR 1.035 (95% CI: 1.025–1.045); *p* < 0.001) in the post periods in Center C.

Amoxicillin dosage adjusted from 53.7 (IQR:7.0) to 51.6 mg/kg/die (IQR:3.8) in Center B (Table 1 and pair-wise comparison in Supplementary Materials, Figure S1) and the median days of therapy (DOT) met the recommended 10 days (8.0 vs. 10.0; *p* < 0.001) after clinical pathways implementation in Center A (Table 1 and pair-wise comparison in Supplementary Materials, Figure S2).

*Antibiotics* **2020**, *9*, 867


**Table 1.** Treatment option for pharyngitis in the different periods in the three Centers.

Center B, as illustrated in Figure 1.

The interrupted time series model strongly suggest a broad-spectrum antibiotic prescriptions reduction following the introduction of the clinical pathways by 81.6% (relative risk (RR) 0.184 (95%

**Figure 1.** Interrupted time series of monthly broad-spectrum antibiotics prescriptions (dots) expressed as percentages with 95% confidence intervals (bars) for (**A**) pharyngitis, (**B**) acute otitis media, and (**C**) non-complicated acute otitis media in the three centers. The lines represent the broadspectrum prescriptions trend in the different centers. **Figure 1.** Interrupted time series of monthly broad-spectrum antibiotics prescriptions (dots) expressed as percentages with 95% confidence intervals (bars) for (**A**) pharyngitis, (**B**) acute otitis media, and (**C**) non-complicated acute otitis media in the three centers. The lines represent the broad-spectrum prescriptions trend in the different centers.

#### *2.2. Acute Otitis Media—Total*

Overall, 3980 AOM visits were assessed, and 3039 met the inclusion criteria. In Center A and C, a significant difference was reported in age class in the various periods (Supplementary Materials, Table S2).

After ASP implementation, "wait-and-see" approach rates were higher in Center A (from 21.6% to 34.1%; *p* = 0.006) and amoxicillin prescriptions rates increased in Center A and B, with a concomitant decrease in broad-spectrum antibiotic prescriptions rates (Center A: from 67.3% to 38.1%; *p* < 0.001, Center B: from 56.6% to 33.6%; *p* < 0.001), especially cephalosporins prescriptions (Table 2).

In Center C, the highest "wait-and-see" approach rate (32.7%) and the lowest broad-spectrum antibiotic prescriptions rate (68.3%) were reported in the semester following the ASP. Initially, the intervention doubled the "wait-and-see" approach rates (RR 2.510 (95% CI: 1.832–3.349); *p* < 0.001), even if rates did not remain stable, but decreased by 3.6% monthly (RR 0.036 (95% CI: 0.964–0.955); *p* < 0.001).

The broad-spectrum antibiotic prescriptions rates decreased by 29.5% (RR 0.705 (95% CI: 0.538–0.923); *p* = 0.011) in Center A and by 55.2% (RR 0.448 (95% CI: 0.235–0.856); *p* = 0.015) in Center B after the clinical pathways implementation. In Center C, the intervention reduced broad-spectrum antibiotic prescriptions by 41.1% (RR 0.589 (95% CI: 0.470–0.737); *p* < 0.001), but in the following semesters, the broad-spectrum antibiotic prescriptions rates increased by 1.5% monthly (RR 1.015 (95% CI 1.009–1.022); *p* < 0.001). The interrupted time series are shown in Figure 1.

Amoxicillin dosage increased from 50.0 (IQR:0.0) to 75.0 (IQR:5.0) mg/kg/die in Center A (*p* < 0.001), similarly to Center B (from 56.7 (IQR:14.3) to 75.0 (IQR:7.1) mg/kg/die). Co-amoxiclav dosage increase was also significant for both centers (Supplementary Materials, Figure S3).

DOT in children <24 months varied significantly from 8.0 (IQR:2.0) to 10.0 (IQR:0.0) just in Center *Antibiotics* **2020**, *9*, x FOR PEER REVIEW A (Figure 2). 8 of 19

**Figure 2.** Distribution of days of therapy for non-complicated acute otitis media in the different periods in Center A (**A**,**D**), Center B (**B**,**E**), and Center C (**C**,**F**) stratified by age class (<24 months: **A**,**B**,**C**; ≥24 months: **D**,**E**,**F**) with pair-wise comparison. The dots represent the granular data, horizontal lines are median and IQR; whiskers extend to the minimum and maximum within 1.5 times the IQR. Violin plots present quantifications. The dotted green line represents the DOT recommended in the clinical pathways for acute otitis media in children <24 months, while the dotted blue line represents clinical pathways recommended DOT for non-complicated AOM in children ≥24 months and the dotted red line represents clinical pathways recommended DOT for complicated AOM in children ≥24 months . **Figure 2.** Distribution of days of therapy for non-complicated acute otitis media in the different periods in Center A (**A**,**D**), Center B (**B**,**E**), and Center C (**C**,**F**) stratified by age class (<24 months: **A**–**C**; ≥24 months: **D**–**F**) with pair-wise comparison. The dots represent the granular data, horizontal lines are median and IQR; whiskers extend to the minimum and maximum within 1.5 times the IQR. Violin plots present quantifications. The dotted green line represents the DOT recommended in the clinical pathways for acute otitis media in children <24 months, while the dotted blue line represents clinical pathways recommended DOT for non-complicated AOM in children ≥24 months and the dotted red line represents clinical pathways recommended DOT for complicated AOM in children ≥24 months.



**Table 2.** Treatment option for acute otitis media in the different periods in the three centers.

In older children, a variation was noted in Center A (from 8.0 (IQR:1.0) to 7.0 (IQR:2.0) DOT; *p* < 0.001) and in Center B (from 10.0 (IQR:2.0) to 8.0 (IQR:5.0) DOT; *p* < 0.001), while in Center C the median DOT remained 7.0 for the all periods in both age class.

#### *2.3. Acute Otitis Media—Sensitivity Analysis*

In 58.0% (2310/3980) of the AOM diagnoses, there was no sign of otorrhea reported.

In Center A and Center B, co-amoxiclav prescriptions decreased after clinical pathway introduction (Center A: from 43.6% to18.6%; *p* < 0.001, Center B: from 33.3% to 19.7%; *p* = 0.035), whereas a pattern similar to total AOM prescription was noted for Center C (Supplementary Materials, Table S3).

The "wait-and-see" approach rates increased by 68.8% after the intervention and stabilized in Center A (RR 1.688 (95% CI: 1.116–2.552); *p* = 0.013), while in Center C, the intervention was significant in increasing the "wait-and-see" approach rates immediately after the intervention (RR: 2.363 (95% CI: 1.774–3.148); *p* < 0.001); then, rates decreased monthly by 3.3% (RR: 0.967 (95% CI: 0.959–0.976); *p* < 0.001)).

The interrupted time series analysis (Figure 1) confirmed previous findings for Center A and Center C on broad-spectrum antibiotic prescriptions rates, but the reduction by 54.1% (RR 0.459 (95% CI: 0.198–1.060)) in Center B was non-significant (*p* = 0.068).

Amoxicillin and co-amoxiclav dosages did not differ from total AOM findings (Supplementary Materials, Figure S4), whereas median DOT for older children decreased significantly from 8.0 (IQR:2.5) to 5.0 (IQR:2.0) DOT (Supplementary Materials Figure S5) in Center A.

#### **3. Discussion**

In this multicentric study, clinical pathways combined with educational lectures proved to be a feasible ASP in decreasing and maintaining broad-spectrum antibiotic prescriptions rates over time for both AOM and pharyngitis (Center A and B). On the other hand, clinical pathways alone failed to maintain the low broad-spectrum prescription rates achieved after the intervention (Center C).

The ASPs reduced co-amoxiclav and cephalosporins prescription rates for pharyngitis by around 80% in all centers. In Center C, the reduction was not maintained after the first semester, possibly due to prescribers' fear of coinfection in Group A Streptococcus carriers [11]. DOT changes reflected prescriptions variation in Center A and Center C, reaching the 10 DOT suggested for amoxicillin in order to decolonize the oropharynx in the Center A. It is unlikely that the recommended duration of amoxicillin is less than the clinical pathway indications, even if penicillin V administered four times daily for five days could represent an alternative regimen in older children with Group A Streptococcus pharyngitis, where commercially available [12,13].

"Wait-and-see" approach rates increased significantly in non-complicated AOM with stable rates after the ASPs implementation. The "wait-and-see" approach was introduced in the early 2000s and was included in most guidelines for the treatment of AOM [14,15] with variation in its applicability. The clinical pathways suggested a "wait-and-see" approach in children with non-severe AOM: if aged 6–24 months just with the unilateral form. In all cases, parents' compliance and the possibility of a follow-up 48–72 h after was needed. The lack of rate variation in Center B could imply that emergency physicians did not feel that it was an appropriate strategy for their setting [16]. Similar to pharyngitis, ASPs were effective in broad-spectrum antibiotic prescriptions rate reduction for AOM. Nonetheless, in Center C, broad-spectrum antibiotic prescriptions rates for non-complicated AOM raised in the last semesters, increasing the possibility to cause more adverse events (i.e., vomiting, diarrhea, rash), in a country where only fixed 7:1 ratio packages are marketed, and the risk of selection of resistant bacteria in the community [17–19]. The same might be the reason why in Center B a more cautious behavior was noted for administering co-amoxiclav, especially at high dosage. Literature findings suggest that a higher co-amoxiclav ratio seems to be associated with fewer side effects without reducing clinical efficacy, but clinical pathways are based on the available medicine formulary to be easily adaptable.

In non-complicated AOM, five median DOT for older children was achieved in Center A's final semester, reflecting prescribers' initial discomfort with short-course treatment. Overall median DOT might not be the most suitable indicator in assessing ASP efficacy on treatment duration when higher rates of second-line therapy (i.e., cephalosporins) are observed, such as in Center C. A possible solution is calculating the median DOT stratified by different drugs.

Our study has several caveats: first, its retrospective nature and difficulties in assessing the reasons why broad-spectrum antibiotics were prescribed. However, the same study nature allowed us to exclude the Hawthorne effect that sometimes could be argued to play a major role in the ASP success. Second, no control groups were selected; hence, we cannot conclude with a high degree of certainty that the variation in broad-spectrum antibiotic prescriptions rates is caused by the ASP implementation without considering other possible explanations. On the other hand, no policy restricting antibiotic prescriptions was implemented in the different centers during the study period, nor were there any shortages in Group A Streptococcus-rapid tests. Third, patient follow-up was beyond the aim of the study and was, therefore, not performed. Although it is possible that patients receiving broad-spectrum antibiotic prescriptions had better clinical outcomes, preliminary results reported that there was no difference in treatment failure rate nor in adverse event rate in the first and second semesters [10]. Fourth, even if the first part of clinical pathways was focused on diagnosis, with a particular emphasis on signs to be considered, clinicians were not tested on the use of pneumatic otoscope nor Group A Streptococcus -rapid tests. Finally, it can be argued that differences in prescriptions may reflect variations in local bacteria resistance, but clinical pathways were developed with microbiologists from different centers, and clinicians were able to adapt them according to local microbiological data.

Despite clinical pathways proving to be a feasible ASP tool with rapid implementation and reduced applicability cost, this study revealed that without combining it with continuous education, it might have no lasting effect. In fact, the Infectious Disease Society of America recommends two core strategies to be implemented together, even if most pediatric ASPs consist of just one intervention. According to a recent systematic review, most of the studies do not report a long follow up, and few report negative results, though publication bias might contribute to this [6].

A study trial where an ASP based on continuous clinician-specific education combined with audit and feedback was implemented in the USA outpatients setting found that following the removal of the audit and feedback, the initial reduction in broad-spectrum antibiotic prescriptions to children with acute respiratory tract infection was lost [20]. The authors believed that audit and feedback was a vital element of the ASP and continuous, active efforts are required to sustain initial improvements in prescribing attitude. Moreover, another trial in a similar setting, comparing different ASPs, found out that only the peer-comparison approach maintained prescription rates lower than the control group after stopping the interventions [21].

Clinicians and researchers interested in implementing an ASP should carefully consider their options in order to avoid inefficiency. A possible improvement in the ASPs proposed lay on the analysis timing. Data were manually collected in condition-specific data collection forms, requiring ad hoc specialists to perform data entry. One solution could be conducting random day-point prevalence surveys every couple of weeks or months in setting with rapid patients turn-over, thus limiting the time dedicated to the collection and providing more rapid estimates on prescribing behaviors [22–24]. Secondly, having IT support to aid in developing real-time indicators will allow for rapid intervention and identification of root causes in cases of prescriber non-adherence to the ASPs [25]. Lastly, in our study, we assessed ASPs' impact and sustainability for conditions with a higher incidence in the cold season, and for this reason, lectures were specifically held in the first months of the season; in the case of developing an ASP for a condition with no such seasonal variation (i.e., sepsis), the ASP team could opt for lectures closer together in time.

#### **4. Materials and Methods**

#### *4.1. Study Design*

This multicenter quasi-experimental study assesses the ASPs impact on antibiotics prescribing in the pediatric emergency department of three different hospitals: two tertiary-level university hospitals (Azienda Ospedale-Università, Padova and Ospedale Pediatrico Bambino Gesù, Rome, having around 24,000 and 56,000 yearly emergency room visits, respectively) and one secondary-level hospital (Ospedale Ca' Foncello, Treviso, having around 14,000 yearly emergency room visits). Each institution was randomly named with a capital letter (Center A, B, C) to keep them intentionally anonymous.

The different periods considered were: one semester before and three semesters after implementation for Center A and C (Center A: 15 October 2014–15 April 2015; 15 October 2015–15 April 2016; 15 October 2016–15 April 2017; 15 October 2017–15 April 2018; Center C: 1 January 2016–30 June 2016; 1 January 2017–30 June 2017; 1 January 2018–30 June 2018; 1 January 2019–30 June 2019) and one semester before and two semesters after implementation for Center B (1 January 2017–30 June 2017; 1 January 2018–30 June 2018; 1 January 2019–30 June 2019). The study flowchart is shown in Figure 3. *Antibiotics* **2020**, *9*, x FOR PEER REVIEW 12 of 19

**Figure 3.** Study flow chart of the ASPs' implementation in the three centers. **Figure 3.** Study flow chart of the ASPs' implementation in the three centers.

During the study period, in the pediatric emergency departments of the three hospitals, there were physicians that worked on a daily basis, specialist consultants that worked depending on the requests and residents that worked on a daily basis but changed every couple of months.

#### *4.2. Intervention*

Two different ASPs were adopted: for Center A and B, the intervention consisted of clinical pathway implementation and dissemination as laminated pocket-cards with yearly educational lectures for residents and pediatricians, and for Center C, the intervention consisted only of the implementation of clinical pathways with the possibility of consulting in the hospital intranet.

A multidisciplinary group of experts from each center, in collaboration with the Division of Pediatric Infectious Diseases of the Children's Hospital of Philadelphia, developed the clinical pathways for pharyngitis and AOM that were adapted to the centers' standard of care with no changes in the algorithm (Appendix A Figure A1 and Appendix B Figure A2).

The educational lectures addressed to residents, structured physicians and specialists and held by the center ASP team, consisted of two hours of training on the diagnosis and treatments of AOM and pharyngitis with a focus on the rational for antibiotic prescribing.

#### *4.3. Population and Case Definition*

All consecutive patients aged two months to 14 years with an International Classification of Diseases, 9th Revision, Clinical Modification code, or descriptive diagnosis of AOM or pharyngitis admitted to the pediatric emergency department in one of the three centers were included.

General exclusion criteria were immunodeficiency or immunosuppressive therapy, concomitant bacterial infections or systemic bacterial infection, craniofacial abnormalities, chronic diseases (i.e., diabetes, cystic fibrosis), and ongoing antibiotic therapy at admission. Pharyngitis exclusion criteria were previous tonsillectomy, periodic fever, aphthous stomatitis, pharyngitis and adenitis syndrome, and admission to the pediatric emergency department for feeding difficulties. AOM exclusion criteria were tympanostomy tubes at the time of diagnosis, chronic otitis media, and AOM complicated by mastoiditis.

Pediatric emergency department visits occurring for the same patient greater than 30 days apart were analyzed as separate events.

All AOM episodes with otorrhea were considered as complicated AOM; the remaining episodes were considered as non-complicated.

Broad-spectrum antibiotics were defined as β-lactam and β-lactamase inhibitor combinations, second- and third-generation cephalosporins, fluoroquinolones, and macrolides. Topic antibiotics (i.e., ciprofloxacin ear drops) were not considered.

#### *4.4. Outcomes*

The following aspects of antibiotic prescriptions for pharyngitis and AOM were assessed:


The "wait-and-see" approach was defined as AOM episodes with no antibiotic prescription.

#### *4.5. Data Collection and Sample Size Calculation*

All clinical, demographic, diagnostic, and prescription data were manually collected from electronic medical records, using a password protected REDCap 10.0.1-© 2020 (Vanderbilt University) data collection form and stored on a secured server at the University of Padova. Privacy was guaranteed by assigning each patient a unique study-specific number and not collecting personally identifying data.

Assuming that before ASP implementation, (i) in 10% of AOM episodes a "wait-and-see" approach would be chosen and in 45% of pharyngitis episodes no antibiotic would be prescribed, (ii) the broad-spectrum antibiotic prescriptions rates would be 50%, (iii) broad-spectrum antibiotic prescriptions would decrease by 25%, (iv) 15% of the episodes did not fulfill the inclusion criteria, (v) a two-tailed Type I error of 0.05 is used, and (vi) the study is required to have at least a power of 70%, we estimated a minimum sample size of 330 pharyngitis and 260 AOM episodes per period per center to detect a significant decrease in broad-spectrum antibiotic prescriptions. The power for estimating the difference between independent proportions was calculated using G Power 3.1.9.4-© 1992–2019 (Universitat Kiel, Germany) [26].

The investigations were carried out following the rules of the Declaration of Helsinki of 1975 (https://www.wma.net/what-we-do/medical-ethics/declaration-of-helsinki/), revised in 2013. This study was approved by the Ethical Committees of all Centers (3737/AO/16). Due to the nature of the study (observational retrospective), no informed consent was required from the patients.

#### *4.6. Data Analysis*

Single center results in the different periods were summarized as numbers and percentages (categorical variables) and as median and interquartile range (continuous variables). Categorical variables were compared with χ2 or Fisher's 2-tailed exact test in a contingency table *r* x *c*; a Fisher test was used when the value in any of the cells of the contingency table was below five. Continuous variables were compared with a non-parametric Kruskal-Wallis rank sum test; for pair-wise comparisons, we used Dwass-Steele-Critchlow-Fligner all-pairs test adjusted with Holm method. Since different DOT are recommended depending on child age, DOT analysis was stratified according to age class (2–23 months of age vs. 2–14 years of age).

An interrupted time series analysis supposing an abrupt step change in monthly significative outcomes (1 and 2) using quasi-Poisson regression models was used to determine the effect of the intervention. [27]"Wait-and-see" approach, log-transformed total AOM episodes, a variable representing the frequency in months in which observations were taken, and a dummy variable indicating the pre- and post-intervention periods were considered. For outcome (2), broad-spectrum antibiotic prescriptions and log-transformed total antibiotic prescriptions were considered together with a frequency variable and a dummy variable previously specified. A seasonal adjustment was not necessary since the same calendar months were considered to control for effects. Autocorrelation was assessed, examining the plot for residuals and the partial autocorrelation function. The corresponding relative risk and 95% confidence interval (95% CI) according to normal approximation were calculated.

Outcome data were sometimes missing (0–20%, Supplementary Materials, Table S1). If variable data were missing completely at random [28] and restricting the analysis would not have resulted in a significant loss of information or biased estimation, listwise deletion was performed (i.e., dosage); in the opposite case (i.e., DOT), group-wise predictive mean matching within the fully conditional specification algorithm was used to fit the missing data [29].

A sensitivity analysis was conducted for non-complicated AOM episodes. Data were analyzed using R statistical software (version 3.6.3, Vienna, Austria) for Windows [30]. The multiple imputation was performed with the "mice" and "miceadds" packages [31]. Figures were created with the packages "ggplot2" [32] and "ggstatsplot" [33]. For brevity, statistical parameters were included in figures displaying pair-wise comparisons. Statistical significance was set at the 0.05 level and *p* values were two-sided.

#### **5. Conclusions**

To the best of our knowledge, this is the first attempt to study the efficacy and sustainability of ASPs over time based on clinical pathways in pediatric emergency departments. Our findings suggest that clinical pathways paired with continuous education can be effective in reducing broad-spectrum antibiotic prescription and in reaching target treatment duration. Researchers should push for efficient assessment and publication of intervention sustainability in order to help other clinicians in choosing the most suitable ASP for their setting.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6382/9/12/867/s1, Table S1. Missing data in the different periods in the three centers (A, B, C); Table S2. Demographic characteristics of the included and excluded patients with pharyngitis and AOM in the different periods in the three centers (A, B, C); Table S3. Treatment option for non-complicated AOM in the different periods in the three centers (A, B, C); Figure S1. Distribution of amoxicillin dosage for pharyngitis among different periods in Center A and B with pairwise comparison; Figure S2. Distribution of days of therapy for pharyngitis among different periods in the three centers (A, B, C) with pairwise comparison; Figure S3. Distribution of co-amoxiclav (A, B) and amoxicillin (C, D) dosage for acute otitis media among different periods in Center A (A, C) and Center B (B, D) with pairwise comparison; Figure S4. Distribution of co-amoxiclav (A, B) and amoxicillin (C, D) dosage for non-complicated acute otitis media among different periods in Center A (A, C) and Center B (B, D) with pairwise comparison; Figure S5. Distribution of days of therapy (DOT) for acute otitis media in the different periods in Centre A (A, D), Centre B (B, E) and Centre C (C, F) stratified by age class (<24 months: A, B, C; ≥24 months: D, E, F) with pair wise comparison.

**Author Contributions:** Conceptualization, D.D., T.Z., L.D.D., and C.G.; methodology, D.D., T.Z., L.D.D., and C.G.; software, E.B.; validation, E.B., M.D.L., and M.M.; formal analysis, E.B.; investigation, D.D., M.D.L., and M.M.; resources, C.G., L.D.D., C.D., M.L.C.D.A., and S.M.; data curation, E.B., M.D.L., and M.M.; writing—original draft preparation, E.B.; writing—review and editing, D.D., T.Z., M.D.L., and M.M.; visualization, E.B. and D.D.; supervision, T.Z., L.D.D., C.D., M.L.C.D.A., S.M., and C.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The authors would like to thank Giulia Brigadoi, Giulia Cesca, Sofia Mezzalira, Antonino Reale, Livia Gargiullo, Lorenza Romani, Simona Pipino, Nicole Colantoni, and Anna Maselli for collecting the data, and Lindsey Hunter for helping to edit the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest. *Antibiotics* **2020**, *9*, x FOR PEER REVIEW 16 of 19

#### **Appendix A**

**Appendix A** 

**Figure A1.** Clinical pathway for acute otitis media for children older than 30 days of age. **Figure A1.** Clinical pathway for acute otitis media for children older than 30 days of age.

*Antibiotics* **2020**, *9*, 867

#### **Appendix B**

**Appendix B**

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 17 of 19

**Figure A2.** Clinical pathway for acute pharyngitis for children older than 30 days of age. **Figure A2.** Clinical pathway for acute acute pharyngitis for children older than 30 days of age.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

#### **E**ff**ectiveness and Acceptance of Multimodal Antibiotic Stewardship Program: Considering Progressive Implementation and Complementary Strategies**

**Flavien Bouchet 1,2,\*, Vincent Le Moing <sup>1</sup> , Delphine Dirand <sup>2</sup> , François Cros <sup>3</sup> , Alexi Lienard <sup>4</sup> , Jacques Reynes <sup>1</sup> , Laurent Giraudon <sup>2</sup> and David Morquin <sup>1</sup>**


Received: 31 October 2020; Accepted: 25 November 2020; Published: 27 November 2020

**Abstract:** Multiple modes of interventions are available when implementing an antibiotic stewardship program (ASP), however, their complementarity has not yet been assessed. In a 938-bed hospital, we sequentially implemented four combined modes of interventions over one year, centralized by one infectious diseases specialist (IDS): (1) on-request infectious diseases specialist consulting service (IDSCS), (2) participation in intensive care unit meetings, (3) IDS intervention triggered by microbiological laboratory meetings, and (4) IDS intervention triggered by pharmacist alert. We assessed the complementarity of the different cumulative actions through quantitative and qualitative analysis of all interventions traced in the electronic medical record. We observed a quantitative and qualitative complementarity between interventions directly correlating to a decrease in antibiotic use. Quantitatively, the number of interventions has doubled after implementation of IDS intervention triggered by pharmacist alert. Qualitatively, these kinds of interventions led mainly to de-escalation or stopping of antibiotic therapy (63%) as opposed to on-request IDSCS (32%). An overall decrease of 14.6% in antibiotic use was observed (*p* = 0.03). Progressive implementation of the different interventions showed a concrete complementarity of these actions. Combined actions in ASPs could lead to a significant decrease in antibiotic use, especially regarding critical antibiotic prescriptions, while being well accepted by prescribers.

**Keywords:** antibiotic stewardship program; complementarity; prospective audit and feedback

#### **1. Introduction**

In 2015 in Europe, 671,689 cases of infections with antibiotic-resistant bacteria features occurred, leading to 33,110 deaths, corresponding to 6.44 deaths per 100,000 population and 874,541 disability adjusted life-years (DALYs) [1]. Without any practical measures, the current state could worsen exponentially with 390,000 deaths every year expected in Europe by 2050. Moreover, this concerning healthcare issue also represents a dramatic economic burden; that is, if the antibiotic-resistant bacteria infection rate remains at the same level as today, this could lead to a loss of 100 trillion of USD

worldwide [2,3]. Implementing antibiotic stewardship programs (ASPs) in hospitals is a major way to improve this issue [4–7]. Many studies have shown the positive impact of antibiotic stewardship programs (ASPs) on antibiotic use and antibiotic resistance, improvement of morbidity and mortality, reduction of *Clostridium di*ffi*cile* infections incidence, and health costs savings [7–13]. One of the key points of ASP success is to gather a multidisciplinary team including pharmacists, microbiologists, and infectious diseases physicians with a specific time dedicated to this task [5,14,15].

Nowadays, cross-disciplinary medical project funding is limited by the current economic healthcare situation. Despite many warnings from French and European infectious diseases societies about the crucial need for ASPs, raising funds and dedicating time to implement these strategies are still difficult, especially when the short-term economic benefit is not obvious [16–18].

In France, ASPs are not fully implemented and the current system relies on supporting prescribers mainly through training and on-request infectious diseases specialist consulting [19]. Training may be a key point to improve antibiotics prescription, yet a recent multicenter web-based survey brings to light that most final-year European medical students feel they still need more education on antibiotic use for their future practice as junior doctors [20]. In this context, the association of the improvement of medical student training and a more interventionist strategy including microbiological laboratory alerts and prospective audit and feedback (PAF) interventions, such as prescription review with assistance by pharmacists, could be useful [6,21,22]. Indeed, PAF allows clinicians to prescribe any empiric antibiotic regimen, then the ASP can advise the clinician on discontinuing or adjusting therapy after prescription analysis. Although feedback further increased the intervention effect, it is used in only a minority of enabling interventions, as shown in a Cochrane meta-analysis [23]. This study raises the need for new studies to assess different stewardship interventions and to explore the facilitators to implementation.

Indeed, the practical way to link together these interventions is not clear and neither the complementarity of these actions nor the acceptance of physicians towards unsolicited advice have yet been evaluated. Based on recent publications [24–27], we progressively implemented an innovative multimodal ASP in 2018 in a secondary care hospital. The aim of this study was to evaluate the complementarity of different interventions in an ASP and the impact on antibiotic use. We also analyzed the impact on mean of length of stay (LOS), 30-day readmission rate (30-DRR), and mortality. Prescribers' acceptance was also assessed in the perspective of long-term system development.

#### **2. Results**

#### *2.1. Interventions Complementarity*

Over the entire analysis period, 7508 stays involved the administration of antibiotic therapy, of which 1316 received an intervention. At least one intervention was carried out for 1430 stays, corresponding to 2046 interventions noted in the electronic medical records (EMRs) (Figure 1). *Antibiotics* **2020**, *9*, x FOR PEER REVIEW 3 of 15 with a 76% acceptance rate. Only 69 proposals (14%) consisted of a therapeutic escalation or an extension of antibiotic therapy with an 84% acceptance rate (Figure 4).

**Figure 1.** Interventions' distribution. Distribution of the 2046 interventions noted in the electronic patient record related to 1430 hospital stays and 1243 patients. Abbreviations: ICU, intensive care unit; IDSCS, infectious diseases specialist consulting service*.* **Figure 1.** Interventions' distribution. Distribution of the 2046 interventions noted in the electronic patient record related to 1430 hospital stays and 1243 patients. Abbreviations: ICU, intensive care unit; IDSCS, infectious diseases specialist consulting service.

**Figure 2.** Compliance depending on intervention type. Compliance rate of the 2046 interventions noted in electronic patient record. Abbreviations: ICU, intensive care unit; IDSCS, infectious diseases

specialist consulting service**;** N = number of interventions.

The overall acceptance rate for the proposals was 88%, with a variation according to intervention types ranging from 68 to 92% (Figure 2). The distribution analysis of intervention-types normalized to working days on site highlights a true complementarity between interventions (Figure 3). In summary, the implementation of PAF interventions in a second phase widens the ASP field of action without impacting other types of intervention. **Figure 1.** Interventions' distribution. Distribution of the 2046 interventions noted in the electronic patient record related to 1430 hospital stays and 1243 patients. Abbreviations: ICU, intensive care unit; IDSCS, infectious diseases specialist consulting service*.*

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 3 of 15

with a 76% acceptance rate. Only 69 proposals (14%) consisted of a therapeutic escalation or an

extension of antibiotic therapy with an 84% acceptance rate (Figure 4).

**Figure 2.** Compliance depending on intervention type. Compliance rate of the 2046 interventions noted in electronic patient record. Abbreviations: ICU, intensive care unit; IDSCS, infectious diseases specialist consulting service**;** N = number of interventions. **Figure 2.** Compliance depending on intervention type. Compliance rate of the 2046 interventions noted in electronic patient record. Abbreviations: ICU, intensive care unit; IDSCS, infectious diseases specialist consulting service; N = number of interventions. *Antibiotics* **2020**, *9*, x FOR PEER REVIEW 4 of 15

This complementarity is also illustrated when comparing advice type according to different kinds of interventions. Regarding IDSCS interventions, 341 of 1053 interventions (32%) were either a lack of antibiotic initiation, a therapeutic de-escalation, a cessation of all antibiotics, or a reduction in treatment duration with a 92% acceptance rate. On the other hand, 535 proposals (51%) were either a therapeutic escalation or an extension of antibiotic therapy with a 92% acceptance rate.

Of the 501 proposals made during PHARM-cATB interventions, 316 (63%) were either a therapeutic de-escalation, a cessation of all antibiotics, or a reduction in the duration of treatment with a 76% acceptance rate. Only 69 proposals (14%) consisted of a therapeutic escalation or an extension of antibiotic therapy with an 84% acceptance rate (Figure 4). *Antibiotics* **2020**, *9*, x FOR PEER REVIEW 5 of 15

**Figure 4.** Advice type according to different kind of interventions. Distribution of the 2046 interventions noted in the electronic patient record related to 1430 hospital stays and rate of propositions according to different kind of interventions. N = number of interventions. IDSCS = infectious diseases specialist consulting service. **Figure 4.** Advice type according to different kind of interventions. Distribution of the 2046 interventions noted in the electronic patient record related to 1430 hospital stays and rate of propositions according to different kind of interventions. N = number of interventions. IDSCS = infectious diseases specialist consulting service.

#### *2.2. Impact on Mortality, 30-Day Readmission Rate, and Mean Length of Stay 2.2. Impact on Mortality, 30-Day Readmission Rate, and Mean Length of Stay*

There were 3561 inpatients with deep infections hospitalized in the eight wards who benefited from the whole ASP from January 2016 to May 2017 versus 3839 from January 2018 to May 2019. The clinical and demographic characteristics of these patients are summarized in Table 1. There was a downward trend in the mean LOS in patients with deep infections, from an LOS of 11.03 days before the implementation of the system to a LOS of 10.44 days after implementation, but this difference was not statistically significant (*p* = 0.096). Nor was there any significant difference regarding inhospital mortality in patients with deep infections (267 (7.31%) versus 266 (6.9%); *p* = 0.37) or 30-DRR (8.2% (293) versus 7.7% (296) in 2019, *p* = 0.44). There were 3561 inpatients with deep infections hospitalized in the eight wards who benefited from the whole ASP from January 2016 to May 2017 versus 3839 from January 2018 to May 2019. The clinical and demographic characteristics of these patients are summarized in Table 1. There was a downward trend in the mean LOS in patients with deep infections, from an LOS of 11.03 days before the implementation of the system to a LOS of 10.44 days after implementation, but this difference was not statistically significant (*p* = 0.096). Nor was there any significant difference regarding in-hospital mortality in patients with deep infections (267 (7.31%) versus 266 (6.9%); *p* = 0.37) or 30-DRR (8.2% (293) versus 7.7% (296) in 2019, *p* = 0.44).


**Table 1.** Demographic characteristics and pre-existing medical conditions of patients with a diagnosis of deep infection before and after implementation of the antibiotic stewardship program in the eight wards that benefited from all types of interventions.

Results are presented as No (and rate %). All patients with a diagnosis of deep infection regarding the International Classification of Diseases were included. Abbreviation: HIV, human immunodeficiency virus; ICU, intensive care unit.

#### *2.3. Impact on Antibiotic Consumption*

Overall, antibiotic use was significantly decreased by 14.6% in the whole hospital after ASP implementation (336 daily dose of antibiotics per 1000 patient-days (DDD1000PD) in 2017 versus 287 DDD1000PD in 2019; *p* = 0.03). Carbapenems use was moderate and stable over time (from 5 DDD1000PD in 2017 to 4 DDD1000PD in 2019; *p* = 0.82). A slight increase in injectable third-generation cephalosporins use was observed (from 53 DDD1000PD in 2017 to 60 DDD1000PD in 2019; *p* = 0.12). There was a significant decrease of fluoroquinolones use of 63% (51 DDD1000PD in 2017 versus 19 DDD1000PD in 2019; *p* = 0.03) (Figure 5). We also observed a significant decrease in overall antibiotic use for the eight departments included in PAF interventions from 543 DDD1000PD in 2017 versus 474 DDD1000PD in 2019 (*p* = 0.016). Moreover, the reduction in fluoroquinolones use was more noticeable between April and December 2018 (60 DDD1000PD versus 34 DDD1000PD, respectively). This decrease continued until the end of the analysis. An effect on carbapenems use took longer to appear, but a clear decrease was observed from January to June 2019 (13 DDD1000PD versus 7 DDD1000PD, respectively) (Figure 6).

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 7 of 15

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 7 of 15

**Figure 5.** Evolution over time of antibiotic consumption in secondary care hospital of Bassin de Thau. Data are presented as defined daily dose of antibiotics per 1000 patient-days (DDD1000PD). **Figure 5.** Evolution over time of antibiotic consumption in secondary care hospital of Bassin de Thau. Data are presented as defined daily dose of antibiotics per 1000 patient-days (DDD1000PD). **Figure 5.** Evolution over time of antibiotic consumption in secondary care hospital of Bassin de Thau. Data are presented as defined daily dose of antibiotics per 1000 patient-days (DDD1000PD).

**Figure 6.** Evolution over time of antibiotic consumption in the eight units that tested the prospective audit and feedback intervention (antibiotics prescription review). Dotted lines represent initiation of critical antibiotics review in association with pharmacy unit members. Data are presented as defined daily dose of antibiotics per 1000 patient-days. **Figure 6.** Evolution over time of antibiotic consumption in the eight units that tested the prospective audit and feedback intervention (antibiotics prescription review). Dotted lines represent initiation of critical antibiotics review in association with pharmacy unit members. Data are presented as defined daily dose of antibiotics per 1000 patient-days. **Figure 6.** Evolution over time of antibiotic consumption in the eight units that tested the prospective audit and feedback intervention (antibiotics prescription review). Dotted lines represent initiation of critical antibiotics review in association with pharmacy unit members. Data are presented as defined daily dose of antibiotics per 1000 patient-days.

#### *2.4. User Experience Assessment: Satisfaction Survey*

Ninety-five physicians that participated in the ASP were surveyed for satisfaction, of which 49 responded. All physicians were satisfied with the dedicated phone line provided and wanted on-request IDSCS to be continued, as well as LAB-M interventions. Regarding PAF actions, only 2/32 physicians were not satisfied with this kind of intervention and did not wish for it to be carried on. All results are summarized in Supplementary Table S1.

#### **3. Discussion**

We demonstrated that implementation of a whole ASP combining solicited and unsolicited interventions is possible and that the different modes of intervention are complementary. The multimodal ASP implemented in the hospital of Bassin de Thau (HBT) led to a decrease of antibiotic use, especially fluoroquinolones, without impacting deep infection mortality. We also observed a slightly decreased trend in the length of hospital stay in these patients.

Many ASP strategies have already shown their efficacy; for example, development and implementation of facility-specific clinical practice guidelines for common infectious diseases syndromes, IDS systematic referral for *Staphylococcus aureus* bacteremia clinical cases, PAF, preprescription authorization for certain antibiotics, specific interventions depending on infection type or clinical department, microbiological laboratory interventions, and so on [10,11,24,28,29]. To the best of our knowledge, our study is the first to evaluate the potential synergy between different kinds of actions. Our main contribution is to demonstrate that multimodal interventions are synergic. Indeed, opposite to former studies evaluating the impact of specific different antibiotic stewardship interventions, our global approach highlighted the complementarity of each intervention in the success of the holistic ASP. It is noteworthy that this study was set up in the French healthcare system where interfering methods are not developed [19]. Thanks to the progressive implementation of this system, we were able to highlight the complementarity of the interventions. Indeed, the on-request IDSCS, LAB-M, and ICU-M interventions number was stable overtime, even after implementation of PAF strategies, i.e., PHARM-cATB and PHARM-7d review. This highlights a cumulative effect between the different modes of intervention, suggesting that each kind of intervention responded to a specific type of problem. Indeed, on-request IDSCS led to escalation or lengthening of antibiotic therapy in more than 50% of cases, while PHARM-cATB resulted in escalation or lengthening in only 14%. Conversely, PHARM-cATB reviews resulted in de-escalation, stopping, or shortening of antibiotic therapy in 60% of cases. This proportion was even higher within PHARM-7d review. Thus, the different modes of intervention were complementary, both qualitatively and quantitatively.

Antibiotic use analysis also revealed the same pattern. Interestingly, the decreased consumption accelerated after implementation of PAF methods, suggesting a stronger impact of interfering methods on overall antibiotic use. Similar results were found by Tamma et al., whose study highlighted the effectiveness of this method and its major impact on antibiotic use in a cross-over trial [30]. According to these results, a recent retrospective study analyzed the impact of interventionist strategies as PAF or preprescription authorization on fluoroquinolone consumption in 48 U.S. hospitals [31]. Fluoroquinolone use was significantly decreased by 26% over two years between establishment with ASP targeting fluoroquinolone and those with no ASP.

We did not find evidence of any statistically significant differences on mean LOS, 30-DRR, or mortality between the two periods. However, we observed a downward trend in the mean LOS with a decrease of 0.6 days of hospitalization per stay. Indeed, the reduction in antibiotic consumption, particularly intravenous antibiotics, might lead to a reduction in adverse effects and an earlier discharge of patients. The absence of statistical significance might be because of a lack of power for this criteria; however, it was not the primary endpoint of this study. In the literature, some arguments tend to confirm this hypothesis: Sasikumar et al. showed a significant impact of IDS interventions on mortality and medical stay costs, especially for ICU stays [32]. Although there was no significant positive impact on mortality, 30-DRR, and mean LOS in our study, we did not observe any negative impact on

patients' clinical outcomes. Moreover, the high prevalence of patients with chronic respiratory failure, end-stages renal diseases, immunosuppression, and diabetes in the second analysis period could lead to an underestimation of the potential positive impact of our ASP on these outcomes.

These data highlight the importance of using a multimodal strategy when setting up an ASP, keeping in mind that different interventions would respond to different needs. The 2016 IDSA guidelines emphasized PAF and preprescription authorization methods, while underlining the potential for better acceptance of PAF as prescriber autonomy is maintained [24]. In our study, the acceptance rate of PAF intervention was high (79%), despite the fact that unsolicited specialist consulting is not culturally ingrained into the French medical community. Most physicians interviewed in the satisfaction survey agreed that on-request IDSCS and LAB-M actions were improving clinical outcomes and should be continued, whereas only two physicians viewed unsolicited interventions as intrusive to their practice and were reluctant to maintain these methods. PAF acceptance was probably better than expected thanks to its implementation over a second phase of the program, whereas more conventional methods, i.e., on-request IDSCS and LAB-M, were already set up. So, sequential implementation can be identified as a facilitator regarding acceptance of interventionist methods. These results reinforce IDSA recommendations to develop and promote PAF strategies.

We show that they may be implemented within French hospitals considering their efficiency and their complementarity to other methods. It is important to note that, without an EMR, it is challenging to set up such a program with unsolicited interventions.

Nevertheless, this kind of program is time-consuming and labor-intensive; indeed, PAF interventions represented 10 h of work per week for one IDS and one pharmacist, while LAB-M interventions counted for 5 h of work, without including intervention retranscription in the electronic patient record (about 30 min for each intervention, i.e., 25 h weekly). This organization requires dedicated medical time for this activity, as recommended by European Society of Clinical Microbiology and Infectious Diseases (ESCMID) [26].

There are limitations to our study. The impact on antibiotic resistance was not assessed because of the short-term study design. This key outcome will be analyzed after several years of operating under the program in order to compare antimicrobial resistance before and after implementation of this system. Our study did not include medico-economic analysis. Nevertheless, the 0.6 days of stay decrease for inpatients with ID diagnosis would allow some healthcare cost saving, despite this result not being statistically significant. We were also not able to set up, in parallel to our ASP, an educational program that could lead to improved practitioner adherence as well as antibiotics prescribing over the long term [33]. Indeed, in a recent Spanish study, the quality of antimicrobial prescribing improved markedly, and the inappropriate treatment rate was significantly lower over 3 years thanks to regular educational interviews [34]. In addition, we could not evaluate antibiotic prescription at the discharge because of the lack of computerization. Indeed, Vaughn et al. recently highlight that hospital-based stewardship interventions did not affect antibiotic prescription at the discharge [31]. In this study, 14/48 hospitals reported using pre-prescription approval and/or PAF to target fluoroquinolone prescriptions, but hospitals with fluoroquinolone stewardship had twice as many new fluoroquinolone starts after discharge as hospitals without. Weber et al. analyzed discharge prescriptions in a 576-bed academic hospital in Portland. Among 6701 discharges, 22.9% were prescribed antibiotics upon discharge [35]. To complete these data, Scarpato et al. analyzed the appropriateness of antimicrobial agents prescribed on discharge [36]. They found that 70% of discharge antibiotics were inappropriate in antibiotic drug choice, dose, or duration. Analysis of discharge prescriptions should be the next step of our ASP with the implementation of an educational program to improve the prescription of discharge antibiotics. Moreover, there are biases inherent to the design of our study. Indeed, as for many "before–after" studies, the two groups we compared are heterogeneous. However, we found more pre-existing medical conditions for patients in the period after implementation of our ASP; therefore, this might lead to underestimation of the impact on the mean LOS downward trend we observed.

Finally, there was a center's effect limiting the extrapolation of our results as our study took place in a small hospital with less than 300 beds for the medicine, surgery, and obstetrics departments. The small hospital size likely facilitated the rapid establishment of this multidisciplinary system. One of the reasons of our success is probably the direct and confident relationship established between the IDS, the pharmacist, the microbiologist, and the prescribers, which may not be possible to install in other settings. Additional multicentric studies are needed to confirm our results and go further.

#### **4. Materials and Methods**

#### *4.1. Study Setting and Interventions*

Two hospitals (secondary care hospital of Bassin de Thau (HBT) and university hospital of Montpellier (UHM)) created a shared infectious disease specialist (IDS) position to sequentially implement an innovative multimodal ASP within HBT. The HBT is a 938-bed hospital with establishments providing care for dependent elderly people (376 beds); psychiatric unit (57 beds); geriatric and follow-up care and rehabilitation unit (167 beds); and acute care unit, medicine, surgery, and obstetrics unit (274 beds). On the whole, 406 beds are provided with an electronic medical record (EMR).

The infectious disease EMR pattern was duplicated from UHM to HBT to allow IDS response in real time with a complete traceability in the patient EMR for each intervention [37].

Several interventions centralized by the same IDS were progressively implemented:

January 2018: Simultaneous implementation of (i) a dedicated phone line for the infectious diseases specialist consulting service (IDSCS), (ii) weekly intensive care unit multidisciplinary clinical team meetings (ICU-M), and (iii) IDS intervention triggered by a bi-weekly microbiological laboratory meeting (LAB-M) for the revision of antibiotics based on microbiological data (blood cultures, per-operative samples, lumbar, pleural, and joint punctures). In addition, monthly educational training on antibiotic use was proposed to all residents of the hospital.

April to December 2018: Establishment of PAF interventions in association with pharmacy unit members. Initiation of critical antibiotics prescription review (PHARM-cATB) in April 2018. This consists of a systematized analysis of critical antibiotics prescription (injectable third-generation cephalosporins, fluoroquinolones (FQ), and carbapenems) performed twice a week within the eight wards with the greatest antibiotics use, with feedback to the prescriber.

Initiation of longer than 7 days antibiotics prescription review (PHARM-7d) in December 2018. A systematized analysis was performed on the whole hospital with feedback to the prescriber.

Each intervention was noted in the EMR in real time and was analyzed to evaluate intervention acceptance. An intervention was defined as having been followed if the proposed antibiotic type, duration, and dosage were accepted by the prescribing physician. The whole system organization is summarized in Figure 7.

#### *4.2. Outcomes*

The complementarity of the different actions was assessed by the quantitative and qualitative analysis of all interventions traced in the EMR (number of different types of interventions over time, analysis of proposal for each intervention, and impact of interventions on antibiotic use). Antibiotic consumption, defined in daily dose of antibiotics per 1000 patient-days (DDD1000PD), was calculated with ConsoRes® software [38] for the years 2016–2019 in the whole hospital. The impact on mean LOS, 30-DRR, and in-hospital mortality was assessed on patients with deep infections from the eight wards (medicine, surgery, and intensive care unit) representing 274 acute care beds that benefited from all the interventions of ASP by comparing two groups of patients over two periods of 18 months: January 2016–May 2017 versus January 2018–May 2019. All patients who were diagnosed with deep infections based on the International Classification of Diseases, Information System Medicalization Program were included in this comparison.

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 11 of 15

**Figure 7.** Antibiotic stewardship program organization in the studied hospital when fully established in December 2018. Dotted lines represent pre-existent relationship before implementation of the antibiotic stewardship program. On request IDSCS: on request infectious diseases specialist consulting service via a dedicated phone line. ICU-M: intensive care unit multidisciplinary clinical team meetings for weekly antibiotics prescription review in the ward. LAB-M: microbiological laboratory meetings for bi-weekly analysis of microbiological samples with microbiologists and infectious diseases specialist. PHARM-cATB: critical antibiotics prescription review twice a week with pharmacist and infectious diseases specialist. PHARM-7d: longer than 7 days antibiotic prescriptions review twice a week with pharmacist and infectious diseases specialist. EMR: electronic medical records. ICU: intensive care unit. *4.2. Outcomes*  **Figure 7.** Antibiotic stewardship program organization in the studied hospital when fully established in December 2018. Dotted lines represent pre-existent relationship before implementation of the antibiotic stewardship program. On request IDSCS: on request infectious diseases specialist consulting service via a dedicated phone line. ICU-M: intensive care unit multidisciplinary clinical team meetings for weekly antibiotics prescription review in the ward. LAB-M: microbiological laboratory meetings for bi-weekly analysis of microbiological samples with microbiologists and infectious diseases specialist. PHARM-cATB: critical antibiotics prescription review twice a week with pharmacist and infectious diseases specialist. PHARM-7d: longer than 7 days antibiotic prescriptions review twice a week with pharmacist and infectious diseases specialist. EMR: electronic medical records. ICU: intensive care unit.

#### The complementarity of the different actions was assessed by the quantitative and qualitative *4.3. User Experience Assessment*

analysis of all interventions traced in the EMR (number of different types of interventions over time, analysis of proposal for each intervention, and impact of interventions on antibiotic use). Antibiotic consumption, defined in daily dose of antibiotics per 1000 patient-days (DDD1000PD), was calculated with ConsoRes® software [38] for the years 2016–2019 in the whole hospital. The impact on mean An anonymous satisfaction survey was sent after 12 months of implementation of the system to all clinicians, followed by two reminder letters.

#### LOS, 30-DRR, and in-hospital mortality was assessed on patients with deep infections from the eight *4.4. Statistical Approach*

wards (medicine, surgery, and intensive care unit) representing 274 acute care beds that benefited from all the interventions of ASP by comparing two groups of patients over two periods of 18 months: January 2016–May 2017 versus January 2018–May 2019. All patients who were diagnosed with deep infections based on the International Classification of Diseases, Information System Medicalization Program were included in this comparison. Comparisons between the two periods were made using a χ 2 test for qualitative variables. Comparisons between the two periods were performed using a χ 2 test for mortality and readmission rate, a Mann–Whitney U test for the mean LOS, and a linear regression test to analyze antibiotic consumption.

#### *4.3. User Experience Assessment 4.5. Ethics*

An anonymous satisfaction survey was sent after 12 months of implementation of the system to all clinicians, followed by two reminder letters. This study was conducted according to the principles of the declaration of Helsinki and in compliance with International Conference on Harmonization/Good Clinical Practice regulations. According to the French law, the study was in accordance with the recommendations of the local ethics committee, without the need for consent.

#### **5. Conclusions**

This study is among the first to analyze the complementarity and impact of combining different strategies, especially interventionist methods, developed within ASPs. This system set up with reasonable human resources could easily be transposable to size-equivalent hospitals. A good acceptance rate of PAF interventions and clear complementarity of the different types of actions, leading to a major decrease in fluoroquinolones use and overall antibiotic use, without a negative impact on mortality or 30-DRR, are key points of this study.

Further studies are needed to strengthen the scope of our results, including multidisciplinary and educational programs; long-stay healthcare structures; analysis of discharge prescriptions; and giving a more important role to PAF interventions, which currently are likely not sufficiently developed [39,40]. As this type of system is probably cost-effective, the economic aspect should not be an obstacle to its implementation.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6382/9/12/848/s1, Table S1: Satisfaction survey regarding the antibiotic stewardship program in a tertiary hospital.

**Author Contributions:** Conceptualization, F.B., D.D., L.G., and D.M.; methodology, F.B., L.G., A.L., and D.M.; software, F.C., D.D., and A.L.; writing—original draft preparation, F.B.; writing—review and editing, V.L.M. and D.M.; supervision, V.L.M. and J.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

#### **Impact of a Rapid Diagnostic Meningitis**/**Encephalitis Panel on Antimicrobial Use and Clinical Outcomes in Children**

#### **Danielle McDonald <sup>1</sup> , Christina Gagliardo 2,3, Stephanie Chiu <sup>4</sup> and M. Cecilia Di Pentima 2,3,5,\***


Received: 9 October 2020; Accepted: 13 November 2020; Published: 18 November 2020

**Abstract:** Rapid molecular diagnostic assays are increasingly used to guide effective antimicrobial therapy. Data on their effectiveness to decrease antimicrobial use in children have been limited and varied. We aimed to assess the impact of the implementation of the FilmArray Meningitis Encephalitis Panel (MEP) on antimicrobial use and outcomes in children. In an observational retrospective study performed at Atlantic Health System (NJ), we sought to evaluate the duration of intravenous antibiotic treatment (days of therapy (DoT)) for patients <21 years of age hospitalized and evaluated for presumptive meningitis or encephalitis before and after the introduction of the MEP. A secondary analysis was performed to determine if recovery of a respiratory pathogen influenced DoT. The median duration of antibiotic therapy prior to the implementation of the MEP was 5 DoT (interquartile range (IQR): 3–6) versus 3 DoT (IQR: 1–5) (*p* < 0.001) when MEP was performed. The impact was greatest on intravenous third-generation cephalosporin and ampicillin use. We found a reduction in the number of inpatient days associated with the MEP. In the regression analysis, a positive respiratory pathogen panel (RPP) was not a significant predictor of DoT (*p* = 0.08). Furthermore, we found no significant difference between DoT among patients with negative and positive RPP (*p* = 0.12). Our study supports the implementation of rapid diagnostics to decrease the utilization of antibiotic therapy among pediatric patients admitted with concerns related to meningitis or encephalitis.

**Keywords:** meningitis; encephalitis; FilmArray; multiplex PCR; antimicrobial; rapid diagnostic technology; stewardship; children; adolescents; outcomes

#### **1. Introduction**

With the aid of rapid molecular diagnostics and the introduction of effective vaccines against *Haemophilus influenzae* type b, *Streptococcus pneumoniae*, and most recently *Neisseria meningitides*, the epidemiology of meningitis and encephalitis remains a rapidly evolving field [1]. The impact of vaccines has mainly affected children in developed countries, with an over 60% reduction in the incidence of bacterial meningitis in this patient population [2]. In a study performed in the United States in 2006, roughly 72,000 adult hospitalizations were related to meningitis [3]. While the majority of

these were due to viral etiologies (54.6%), the estimated healthcare cost reached USD 1.2 billion [3]. More recent data show that the global incidence of meningitis increased from 2.5 million cases in 1990 to 2.82 million cases in 2016, with the highest rates found in sub-Saharan African countries, also known as the meningitis belt [4]. Kwambana-Adams et al. published the prevalence of bacterial, viral and parasitic infection in children younger than 5 years of age in West Africa following the rollout of conjugate vaccines against pneumococcus (PVC), meningococcus (MenAfriVac) and *Haemophilus influenzae* [5]. *Escherichia coli* (4.8%), followed by *S. pneumoniae* (3.5%) and *Plasmodium* (3.5%), were the most prevalent etiologies of meningitis in this age group. Because serotyping for pneumococcal isolates was not reported, the impact of PVC could not be determined. Gram negative rods, particularly *Escherichia coli* and *Klebsiella pneumoniae*, were more commonly identified in newborns.

The initial clinical manifestations of central nervous system (CNS) infections in neonates and children can be non-specific, difficult to diagnose and devastating if not treated correctly. The implementation of diagnostic stewardship entails optimization of clinical care and antimicrobial therapy guided by timely and personalized effective testing [6,7]. Rapid diagnostics have been shown to improve clinical outcomes in patients with bacteremia and infections with multidrug-resistant organisms when the introduction of these tests are linked to effective antibiotic stewardship strategies [7,8]. Data on the performance and impact of the FilmArray Meningitis Encephalitis Panel (MEP) in children are limited [9]. The MEP is a rapid multiplex polymerase chain reaction (PCR) assay designed to detect 14 pathogens in the cerebrospinal fluid (CSF). These pathogens include six bacteria, seven viruses, and one yeast group. In the cases of meningitis or encephalitis, quick pathogen identification aids in the initiation/continuation of appropriate targeted therapy as well as discontinuation of unnecessary empiric antimicrobials. Timely diagnosis directly impacts patient outcomes and healthcare costs.

Prior to Food and Drug Administration (FDA) approval of the MEP, a large, prospective, multicenter study of 1560 CSF specimens was conducted to compare the MEP to standard diagnostics, bacterial culture and viral PCR [10]. In this study, the MEP yielded a percent positive agreement (PPA) of 100% for 9 of 14 analytes. Enterovirus yielded a 95.7% PPA, and human herpes virus type 6 had an agreement of 85.7%. *Streptococcus agalactiae* had one false-positive and one false-negative result. *Listeria monocytogenes* and *Neisseria meningitides* were not evaluated.

Additional studies augmented the results and strengthened the findings of this initial study. Liesman et al. evaluated 291 CSF specimens and found a PPA of 85.6% [11]. When results for *Cryptococcus neoformans*/*gattii* were excluded, the PPA increased to 92.5%. Naccache et al. evaluated 251 samples and showed a low false positivity rate [12]. Piccirilli et al. demonstrated 90.9% concordance between the FilmArray MEP and conventional microbiological procedures in 77 CSF samples studied [13]. Additionally, two published reviews had a pediatric focus. Graf et al. used 67 retrospective viral PCR or bacterial culture-positive samples and identified 92.5% that were positive for the same target on the panel [14]. Messacar et al. tested 138 CSF samples and concluded an overall agreement of 96% as compared to conventional diagnostic methods in children with CNS infections [15]. In a recently published meta-analysis by Tansarli and Chapin, and as previously reported by Liesman et al., the MEP was found to have higher rates of false-negative results for herpes simplex virus 1 and 2 and enterovirus when compared with standard PCR assays [11,16].

More recently, several studies demonstrated cost savings and reductions in antibiotic days of therapy (DoT) with implementation of the MEP [17–21]. Nabower et al. demonstrated decreased length of stay (LOS) and fewer acyclovir doses administered, while Weber et al. demonstrated hospital cost savings in a military treatment facility [17,18]. Posnakoglou et al. supplemented these findings, demonstrating decreased LOS, a reduction in antimicrobial use, and a decrease in total cost [19]. Similarly, Hagen et al. noted a decreased duration of empiric therapy, with the largest effect documented in infants [20]. Messacar et al. focused on herpes simplex virus in patients >60 days of age and observed a doubling of herpes simplex virus testing with a reduction in acyclovir duration of therapy [21]. These studies begin to validate the clinical utility of the MEP in pediatric patients with results that support opportunities for antimicrobial stewardship.

The purpose of this study was to evaluate the impact of the implementation of the FilmArray MEP in pediatric patients receiving empiric therapy for meningitis and/or encephalitis. The potential confounding role of respiratory pathogens was examined in a secondary analysis.

#### **2. Methods**

In an observational retrospective study performed at Atlantic Health System (AHS), we reviewed 297 medical records of patients <21 years of age evaluated for meningitis and/or encephalitis between January 2015 and September 2018. AHS is a not-for-profit private healthcare corporation operating five hospitals in northern New Jersey. Subjects evaluated at two AHS hospitals were included in the study: Goryeb Children's Hospital in Morristown and Goryeb Children's Center at Overlook Medical Center in Summit.

Admitted patients evaluated for meningitis by lumbar puncture prior to MEP incorporation were categorized and analyzed as "pre-implementation" subjects (January 2015–October 2016), whereas admissions on or after incorporation were categorized as "post-implementation" subjects (November 2017–September 2018). In order to only assess duration of empiric therapy, patients with confirmed bacterial infections and patients with herpes simplex meningoencephalitis were excluded from the study. Confirmation was based on positive MEP and CSF, blood cultures, and urine cultures. Hematology–oncology and neurosurgery patients were also excluded. A total of 247 patients were included in the final analysis.

The FilmArray MEP (BioFire Diagnostics, Salt Lake City, UT, USA) was incorporated at AHS on November 1, 2016. Analytes on the MEP include *Escherichia coli* (K1 capsular type), *Haemophilus influenzae*, *Listeria monocytogenes*, *Neisseria meningitides*, *Streptococcus agalactiae*, *Streptococcus pneumoniae*, *Cytomegalovirus*, *Enterovirus*, *Epstein-Barr virus*, *Herpes simplex viruses 1 and 2*, *Human Herpes virus 6*, *Varicella zoster*, *Human parechovirus*, and *Cryptococcus neoformans*/*gattii*. The sample size required is 200 microliters (µL), and the turnaround time for reporting the MEP at AHS is approximately 2 h. FilmArray Respiratory Pathogen panel (RPP; BioFire Diagnostics, Salt Lake City, UT, USA) was incorporated in 2011. Analytes of the RPP include Adenovirus, Coronaviruses (HKU1, NL63, 229E, OC43), Human metapneumovirus, Human rhinovirus/Enterovirus, Influenza A (A/H1, A/H#, A/H1-2009) and Influenza B viruses, Parainfluenza (1–4) viruses, Respiratory Syncytial Virus, *Bordetella pertussis*, *Bordetella parapertussis*, *Chlamydia pneumonia*, and *Mycoplasma pneumoniae*.

Data collected included patient age, gender, admission date, event date, CSF studies, diagnosis, antimicrobial therapy, RPP if performed, mortality and 30-day readmission.

The primary outcome of the study was to evaluate the duration of empiric antimicrobial therapy measured as DoT before and after incorporation of the MEP. Secondary outcomes included length of stay (LOS), all-cause mortality and 30-day readmission rates. Patient outcomes were compared preand post-implementation.

#### **3. Statistical Analysis**

Patient characteristics were summarized using medians and interquartile range (IQR) for continuous variables and proportions for categorical data. Total DoT and LOS failed normality tests, so non-parametric comparative analyses, Mann–Whitney, were performed to assess the data between the two groups. Binomial variables were compared using 2 proportions, and binary regression analyses were used to determine significant predictor variables. Categorical variables were evaluated using a chi-square test. All tests were 2-tailed at a level of significance of less than 0.05.

The AHS institutional review board approved this study (Protocol Number: 1107015-1).

#### **4. Results**

Two-hundred and forty-seven children with suspected meningitis or encephalitis who received empiric antimicrobial therapy were included in the study analysis. Of these, 186 patients were part of the pre-implementation period while 61 patients had an MEP performed during the post-implementation period. Patient characteristics for each group are depicted in Table 1. Age and gender were similar in both groups. The median age for all patients was less than 1 year of age. A total of 113 (60%) and 37 (64%) patients were males before and after the implementation of the MEP, respectively. Even when a higher proportion of patients was admitted to intensive care units during the pre-implementation phase, this difference was not statistically significant (*p* = 0.16). Patients were more likely to have a positive RPP prior to the implementation of the MEP (*p* < 0.01). gender were similar in both groups. The median age for all patients was less than 1 year of age. A total of 113 (60%) and 37 (64%) patients were males before and after the implementation of the MEP, respectively. Even when a higher proportion of patients was admitted to intensive care units during the pre-implementation phase, this difference was not statistically significant (*p =* 0.16). Patients were more likely to have a positive RPP prior to the implementation of the MEP (*p <* 0.01).

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 4 of 9



† Mann–Whitney, \* chi-square test, IQR: interquartile range, NICU: neonatal intensive care unit, PICU: pediatric intensive care unit, CSF: cerebrospinal fluid, WBC: white blood cells. † Mann–Whitney, \* chi-square test, IQR: interquartile range, NICU: neonatal intensive care unit, PICU: pediatric intensive care unit, CSF: cerebrospinal fluid, WBC: white blood cells.

The median duration of antibiotic therapy in the pre-implementation group was five DoT (IQR: 3–6) versus three DoT (IQR: 1–5) (*p* < 0.001) post–implementation. Figure 1 illustrates antibiotic utilization before and after the MEP was introduced into clinical practice in our study population. The median duration of antibiotic therapy in the pre-implementation group was five DoT (IQR: 3–6) versus three DoT (IQR: 1–5) (*p <* 0.001) post–implementation. Figure 1 illustrates antibiotic utilization before and after the MEP was introduced into clinical practice in our study population.

**Figure 1.** Patient-specific empiric antimicrobial utilization. **Figure 1.** Patient-specific empiric antimicrobial utilization.

During the pre-implementation period, the median DoT for individual antibiotics was 3 DoT (IQR: 1) for third-generation cephalosporins, including ceftriaxone and cefotaxime (*n*:23), 3 DoT (IQR: 3) for ampicillin (*n*:113) and 2 DoT (IQR: 2–3) for vancomycin (*n*:40). Ceftazidime was not used in either cohort. During the pre-implementation period, the median DoT for individual antibiotics was 3 DoT (IQR: 1) for third-generation cephalosporins, including ceftriaxone and cefotaxime (*n*:23), 3 DoT (IQR: 3) for ampicillin (*n*:113) and 2 DoT (IQR: 2–3) for vancomycin (*n*:40). Ceftazidime was not used in either cohort.

The median duration of empiric antibiotic therapy in patients with suspected meningitis or encephalitis during the post-implementation period was 2 DoT (IQR: 2–3) for third-generation cephalosporins (*n*:31) (*p =* 0.02), 2 DoT (IQR: 2–3) for ampicillin (*n*:22) (*p =* 0. 017) and 2 DoT (IQR: 2– 3) for vancomycin (*n*:7) (*p* = 0.2). Gentamicin was used in 65 and 17 subjects, before and after the implementation of the MEP with a median duration of 2 DoT for both patient groups (*p =* 0.13). We found no statistical differences in the median duration of cefepime (*p =* 0.70), doxycycline (*p =* 0.9) or piperacillin-tazobactam (*p =* 0.95) between the two cohorts. The median duration of empiric antibiotic therapy in patients with suspected meningitis or encephalitis during the post-implementation period was 2 DoT (IQR: 2–3) for third-generation cephalosporins (*n*:31) (*p* = 0.02), 2 DoT (IQR: 2–3) for ampicillin (*n*:22) (*p* = 0.017) and 2 DoT (IQR: 2–3) for vancomycin (*n*:7) (*p* = 0.2). Gentamicin was used in 65 and 17 subjects, before and after the implementation of the MEP with a median duration of 2 DoT for both patient groups (*p* = 0.13). We found no statistical differences in the median duration of cefepime (*p* = 0.70), doxycycline (*p* = 0.9) or piperacillin-tazobactam (*p* = 0.95) between the two cohorts.

Few patients received acyclovir before (*n*:32) and after implementation of the MEP (*n*:9). Median

In the regression analysis, in patients evaluated for meningitis or encephalitis, a positive RPP was not a significant predictor of duration of antibiotic therapy (Odds Ratio: 1.15; 95% Confident

utilization of acyclovir was 3 DoT (IQR: 3–4) and 2 DoT (IQR: 2–3), respectively (*p =* 0.76).

Few patients received acyclovir before (*n*:32) and after implementation of the MEP (*n*:9). Median utilization of acyclovir was 3 DoT (IQR: 3–4) and 2 DoT (IQR: 2–3), respectively (*p* = 0.76).

In the regression analysis, in patients evaluated for meningitis or encephalitis, a positive RPP was not a significant predictor of duration of antibiotic therapy (Odds Ratio: 1.15; 95% Confident Interval: 0.1–1.34). Furthermore, we found no significant differences between DoT among patients with negative (median: 4 DoT; range 0–6) and positive (median: 4 DoT; range 0–21) RPP (*p* = 0.12).

Secondary outcomes are summarized in Table 2. We found a statistically significant reduction in the median number of inpatient days after the implementation of the MEP (*p* < 0.01). All-cause readmission was higher in the pre-implementation group but did not reach statistical significance (*p* = 0.24). No deaths occurred in either cohort.


† Mann–Whitney, \* chi-square test.

#### **5. Discussion**

In our experience, implementation of the MEP decreased antimicrobial use and LOS among hospitalized children evaluated for presumptive meningitis or encephalitis, without having a negative impact on readmissions or mortality.

Performing a lumbar puncture in young infants and children can be challenging, limiting the ability to obtain large volumes of CSF to submit for multiple tests, especially when standard antigen, PCR and/or antibody testing must be performed at different reference laboratories. The MEP uses only 200 µL of CSF. Reference laboratories usually request a minimum of 500 µL of CSF to perform individual pathogen testing such as *Cryptococcus* antigen, HSV or enterovirus PCRs. CSF culture, although still the gold standard for diagnosis, takes a longer time to result. Standard microbiological methods for recovery and identification of an organism can take up to 48–72 h to report, and turnaround times for reference laboratories mean that it can take days to deliver results. At our institution, the MEP is reported within 2 h of obtaining the CSF sample. Furthermore, culture results can be difficult to interpret in patients who previously received antibiotic treatment. In a small study of 62 CSF samples from young infants with suspected meningitis, seven samples were positive on the PCR panel with no culture growth [22]. These seven samples were obtained from infants who had been pretreated with antibiotics. While false-positive and false-negative results from the MEP are possible, and results need to be interpreted in the context of the patient's clinical condition, the MEP may increase the ability to recover a clinically significant organism in children who have been pretreated with antimicrobials.

We noted that impact on antibiotic utilization mainly affected intravenous ampicillin, commonly used in newborns with suspected early or late onset sepsis and/or meningitis when *Streptococcus agalactiae*, *Listeria monocytogenes*, *E. coli*, and other Gram-negative pathogens are a consideration. With rates of ampicillin-resistant *E. coli* surpassing 50%, early identification of a potential etiology can guide appropriate antibiotic therapy. Similar impact was noted on third-generation cephalosporins, the antibiotics of choice for empiric therapy for infants and children with suspected CNS infections. Despite the intermittent shortages and eventual discontinuation of cefotaxime, and the age limitations for the use of ceftriaxone during the newborn period, we did not find a statistically significant change in the use of cefepime. Ceftazidime, an alternative to cefotaxime recommended by the American Academy of Pediatrics for infants under 2 months of age with suspected meningitis, was not used in our patient population. While the duration of empiric gentamicin use did not change with the rollout of the MEP, fewer patients were started on this antibiotic after the MEP was implemented.

The impact found on antibiotic utilization was independent of patients diagnosed with a respiratory viral pathogen. Furthermore, a positive RPP was not associated with a shorter duration of antibiotic therapy, implying that diagnosis of a viral respiratory infection did not drive antibiotic management in this patient population. Studies assessing the impact of rapid diagnostics in children with acute respiratory tract infections found that these assays reduce LOS and empiric antibiotic utilization [23]. To the best of our knowledge, prior studies assessing the impact of RPP in children admitted with possible CNS infection has not been published. Our data suggest that rapid syndromic molecular testing has a more meaningful impact when aimed at the diagnosis of concern rather than in combination.

Herpes simplex virus can be a devastating CNS infection in newborns. Although rare, with rates in the United States ranging between 1 in 2000 to 1 in 13,000 live births, early diagnosis and treatment remain critical to impact mortality and neurologic outcomes in this patient population [24]. In our institution, prior to the implementation of the MEP, CSF herpes simplex virus DNA testing was sent out to an outside laboratory, delaying turnaround time by several days. Though acyclovir use decreased after the implementation of the MEP, the small sample size made it difficult to assess the statistical impact on acyclovir DoT. However, rapid negative herpes simplex virus results from the MEP reduced the number of patients receiving empiric acyclovir therapy for several days pending results, as was seen in the pre-implementation period. These findings are critically important in the phase of recurrent shortages of intravenous acyclovir. Acyclovir shortages trigger the need to use alternative therapies such as IV ganciclovir or high-dose oral valacyclovir [25]. Of particular concern is the potentially negative economic and clinical impact associated with drug shortages described in the literature [26]. As of 22 September 2020, acyclovir remains in the drug shortage list kept by the American Society of Health-Systems Pharmacists [27]. The COVID-19 pandemic has generated an additional challenge to the chronic problem of antimicrobial shortages by creating an imbalance between supply and demand [28]. Furthermore, it is estimated that the global public health crisis generated by the severe acute respiratory syndrome COVID-19 resulted in an increased use of antimicrobials and could amplify the threat of antimicrobial resistance [29].

There are limitations to this study. We excluded hematology, oncology and neurosurgery patients. However, we purposely excluded these populations, in whom empiric antibiotic therapy might be guided based on risks associated with their underlying conditions. We also excluded hospitalized patients with documented infections to assess only the duration of empiric antimicrobial therapy. Our patient population represents a single institution, and the results might not be generalizable to all centers caring for children. Our study did not include an economic analysis. Duff et al. evaluated the financial outcome of the implementation of the BioFire® MEP in adult and pediatric populations at a single institution [30,31]. Greater savings were found when testing was performed in all suspected cases rather than in those with abnormal CSF findings for both pediatric (USD 3481/case) and adult (USD 2213/case) patients [30,31]. While antibiotic use is known to be associated with the emergence of antibiotic resistance, the direct impact of a single rapid diagnostic test is difficult to discern. To the best of our knowledge, studies evaluating the potential impact of the MEP on antibiotic resistance has not been published.

The World Health Organization is committed to decreasing the scourge of bacterial meningitis, especially for pathogens affecting young children, such as *Streptococcus agalactiae*, *Neisseria meningitidis*, *Streptococcus pneumoniae*, and *Haemophilus influenzae* [32]. The "Defeating meningitis by 2030" global roadmap is a multi-organization partnership that calls for the development and wide implementation of molecular-based multiplex meningitis rapid diagnostic assays at the point of care [32]. Newer generation meningitis assays will be needed to accomplish these goals worldwide [33]. Despite all the benefits described, rapid diagnostic tests should not replace routine bacterial and fungal cultures. Moreover, diagnostic testing should be interpreted in the context of the patient's clinical manifestations, and the possibility of either a false-positive or false-negative result should be considered based on the index of clinical and laboratory suspicion.

#### **6. Conclusions**

Meningitis remains a prevalent and devastating disease worldwide. While the timely administration of antimicrobial therapy is critical to optimizing the outcomes of children with CNS infections, unnecessary treatments and prolonged hospitalizations represent a burden on our healthcare system and contribute to antimicrobial shortages and resistance. In our experience, the implementation of rapid CSF multiplex PCR assays aided in antimicrobial stewardship initiatives and shortened the duration of hospital stays in children with suspected meningitis and encephalitis.

**Author Contributions:** Conceptualization, M.C.D.P.; formal analysis, S.C.; investigation, D.M., C.G. and M.C.D.P.; writing—original draft, D.M. and M.C.D.P.; writing—review and editing, C.G. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors have no conflicts of interest or financial relationships relevant to this article to disclose.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

#### **Implementation of a Delayed Prescribing Model to Reduce Antibiotic Prescribing for Suspected Upper Respiratory Tract Infections in a Hospital Outpatient Department, Ghana**

**Sam Ghebrehewet 1,\*, Wendi Shepherd 1,2 , Edwin Panford-Quainoo <sup>2</sup> , Saran Shantikumar <sup>3</sup> , Valerie Decraene <sup>4</sup> , Rajesh Rajendran <sup>5</sup> , Menaal Kaushal <sup>6</sup> , Afua Aku**ff**o 6 , Dinah Ayerh <sup>6</sup> and George Amofah <sup>6</sup>**


Received: 31 August 2020; Accepted: 31 October 2020; Published: 4 November 2020

**Abstract:** *Background*: High levels of antimicrobial resistance (AMR) in Ghana require the exploration of new approaches to optimise antimicrobial prescribing. This study aims to establish the feasibility of implementation of different delayed/back-up prescribing models on antimicrobial prescribing for upper respiratory tract infections (URTIs). *Methods*: This study was part of a quality improvement project at LEKMA Hospital, Ghana, (Dec 2019–Feb 2020). Patients meeting inclusion criteria were assigned to one of four groups (Group 0: No prescription given; Group 1; Patient received post-dated antibiotic prescription; Group 2: Offer of a rapid reassessment of patient by a nurse practitioner after 3 days; and Group 3: Post-dated prescription forwarded to hospital pharmacy). Patients were contacted 10 days afterwards to ascertain wellbeing and actions taken, and patients were asked rate the service on a Likert scale. Post-study informal discussions were conducted with hospital staff. *Results:* In total, 142 patients met inclusion criteria. Groups 0, 1, 2 and 3 had 61, 16, 44 and 21 patients, respectively. Common diagnosis was sore throat (73%). Only one patient took antibiotics after 3 days. Nearly all (141/142) patients were successfully contacted on day 10, and of these, 102 (72%) rated their experiences as good or very good. Informal discussions with staff revealed improved knowledge of AMR. *Conclusions:* Delayed/back-up prescribing can reduce antibiotic consumption amongst outpatient department patients with suspected URTIs. Delayed/back-up prescribing can be implemented safely in low and middle-income countries (LMICs).

**Keywords:** antimicrobial resistance (AMR); antimicrobial stewardship (AMS); delayed/back-up prescribing; upper respiratory tract infections; developing countries; LMICs; Ghana

#### **1. Introduction**

Resistance to antimicrobials poses a substantial threat to individual and public health. Antimicrobial resistance (AMR) is responsible for around 700,000 deaths globally per annum—this figure

is predicted to rise to 10 million by 2050 if current trends continue unabated [1] with a disproportionately heavier burden in developing countries [2]. Antimicrobial stewardship (AMS) as an organisational, healthcare system-wide approach to promoting and monitoring judicious use of antimicrobials to preserve their future effectiveness has a critical role in reversing these trends [3].

Upper respiratory tract infections (URTIs) antibiotic prescriptions account for the vast majority of antibiotic prescribing—usually in primary care [4]—where they are frequently prescribed for conditions where there is limited evidence of benefit, including acute otitis media and pharyngitis, and where there is no evidence of benefit, such as the common cold [5–7]. Delayed/back-up prescribing (where antibiotics can be accessed at a later time after the initial consultation) [8] is one strategy that can be implemented to reduce antibiotic prescribing.

Current British National Institute for Clinical Excellence (NICE) guidance suggests that a delayed antibiotic prescribing strategy "encourages self-management . . . but allows a person to access antimicrobials without another appointment if their condition gets worse" [9]. Delayed/back-up prescribing should not be used where there is evidence of serious illness or complications, or where the patient is in a clinical risk group [10].

A Cochrane systematic review identified 10 randomised clinical trials that investigated the effectiveness of delayed and no prescriptions strategies for respiratory tract infections [11]. The review found that there was no difference for adverse effects or results favoured delayed antibiotics over immediate antibiotic prescribing; significant reduction in antibiotic use compared to immediate prescription; patient satisfaction favoured delayed prescribing over no antibiotics and there was no difference in patient satisfaction.

Delayed/back-up prescribing can be implemented in a wide variety of ways. An English Ipsos-MORI survey of 1625 participants in 2015 showed that 15% of participants that were prescribed an antibiotics received a delayed prescription [12]. The same study showed a lack of awareness by the public of what the term "delayed prescription" means—after explanation, just 30% of respondents were opposed to General Practitioners (GPs) using this prescribing method for throat, urinary tract, ear, or chest infections. Furthermore, another study [13] reported that delayed prescribing is acceptable no matter how the delay is operationalised, but explanation of the rationale is needed and care taken to minimise mixed messages about the severity of illnesses and causation by viruses or bacteria. A Randomised Control Trial (RCT) considering delayed antibiotic prescribing for uncomplicated acute respiratory tract infections surmised that the practice of delayed prescribing was "associated with slightly greater but clinically similar symptom burden and duration and also with substantially reduced antibiotic use when compared with an immediate strategy" [14].

Delayed/back-up prescribing can be undertaken or approached in different ways. For example, practitioner-centred (when the health professional is responsible for completing the delayed prescription process) or patient-centred, where the patient has responsibility. An advantage of delayed/back-up prescribing is that it provides clinicians and patients with a safety net should an infection deteriorate or fail to improve. Other options may include a systems approach (if diagnosis is clearly identified) whereby local prescribing only allows for delayed prescription—however this may have significant limitations as it does not include the opportunity to review the patient's condition and consider appropriateness of delayed prescribing. Different approaches may be more appropriate for separate patient groups, practitioners, health facilities, or health systems. A typography of approaches is shown in Figure S1.

Ghana has high levels of AMR, with one study showing multidrug resistance rates of over 75% for some organisms [15]. This demonstrates an urgent need to introduce models of care that optimise antibiotic prescribing within a Ghanaian setting [16].

In 2019, a health partnership between the UK Faculty of Public Health (Africa Special Interest Group) and Ghana Public Health Association secured a global volunteering grant from the Fleming Fund's Commonwealth Partnerships for Antimicrobial Stewardship (CwPAMS), supported by Tropical Health and Education Trust (THET) and Commonwealth Pharmacists Association (CPA) to undertake a series of stewardship programmes at LEKMA hospital in Ghana. It was felt that there was an opportunity to try different models of antibiotic prescribing within a Ghana healthcare setting to understand what the barriers would be for implementation of a change to existing antibiotic prescribing practices in a low to middle-income country (LMIC) context. Theoretical work to understand challenges to tackling AMR in LMICs has demonstrated that a broad range of factors such as weak governance and poor regulatory measures, compounded by low public awareness of AMR and technological limitations to adequate surveillance, may present complexities not present in non-LMIC settings [2].

The main aim of this study was to explore the feasibility and practical application of different delayed/back-up prescribing models of antibiotics for the management of URTI within a large outpatient facility in a LMIC. This study also aimed to determine if delayed prescribing was safe within this setting, and to test the model's acceptability to both patients and clinicians.

#### **2. Results**

#### *2.1. Quantitative Results*

Over a 3-month period from December 2019 to February 2020, 142 patients who attended LEKMA hospital outpatient's department and were cared for by one of three medical doctors were eligible for delayed/back-up prescribing. Of these, 86 (61%) were female, 53 (37%) were male; and 3 (2%) did not specify gender (Table 1).



With regard to the different models of delayed/back-up prescribing, 61 (43%) patients were managed conservatively without a back-up prescription or reassessment option (Group 0), 16 (11%) had a post-dated prescription issued (Group 1), 44 (31%) were offered a follow-up appointment for reassessment with a nurse in 3-days if required (Group 2), and 21 (15%) had a prescription left for collection at the hospital pharmacy (Group 3) (Table 1).

As shown in Table 2, nearly half the participants (67, 47%) were working age adults, followed by children under 10 years of age (52, 36%).


**Table 2.** Age Profile of Participants.

The most common clinical diagnoses were sore throat (72%; n = 102), common cold (15%; n = 22) and acute sinusitis (5%; n = 10) with a similar distribution across the four groups (Table 3). Clinical diagnosis was not recorded for three participants. All participants were successfully contacted at day 10 to record outcome data. In all, only 12 (9%) patients remained mildly symptomatic at day 10, although they all indicated they were feeling better and none had sought further healthcare advice. A lower proportion of those in Group 0 had symptoms at day 10, compared with the other groups. Only one individual in the entire eligible patient population (from Group 3) took antibiotics based on worsening symptoms after 3 days. This patient was diagnosed with sore throat, which had subsided when contacted at day 10 and they had completed course of antibiotics.


A Likert Scale of 1–5 was used which has been found to have the most reliability and validity of available Likert Scale methodologies [17]. It should be noted that gaps in the dataset exist where patients were contacted for follow-up but either did not answer questions or the answers were not recorded. Most patients [102 (72%)] rated their experience as good or very good. No patients rated the care they received as poor. When considered by group, the group not given a back-up prescription or an appointment for reassessment (Group 0) were most satisfied with their experience (Table 3).

No adverse events or serious deterioration of illness were reported as a result of the delayed/back-up prescribing model during the 10-day follow-up period for all participants.

#### *2.2. Informal Discussions*

Following the period of data collection, informal/unstructured discussions were held with LEKMA Hospital staff (doctors, nurse practitioners and pharmacists) involved in the project. This approach, rather than structured interviews, was adopted to give the LEKMA healthcare workers the flexibility to discuss the key components of the pilot from their perspective and important considerations for their patients. These conversations were based on the rapport that had been built between the researchers and the staff at the hospital during the study. This approach, however, meant that each discussion was unique. From these discussions, we were able to ascertain both areas for development as well as successes to inform future delayed/back-up prescribing projects in LMICs.

The main areas for consideration all appear to stem from the lack of visible senior clinical leadership from project implementation. The senior team at LEKMA Hospital were all very supportive of the project from the outset, but the promotion of the project in the outpatients' department was delegated to the project/study staff. This resulted in misunderstanding by staff about the project aims and the role of clinical staff in delivery of the project. Inevitably, this comprehension impacted on the recruitment of participants to the project as clinical staff did not wish to engage. These issues were addressed once dedicated training had been delivered to the outpatients' department clinical staff, together with the active promotion of the project by senior leadership.

In-depth understanding of some of the root-causes for AMR and the potential solutions are not commonly/widely shared in Ghana—either by clinicians or by patients. The project provided an opportunity to raise awareness and educate both groups on the importance of the issue via face-to-face discussion with patients and regular AMR training sessions for staff in the outpatients' department. Furthermore, dedicated nurse practitioners input to the project provided opportunities for nurse practitioners to discuss other aspects of health and healthy living with the patients—this was particularly important as there is a low level of literacy in Ghana [16], and therefore traditional health messages via printed media may be missed.

#### **3. Discussion**

Our project setting was LEKMA, Greater Accra (Ghana), which has a population of 263,631 representing approximately 5.7% and 0.92% of the Greater Accra and Ghanaian populations, respectively. The population is young, 52% female, with a broad-based population pyramid which tapers towards the top, with very few individuals aged over 65. The catchment area population for LEKMA district general hospital is like that of most hospitals in LMICs, especially in Africa. Therefore, considering some of the cultural and socioeconomic factors, we believe the findings of this study are likely to be applicable to most LMICs, especially African countries.

There was an initial cost to this project (nurse practitioners' time) which, if the model was to be implemented on a long-term basis, would need investment. Depending on the funding arrangements within the setting of application, this could be viewed as a long-term investment initiative through the reduction of antibiotic prescribing and associated reduction in AMR which will offset the on-going staff support costs. The provision of nursing support did not affect the patient experience or clinical outcome—rather this reassured medical colleagues about the safety of the model—so delayed/back-up prescribing within an LMIC setting for URTIs could be a cost-neutral endeavour if medical support for the model is in place from the outset. More detailed work on the financial aspects of delayed prescribing in LMICs is required to determine the specific cost-benefit of the strategy, but this is likely to be dependent on the specific funding structures for healthcare in local settings.

The primary message from this project is the practical demonstration and evidence that delayed prescribing models for URTIs can be safely utilised in primary care in the Ghanaian healthcare setting—albeit with some areas which require further redress for sustainability and professional acceptance.

#### *3.1. Safety and Patient Outcomes*

This study demonstrates that all the models of delayed prescribing, including the group given no further follow-up options (Group 0), were acceptable to both staff and patients. With the exception of one patient, all other patients involved in the project reported no deterioration of symptoms as a result of participation. Indeed, there were no adverse events related to the patients' presenting symptoms during the project and follow-up period (10 days post presentation).

Most patients (91%) who participated in the study had indicated presenting symptoms had resolved by day 10. This reinforces the findings and conclusions from several studies which have demonstrated no differences in antibiotic prescription rates or clinical outcomes between immediate and delayed prescription and, in the short term, there is also little difference in symptom control between delayed prescription, no prescription, or immediate prescription. Delayed/back-up antibiotic prescription resulted in the minority of patients using antibiotics, and any strategy of delayed prescribing is likely to result in fewer than 40% of patients using antibiotics [17]. In this project, a significantly lower proportion of those in Group 0 (2%) had symptoms, albeit milder, at day 10 compared with the other groups. There is a possibility that these patients were generally less unwell at time of presentation, so clinicians felt more comfortable in not providing any treatment or delayed/back-up prescribing options.

It has been found that patients who had been prescribed an antibiotic for cough in the previous two years were over twice as likely to consult for a similar illness, and that a delayed antibiotic prescription strategy reduced re-consultation by 78% in this group [11]. Another study into paediatric antibiotic use also found that "delayed (rather than immediate) antibiotics reduced re-consultations for deterioration for children with URTI in RCT" [18]—a view supported by other studies into antimicrobial prescribing strategies [18,19]. Although the follow-up period in this project was short (10 days), none of the participants sought further heath advice or took antibiotics from any other source. There was also no need for re-consultation in the follow-up period. A longer period of follow-up, however, is required to confirm that there was no further deterioration or need for clinical re-assessment after the 10-day period, as well as the long-term impact on repeat antibiotic usage in the project population.

The research team had expected that there may be some logistical barriers to uptake of different models within the project, such as travel time and expense to return if symptoms did not spontaneously resolve or if clinical condition deteriorated, which may result in poorer clinical outcomes. It is acknowledged that we did not have a complete response to these questions from all participants (39/82 in Groups 1–3), but 37/38 (97.4%) of those who did respond stated that these factors were not an issue meaning that these factors were not a primary concern for patients in the LEKMA project. Although care should be taken in interpreting these results as they are based on a small number of responses, it is not unreasonable to suggest these results are likely to generalisable to the wider catchment area of LEKMA Hospital and beyond in Ghana, i.e., given the similar geographic or population economic circumstances.

#### *3.2. Reduction in Antibiotic Prescribing*

*3.3. Behaviour Change and Acceptability* 

model.

Over the course of the project, there was a reduction of at least 141 antibiotic prescriptions (one prescription per individual but some individuals may have received multiple prescriptions to treat the same infection). It is reasonable to extrapolate the potential number of antibiotic prescriptions that could be saved over a year if this service improvement project is extended to all clinical staff. Based on surveillance data from LEKMA Hospital, and assuming that all patients presenting with URTI will be prescribed antibiotics [personal communication with LEKMA hospital doctors] we estimated that at least, well over 2000 antibiotic prescriptions can be avoided for just two of the commonest URTIs in LEKMA hospital (common cold and sore throat) in the peak URTI season (October to December), i.e., if all LEKMA hospital outpatient clinicians were involved in the implementation of delayed/back-up prescribing. This figure could be at least 3× higher (>6000 antibiotic prescriptions) if delayed/back-up prescribing is implemented for a full 12-month period for all URTI and by all out-patient department clinicians (Figure 1). *Antibiotics* **2020**, *9*, x FOR PEER REVIEW 7 of 12

**Figure 1.** Attendance of Patients with upper respiratory tract infections (URTIs) at LEKMA Hospital, July 2019–June 2020. in addition to senior clinical leadership, cannot be overemphasised. **Figure 1.** Attendance of Patients with upper respiratory tract infections (URTIs) at LEKMA Hospital, July 2019–June 2020.

antibiotic resistance, while maintaining patient safety and satisfaction levels" [1].

The delayed prescribing model was rated as good or very good by 95% of patients. After initial

Contrary to the perceptions of some local clinicians, Group 0 were the most satisfied with their experience. LEKMA clinicians felt that a potential barrier to the project was that individuals would not wish to leave the outpatient department without a prescription of some nature. This mirrors research considering the issue of increasing antimicrobial resistance, where the authors concluded that "where clinicians feel it is safe not to prescribe antibiotics immediately for people with respiratory infections, no antibiotics with advice to return if symptoms do not resolve is likely to result in the least antibiotic use while maintaining similar patient satisfaction and clinical outcomes to delaying prescription of antibiotics. Where clinicians are not confident in using a no antibiotic strategy, a delayed antibiotics strategy may be an acceptable compromise in place of immediate prescribing to significantly reduce unnecessary antibiotic use for URTIs, and thereby reduce

Our findings demonstrate that, not only was the use of no prescribing (delayed/back-up) acceptable to patients, it was also safe. In addition, the delayed/back-up strategy meant that medical staff were reassured that patients had a point of contact in the nursing team if their condition did not improve/deteriorated, so that they could access further advice and treatment—even if they had been assigned to the Group 0. We hope that these findings can contribute to changing the behaviour of healthcare professionals when considering whether or not to prescribe antibiotics for URTIs.

The experiences of staff during the project, while positive overall, indicate the complexities of the healthcare system in Ghana, the need for complete transparency as to the rationale for such a programme, and the importance of visible senior leadership from the outset. The need for dedicated training and awareness raising of the scheme and the rationale prior to commencement of the project,

#### *3.3. Behaviour Change and Acceptability*

The delayed prescribing model was rated as good or very good by 95% of patients. After initial trepidation about the model by clinicians, staff awareness sessions and visible senior clinical leadership were successful methods of ensuring clinician buy-in to the delayed/back-up prescribing model.

Contrary to the perceptions of some local clinicians, Group 0 were the most satisfied with their experience. LEKMA clinicians felt that a potential barrier to the project was that individuals would not wish to leave the outpatient department without a prescription of some nature. This mirrors research considering the issue of increasing antimicrobial resistance, where the authors concluded that "where clinicians feel it is safe not to prescribe antibiotics immediately for people with respiratory infections, no antibiotics with advice to return if symptoms do not resolve is likely to result in the least antibiotic use while maintaining similar patient satisfaction and clinical outcomes to delaying prescription of antibiotics. Where clinicians are not confident in using a no antibiotic strategy, a delayed antibiotics strategy may be an acceptable compromise in place of immediate prescribing to significantly reduce unnecessary antibiotic use for URTIs, and thereby reduce antibiotic resistance, while maintaining patient safety and satisfaction levels" [1].

Our findings demonstrate that, not only was the use of no prescribing (delayed/back-up) acceptable to patients, it was also safe. In addition, the delayed/back-up strategy meant that medical staff were reassured that patients had a point of contact in the nursing team if their condition did not improve/deteriorated, so that they could access further advice and treatment—even if they had been assigned to the Group 0. We hope that these findings can contribute to changing the behaviour of healthcare professionals when considering whether or not to prescribe antibiotics for URTIs.

The experiences of staff during the project, while positive overall, indicate the complexities of the healthcare system in Ghana, the need for complete transparency as to the rationale for such a programme, and the importance of visible senior leadership from the outset. The need for dedicated training and awareness raising of the scheme and the rationale prior to commencement of the project, in addition to senior clinical leadership, cannot be overemphasised.

#### *3.4. Strengths of Project*

This project contributes to the evidence base around the use of delayed prescribing as a strategy to reduce antibiotic usage within outpatient settings in LMICs. We have found no previous studies that have examined or implemented the strategy of delayed/back-up prescribing in LMICs in our extensive literature searches on Medline, CINAHL, or Global Health databases until July 2020.

The real-life setting of this project provides evidence for the applicability of delayed/back-up prescribing models in similar settings. Through implementation of the project in a functioning and very busy outpatient department, we were able to confirm theoretical principles into practice. Furthermore, the use of different grouping to test all the current suggested delayed prescribing models, including an information only option (Group 0), demonstrates that there was no major difference in outcomes based on the model used.

The use of Group 0 also removes doubt regarding the possibility that delayed/back-up prescriptions may have been used on those who did not require them. Coupled with proactive efforts to reduce private antibiotic sales from local community pharmacies, we can be confident that the results outlined above are an accurate reflection of the treatment each participant received for their condition.

Use of a 10-day follow-up period within the project increased patient safety as it ensured that all patients had a point of rapid access in case of lack of improvement in their condition. Provision of a call to the patient at the end of the follow-up period allowed documentation of clinical outcome.

Finally, the project had an excellent response rate from participants—both on enrolment and at the end of the follow-up period—which provides assurance as to the accuracy of the findings.

#### *3.5. Limitations of Project*

This service improvement project had some limitations which require discussion to build on our experiences for future delayed/back-up prescribing projects in similar settings.

This initiative had a relatively small number of eligible participants. In part, this may be due to buy-in and understanding of the project from clinicians within LEKMA outpatients' department in the early period of the project. Although training was offered, it did not reach all outpatients' department doctors, and this may have contributed to the lack of engagement from all the doctors who worked at LEKMA outpatients' department at the time. We did not interview clinicians who were not involved in the project to see what factors may have been responsible for the low take-up of the model by prescribing staff.

Although we have one year's data, we do not know the true trend of URTIs at LEKMA Hospital. Further analysis on general rates of URTIs over the project period may also yield information about whether there was simply a lower burden of URTIs than expected compared with previous years. It was not possible to run the project over all outpatient clinics throughout the week due to project resources only being available on weekdays, so we do not have information on weekend and out of hours attendance for URTIs. We did not collect information on patient's medical history or existing comorbidities, and these may have impacted on patient outcomes, i.e., persistence of illness or need for antibiotics. In addition, outcomes were self-reported and not validated by clinical examination.

More extensive qualitative and quantitative research throughout the project would have been beneficial to understand other reasons for low take-up—for example, we do not have information on number of potential participants approached who did not consent to being in the project, the number of patients with an URTI who came to the clinic who were not included in the study who may have been given antibiotic treatment. We also do not have information on the exact impact of COVID-19 on potential participant attendance at the outpatients' department, especially in February 2020. These low numbers of participants, and the single site setting, may limit the generalisability of these findings and inference to other settings in Ghana and other LMICs. This is compounded by this not being an RCT design, and therefore a comparative, standardised study design was not undertaken.

Data quality and data completeness have affected the comprehensiveness of the findings, as there are significant gaps in participant responses—particularly towards the end of the project period when staff were undertaking additional duties to assist with the COVID-19 efforts. Counter to this, at the start of the project, participants were not sequentially allocated to a group, which may have led to some bias. Staff involved in the project have commented that they were not aware of the importance of sequential allocation initially. It was also not possible to blind the study due to the service nature of the project, which may have affected our outcomes. Additional training of project staff in service improvement methodology prior to the commencement of the project may have increased local ownership and understanding of the rationale behind the data collection tools which may have improved data quality. This would have the added benefit of allowing local adaptation of tools to fit local circumstance based on local knowledge.

We recommend that further studies are conducted to address some of these issues in other settings.

#### **4. Materials and Methods**

The service improvement pilot ran from December 2019 to February 2020 within the outpatients' department of LEKMA Hospital, Accra, Ghana. Inclusion criteria were all patients who presented at the setting with URTI symptoms during the project period and were deemed to be eligible for the study by the examining clinician. Exclusion criteria were any patient who is diagnosed with URTI but clinician considered that delayed antibiotic prescription was inappropriate or any patient that did not verbally consent to take part. As this was not clinical research, the quality improvement project was discussed with LEKMA Hospital management, and all available evidence regarding back-up/delayed prescribing was presented. The consensus by LEKMA management was that this is evidence-based

good practice that would be a key component of quality improvement for the hospital and would not alter patient choices and opportunities; therefore, no ethics approval was required.

There were four different models of delayed/back-up prescribing—no prescription given (only information leaflet); post-dated prescriptions being given to patients at time of first clinical appointment to use if no symptom resolution 3 days after clinic visit; rapid access to a nurse-led clinic for re-assessment after three days, i.e., if symptoms did not improve/patient's condition deteriorate; and prescription forwarded to the hospital pharmacy and clinician/nurse practitioner asking patient to visit the hospital pharmacy to collect a pre-written antibiotic prescription if symptoms did not improve within 3 days. Delayed/back-up antibiotic prescribing was only to take place when it was deemed clinically appropriate to do so—this decision was at the sole discretion of the clinician responsible for the patients' care. The clinical pathways are available in Figure S2.

Two experienced nurses were recruited from the hospital staff to provide support to the project on a full-time basis for the project duration, although their time was diverted to COVID-19 response towards the end of the project. A data collection tool was developed using Microsoft Excel to ensure contemporaneous data capture by these nurses and provide a database of clinical presentation, clinical outcome at the end of follow-up period, and general evaluation information. Patient information leaflets were prepared for different groups. All project materials are available in the Supplementary Information.

Participants were allocated to one of the four groups on a sequential basis by the nursing team. The study was designed so that there was sequential randomisation of participants but, in reality, this was initially more ad-hoc than was planned which resulted in unequal sized groups between the different interventions and may have introduced bias into the study.

Dedicated face-to-face training to medical, nursing, and administrative staff in the outpatients' department and pharmacy was provided by the UK partnership staff in advance of the project to advise them of the project aims and objectives. Further training was provided in January 2020 by senior medical staff at LEKMA Hospital.

Community pharmacists in the area surrounding LEKMA Hospital were visited by the project team to try to minimise non-prescribed antibiotic purchases direct from pharmacists.

All patients were contacted via telephone 10-days after their initial presentation to ascertain if they were still symptomatic; if they had consulted other medical professionals over the 10-day period; antibiotics taken in previous 10-days (and source); and their experience of the care they received using a Likert scale.

Informal discussions with LEKMA staff members actively involved in the project were undertaken, and their transcripts were analysed to understand how the field experience at the local level in Ghana related to the potential barriers outlined in other work on antimicrobial prescribing in LMICs [2].

#### **5. Conclusions**

Despite good evidence for a delayed prescribing approach in other regions (mainly high-income countries), there is sparse published evidence for the use of delayed/back-up prescribing in LMICs—particularly in Africa, where alternative methods of gaining antibiotics may exist (such as direct illicit purchase from community pharmacists, as can occur in Ghana). Furthermore, in LMIC hospital outpatient departments, which see new clinical presentations of illness (much like in primary care), the proportion of antibiotic prescriptions that are written for URTIs is unknown, and inappropriate antibiotic prescribing is likely to be high.

The results from this service improvement project show support from both clinicians and patients for more dedicated interventions to reduce inappropriate prescribing of antibiotics in LMICs with little preference for which model of delay/back-up prescribing used. The success of the models is reflected through a significant reduction in antibiotic use for URTIs in LEKMA outpatients during the project with no serious illness or adverse events recorded over the 10-day follow-up period. Furthermore, upscaling implementation delayed/back-up prescribing in LMICs could contribute to improvement in clinicians' confidence, optimise antibiotic prescribing and reduce antimicrobial resistance.

Further in-depth exploration of clinicians' and patients' experiences and perceptions need to be captured to help optimise delayed/back-up prescribing implementation. Extended project schemes along the same model should be used in different settings and with larger cohorts of patients to prove the clinical applicability of the model to other LMIC settings using a bigger data set.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6382/9/11/773/s1, Figure S1: Typography of approaches to delayed antibiotic prescription, Figure S2: Patient pathways for project of delayed prescribing in LEKMA Hospital outpatients' department, Ghana.

**Author Contributions:** S.G., W.S., E.P.-Q., and S.S. drafted the paper and contributed to the project idea, conceptualisation, design and implementation of the study. V.D., G.A., M.K., A.A., D.A. and R.R. contributed to the design and implementation of the study and commented on the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project was funded as part of the Commonwealth Partnerships for Antimicrobial Stewardship Scheme (CwPAMS) supported by Tropical Health and Education Trust (THET) and Commonwealth Pharmacists Association (CPA) using Official Development Assistance (ODA) funding, through the Department of Health and Social Care's Fleming Fund. The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, PHE, the Fleming Fund, the Department of Health and Social Care, THET or CPA.

**Acknowledgments:** On behalf of the UK Faculty of Public Health (Africa Special Interest Group), and Ghana Public Health Association, we would like to thank all the UK FPH and GPHA partnership members for their invaluable input and support in the implementation of this projects. The partnership members are: Col. Edwin Afari (Rtd.), President of the GPHA; Perdita Hilary Lopes, Financial Secretary & Deputy Project Coordinator; Michael Adjabeng, GPHA Secretary; Amoah James McKeown, Public Relations Officer, GPHA; Samantha Walker, Lead Nurse—Infection Prevention and Control, Countess of Chester Hospital NHS Foundation Trust; Indu Das, Antimicrobial Specialist Pharmacist, East Cheshire NHS Trust. We are also grateful for LEKMA Hospital Senior Management, especially Juliana Ameh, Medical Superintendent (during the project period), for hercommitment to support the project with 2 full-time nurse practitioners. In addition, we would like to thank the following members of LEKMA Hospital staff for providing information, and support in coordination and implementation of the project: Akua Gyimah-Asante, Medical Superintendent; Stella Siriboe, Director of Nursing; Anthonia Aba Bannerman-Quist, Chief Pharmacist; Paul Beson, Infection Prevention & Control (IPC), Focal Lead; Abdul Razak Amari, Head of Laboratory; and Doreen Atta-Fynn. Finally, we would like to thank Merav Kliner and Peter MacPherson for their initial work in 2015 on the evidence base for delayed prescribing models within the UK.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

#### **Is Antimicrobial Dosing Adjustment Associated with Better Outcomes in Patients with Severe Obesity and Bloodstream Infections? An Exploratory Study**

**Stéphanie Sirard <sup>1</sup> , Claire Nour Abou Chakra <sup>1</sup> , Marie-France Langlois <sup>2</sup> , Julie Perron <sup>3</sup> , Alex Carignan <sup>1</sup> and Louis Valiquette 1,\***


Received: 8 September 2020; Accepted: 14 October 2020; Published: 16 October 2020

**Abstract:** The impact of adjusted treatment on clinical outcomes in patients with severe obesity is unclear. This study included adults with severe obesity admitted for bloodstream infections between 2005 and 2015. The patients were grouped according to the percentage of the appropriateness of the dosage of their antimicrobial treatment: 80–100% = good, 20–79% = moderate, and 0–19% = poor. The association between antimicrobial adjustment and a composite of unfavourable outcomes [intensive care unit stay ≥72 h, duration of sepsis >3 days, length of stay ≥7 days or all-cause 30-day mortality] was assessed using logistic regression. Of 110 included episodes, the adjustment was rated good in 47 (43%) episodes, moderate in 31 (28%), and poor in 32 (29%). Older age, Pitt bacteremia score ≥2, sepsis on day 1, and infection site were independent risk factors for unfavourable outcomes. The level of appropriateness was not associated with unfavourable outcomes. The number of antimicrobials, consultation with an infectious disease specialist, blood urea nitrogen 7–10.9 mmol/L, and hemodialysis were significantly associated with adjusted antimicrobial dosing. While the severity of the infection had a substantial impact on the measured outcomes, we did not find an association between dosing optimization and better outcomes.

**Keywords:** obesity; bloodstream infection; antimicrobials; prescription

#### **1. Introduction**

In the last 40 years, the prevalence of obesity has doubled in more than 70 countries, accounting for over 2 million deaths worldwide [1]. Aside from comorbidities such as type 2 diabetes, hypertension, and cardiovascular diseases, obesity is associated with a high risk of infections [2,3]. Physiologic alterations in patients with obesity influence the pharmacokinetics (PK) and pharmacodynamics (PD) of many drugs, including antimicrobials [4,5]. Underdosing of antimicrobials in patients with obesity could lead to sub-inhibitory concentrations, which, in turn, could impair treatment and lead to clinical failure [6–9]. For example, in patients with severe obesity and cellulitis, a low antimicrobial dose upon hospital discharge was associated with either recurrence, emergency room visit, rehospitalization, or 30-day attributable death (odds ratio [OR] 3.6 95% CI 1.4–9.4) [7]. In a cohort of critically ill patients with complicated intra-abdominal and skin and soft tissue infections, high doses of tigecycline resulted

in a significant reduction in mortality, intensive care unit (ICU) length of stay (LOS), and occurrence of bacteremia and septic shock [10].

Bloodstream infections (BSI) are severe infections and one of the leading causes of death in North America and Europe [11]. In one study, the risk of BSI was higher in patients with obesity than in normal-weight patients (31% for body mass index (BMI) of 30–34.9, 87% for BMI of 35–39.9, and 210% for BMI of ≥40) [12]. While obesity had no association with short-term all-cause mortality and clinical outcomes in patients with BSIs [13], another study found that high BMI was associated with organ failure and all-cause hospital mortality in patients with BSIs due to Gram-negative bacteria [14].

Although some studies have investigated the link between obesity and unfavourable outcomes associated with BSI, to our knowledge, none has focused on the impact of antimicrobial dose adjustment in BSI patients with class III obesity [12–16].

In this article, we describe a retrospective cohort of adults with class III obesity hospitalized for BSI, where we assessed factors associated with adjustment of antimicrobial dosing and compared clinical outcomes according to the appropriateness of antimicrobial dose adjustments.

#### **2. Materials and Methods**

#### *2.1. Population and Study Design*

This retrospective study was conducted at the Centre intégré universitaire de santé et de services sociaux de l'Estrie-Centre hospitalier universitaire de Sherbrooke (CIUSSSE-CHUS), a 677-bed academic centre in the Province of Quebec, Canada. Approval was obtained from CIUSSSE-CHUS institutional review board (#12–187). Subjects were identified through a clinical data warehouse. All adult patients with documented class III obesity (BMI ≥ 40 kg/m<sup>2</sup> ) hospitalized between 1 August 2005 and 31 August 2015 for BSI were included.

BSI was defined by the presence of a pathogen in one blood culture or the presence of a skin flora microorganism (coagulase-negative staphylococci, alpha-hemolytic streptococci, *Micrococcus* species, *Propionibacterium*/*Cutibacterium* species, *Corynebacterium* species, and *Bacillus* species) in at least two consecutive blood cultures (from two different sites).

Specific populations for whom the BMI was not reliable, such as pregnant women, patients with dwarfism, those with above-the-knee bilateral amputation, or those with a history of bariatric surgery were excluded. Other exclusion criteria were: presence of fungemia, transfer from another hospital after ≥48 h, receiving palliative care, presence of more than one bacterial infection, or two or more distinct episodes during hospitalization. We excluded patients who were treated for the whole or the majority of the treatment (>80%) with an antimicrobial requiring no adjustment for obesity (cefixime, moxifloxacin, ertapenem, fosfomycin, and tigecycline) and those treated with vancomycin or aminoglycosides only, especially cases where it was the only effective antimicrobial. In addition, patients who died or did not received an antimicrobial within the first 48 h after the initial positive blood culture, or had inadequate antimicrobial coverage for >48 h were excluded.

#### *2.2. Data Collection*

A standardized form was used to collect data on clinical variables from computerized medical charts. Pathogens isolated from blood samples were noted along with antimicrobial susceptibility test results. Immunosuppression was defined as the presence of leukaemia, lymphoma, HIV infection, neutropenia (neutrophils < 1800/µL), organ transplantation, and connective tissue disease or use of immunosuppressive drugs for over one month within the previous six months. To evaluate the severity of illness, the Pitt bacteremia score (PBS) (ranging from 0 to 18) was documented on the day of the positive blood culture and up to 48 h prior [17,18]. The time to effective antimicrobial was determined by the time between admission and the administration of the first effective antimicrobial related to the infection. All antimicrobial prescriptions relevant to the treatment of BSI were reviewed for the route

of administration, dose, and dosing interval and were compared to the local guidelines for adults with class III obesity based on the current literature (see Table 1) [5,19,20].


Abbreviations: CrCl: creatinine clearance, IV: intravenous, PO: oral administration. \* estimated with the Cockcroft-Gault equation with adjusted body weight.

Cockcroft-Gault equation with adjusted body weight was used to estimate creatinine clearance (CrCl) at the beginning of all prescriptions and for every significant change in creatinine values [21]. A prescription was deemed inadequate if either the dose and/or the dosing intervals were lower than expected for class III obesity. When multiple antimicrobials were administered at the same time, we considered the whole prescription adequate if at least one antimicrobial dosing and the spectrum were adequate. The first prescription was also carefully reviewed. The percentage of the appropriateness of the dose and dosing intervals was calculated by dividing the number of days of adequate treatment by the total number of days of treatment and was considered good (80–100%), moderate (20–79%), or poor (0–19%). This classification has been selected after discussion with infectious disease experts, locally.

#### *2.3. Outcomes*

To reflect unfavourable outcomes potentially associated with unadjusted antimicrobial dosing and because of the low frequency of each component, we constructed a composite primary endpoint including clinically relevant components: ICU stay ≥72 h, duration of sepsis >3 days, LOS ≥7 days, or all-cause 30-day mortality. Other secondary endpoints collected per hospitalization were time to defervescence, time to white blood cells (WBC) normalization, time to sepsis normalization, ICU LOS, need for and duration of mechanical ventilation, and hospital LOS. Fever was defined as an increase in body temperature above 37.5 ◦C orally, 38 ◦C rectally and centrally, or 37.3 ◦C axillary. Time to defervescence was the time between the first abnormal value and the first normal value that remained within normal values for at least 48 h. Only the first febrile episode within 48 h of the first positive blood culture was considered in this calculation. WBC normalization associated with the first positive blood culture was defined as a stable return within the normal range during hospitalization. We could not calculate this variable in patients for whom WBC counts were within or below normal range during the entire study period. To define sepsis, we used a modified sequential organ failure assessment (mSOFA) to overcome the limitations due to missing values [22,23]. A serial mSOFA was calculated on days 1, 3, 5, and 7 with the most abnormal values in the 24-h period. Missing values were imputed with normal values, and the corresponding parameter of mSOFA was attributed a score of 0. Sepsis was defined as an mSOFA score of ≥2. All-cause readmission within 30 days of hospital discharge

was assessed, and relapse was considered when patients were hospitalized for the initial infection or a complication.

#### *2.4. Statistical Analysis*

Data were analyzed using IBM SPSS Statistics for Windows, version 25 (IBM Corp., Armonk, NY, USA). Groups of patients were compared on the basis of the appropriateness of the antimicrobial dosage. To account for potential changes linked to the impact of an antimicrobial stewardship program based on a decision support system (APSS, Lumed Inc., Canada) implemented in August 2010, we divided the study period into three segments: pre-APSS (2005–July 2010), early-APSS (August 2010–2012), and late-APSS (2013–2015). The Antimicrobial Prescription Surveillance System (APSS) is an asynchronous system that generates alerts for potentially inappropriate antimicrobial prescriptions based on published recommendations and expert opinions. These alerts are reviewed by pharmacists who are part of the antimicrobial stewardship program team and recommendations are made to physicians. Special rules were developed for patients with class III obesity [24].

Descriptive statistics were used to characterize baseline demographic characteristics, comorbidities, and outcomes, stratified by the level of appropriateness. Descriptive analyses are presented using three groups based on the level of appropriateness (good, moderate and poor). For some comparisons, we combined the moderate and poor groups and compared to the most optimal group (good). Categorical variables were reported as number and percentage for each group and were compared using the χ 2 test or binary logistic regression, when appropriate. Continuous variables were reported as median values with their interquartile range (IQR) and were compared with the Wilcoxon test. Logistic regression was used to assess the association between adjusted antimicrobial dosing and unfavourable outcomes and to identify factors associated with adjusted antimicrobial therapy (0–19% poor compared to 20–100%). Selected variables and variables identified in univariable analysis were included in a multivariable model in order of the lowest *P*-value and results of the likelihood ratio test. The results are presented as unadjusted or adjusted OR (aOR) with 95% confidence interval (CI).

#### **3. Results**

During the study period, 160 clinical episodes of positive blood cultures in adults with class III obesity were identified in our centre, and 110 episodes occurring in 96 patients met the eligibility criteria (Supplementary Data, Figure S1). The excluded patients were similar to the study population, except for higher rates of intra-abdominal (16% vs. 5%, *p* = 0.04) and catheter (16% vs. 4%, *p* = 0.009) infections. Patients' characteristics and comorbidities are presented in Table 2. Antimicrobial treatment was classified as 80–100% adequate (good) in 47, 20–79% adequate (moderate) in 31, and 0–19% adequate (poor) in 32 patients.

Overall, the median BMI was 44.9 kg/m<sup>2</sup> (IQR 42–49), 20% (*n* = 22) of the patients had a BMI over 50 kg/m<sup>2</sup> , and 85% (*n* = 94) had at least one chronic underlying illness. The most frequent comorbidities were diabetes (69%, *n* = 76), coronary artery disease (32%, *n* = 35), and chronic obstructive pulmonary disease (25%, *n* = 27). One in five patients (21%, *n* = 23) had renal failure and 15% (*n* = 17) were immunocompromised. Apart from a significantly greater proportion of hemodialysis patients in the group with good adjustment, all other demographic variables and comorbidities were similar between groups.

The most common source of BSI was urinary tract infections (34%), followed by skin and soft tissue infections (25%). Infections in patients who had good antimicrobial adjustment were more severe, with a lower proportion of urinary tract infections (21% vs. 43%, *p* = 0.018), and a greater proportion of patients with a Pitt bacteremia score (PBS) ≥ 2 (68% vs. 40%, *p* = 0.003); there was a significantly higher frequency of sepsis in this group than in those with moderate or poor levels of adjustment (79% vs. 59%, *p* = 0.027). *Escherichia coli* was the most frequently isolated pathogen (28% of episodes). Enterobacteriaceae were recovered less often from patients with a good adjustment than from patients in the other groups (28% vs. 51%, *p* = 0.015).


**Table 2.** Patient demographics and medical conditions stratified by the level of appropriateness.

Results are reported as number (%) or median (IQR). Abbreviations: IQR: interquartile range, BMI: body mass index, COPD: chronic obstructive pulmonary disease, PBS: Pitt bacteremia score. <sup>a</sup> Bones and joints (6), cardiovascular (5), hepatic/biliary (5), catheter (4), undetermined (3), thoracic (2), central nervous system (1), vascular system (1). <sup>b</sup> Since it was present upon arrival or occurred early after initiation of antimicrobial treatment, sepsis at day 1 was considered a severity factor rather than a clinical outcome. <sup>c</sup> β-hemolytic (groups A, B, C and G) (27) and non-hemolytic streptococci (*S. gallolyticus, S. mitis*) (2). <sup>d</sup> *Escherichia coli* (31), *Klebsiella pneumoniae* (10), *Serratia marcescens* (3), *Citrobacter freundii* (2), *Enterobacter cloacae* (2), *Proteus mirabilis* (2), *Morganella morganii* (1), *Klebsiella oxytoca* (1). <sup>e</sup> *Enterococcus faecalis* (3), *Haemophilus influenzae* (2), *Aerococcus urinae* (1), *Bacteroides thetaiotaomicron* (1), *Bilophila wadsworthia* (1), *Clostridium septicum* (1), *Clostridium ramosum* (1), *Pasteurella multocida* (1), *Prevotella loescheii* (1), *Pseudomonas aeruginosa* (1).

During hospitalization, patients received an average of 3.1 ± 1.2 antimicrobials for their infection, of which 1.8 ± 0.9 had inadequate posology. The first prescription was unadjusted for the dose and/or the interval in 60% of patients (*n* = 66), and the dose was insufficient in 68% of the cases. Piperacillin-tazobactam (25%), ciprofloxacin (20%), and ceftriaxone (10%) were the most frequently non-adjusted antimicrobials.

More than half of the episodes (54%, *n* = 59) occurred after the implementation of APSS. There was a significant increase in the median appropriateness percentage of the treatment in the late-APSS period (84% [IQR 35–100], *p* = 0.031) compared with the other periods (pre-APSS: 27% [IQR 12–86]; early-APSS: 60% [IQR 9–97]) (Suppl. Data, Table S1). The proportion of inadequate prescriptions upon discharge was significantly lower (44% vs. 75% *p* = 0.02) in the late-APSS than in the pre-APSS period. Further, consultation with an infectious disease specialist was more frequent among patients with a good level of appropriateness than in the other categories (66% vs. 37%, *p* = 0.002).

#### *3.1. Outcomes*

The clinical outcomes (hospital outcomes and 30-day outcomes) are presented in Table 3. Overall, 53% (*n* = 58) of patients were admitted to the ICU and the median time to ICU admission was 7.4 h (IQR 4.0–14.9). Patients in the good appropriateness category tended to be admitted sooner (5.6 h IQR 3.7–10.4, *p* = 0.25) compared with the other groups (9.4 h [IQR 5.0–25.4]; 8.0 h [IQR 4.0–21.8]). Time from admission to first effective antimicrobial did not differ between groups (*p* = 0.84). The first antimicrobial was administered before ICU admission in most patients (89%, *n* = 98), but half of the patients who received their first antimicrobial in the ICU were in the good level of appropriateness. The patients in this group experienced more sepsis on days three and five and required more mechanical ventilation.


**Table 3.** Clinical outcomes stratified by the level of appropriateness.

Results are reported as number (%) or median (IQR). Abbreviations: IQR: interquartile range, WBC: white blood cells, ICU: intensive care unit, LOS: length of stay. <sup>a</sup> At least one day with mSOFA ≥ 2, on days 1, 3, 5, 7. <sup>b</sup> In cases with poor appropriateness, only 2 patients were readmitted with time to readmission of 9 and 18 days. \* Statistically significant difference (*p* value < 0.05), reference category: good (80–100%).

Although more patients from the good and moderate appropriateness groups were readmitted within 30 days from discharge and had high mortality rates, these differences did not reach statistical significance.

#### *3.2. Factors Associated with Adjusted Antimicrobial Therapy*

Factors associated with antimicrobial dosing adjusted for obesity are presented in Table 4. In the adjusted model, the number of antimicrobials (aOR 2.2, 95% CI 1.4–3.4,), consultation with an infectious disease specialist (aOR 3.3, 95% CI 1.3–8.6), blood urea nitrogen (BUN) 7–10.9 mmol/L (aOR 7.3, 95% CI 1.8–29.5), and hemodialysis (aOR 10.30, 95% CI 1.62–65.56) were significantly associated with high appropriateness. BMI >50 or weight >120 kg was not associated with adjusted antimicrobial dosing.


**Table 4.** Factors associated with adjusted antimicrobial therapy.

Reference category: 0–19% poor (vs. 20–100%). Abbreviations: BUN: blood urea nitrogen, APSS: Antimicrobial Prescription Surveillance System, PBS: Pitt bacteremia score.

#### *3.3. Factors Associated with Unfavourable Outcomes*

Overall, 55% (*n* = 60) of the patients had at least one of the following components of a composite outcome: ICU stay ≥72 h (33%, *n* = 36), duration of sepsis >3 days (34%, *n* = 37), LOS ≥7 days (55%, *n* = 61), and 30-day mortality (8%, *n* = 9). Risk factors for unfavourable outcomes are shown in Table 5. In multivariable analysis, age (aOR 1.07, 95% CI 1.02–1.12, *p* = 0.009), PBS ≥2 (aOR 7.30, 95% CI 2.09–25.52, *p* = 0.002), sepsis on day 1 (aOR 16.78, 95% CI 3.93–71.63, *p* < 0.001), and infection site (pulmonary aOR 7.52, 95% CI 1.20–47.15, *p* = 0.031, skin and soft tissue aOR 7.79, 95% CI 1.67–36.41, *p* = 0.009, others aOR 9.47, 95% CI 1.99–45.10, *p* = 0.005) were significantly associated with unfavourable outcomes. After adjustment, no measure of treatment appropriateness (first adjusted prescription, adjusted prescription within the first 72 h, and level of appropriateness) was associated with unfavourable outcomes.


**Table 5.** Factors associated with unfavourable outcomes.

Abbreviations: BMI: body mass index, PBS: Pitt bacteremia score, BUN: blood urea nitrogen, PD: pharmacodynamics.

#### **4. Discussion**

Since the prevalence of obesity continues to rise, and as individuals with obesity are likely to receive a high number of antimicrobials [25,26] and complex antimicrobial treatment [27], a better understanding of the impact of optimal dosing adjustment in patients with obesity is needed. In this study, we retrospectively assessed the impact of the appropriateness of antimicrobial dosing in patients with severe obesity hospitalized for BSI.

We observed low adherence to our local guidelines on the adjustment of doses for patients with severe obesity, with the first prescription being adequate in 40% of the episodes and the treatment being fully adequate in only 24% of the cases. These findings are consistent with those of previous studies, where recommendations (published or local guidelines) were rarely followed [28–31]. For instance, in patients with class III obesity, initial doses of vancomycin [28], ciprofloxacin, cefazolin, and cefepime [31]

were adequate in only 0%, 1.2%, 3%, and 8% of the cases, respectively. However, in our centre, a computerized clinical decision support system designed to assist the antimicrobial stewardship program team [24] had an impact on the prescriptions for patients with obesity, as shown by a three-fold increase in the median appropriateness of the antimicrobial treatment from the pre-APSS to the late-APSS period. In addition, patients who benefited from a consultation in infectious diseases had a higher likelihood of receiving a dosage adjusted for severe obesity than those who did not. Other factors associated with a high likelihood of adjustment were the number of antimicrobials, BUN between 7 and 10.9 mmol/L, and hemodialysis.

In the univariable analysis, we initially found a significant association between good prescription adjustment (>80%) and the occurrence of unfavourable outcomes. This association is counterintuitive as it implies that optimized dosage leads to negative outcomes. However, the association ceased to exist after adjustment for disease severity and the presence of sepsis on day one. It is common practice to increase antimicrobial dosage in the sickest patients, given their altered antimicrobial pharmacokinetics [32,33]. The presence of severe obesity in these patients is an additional reason to adjust the dosage upwards [5,19]. Finally, the wide therapeutic index of most antimicrobials used in this setting favours adjustments towards higher doses, given the imbalance between the severity of their condition and the low risk of adverse effects associated with overdosing with most antimicrobials. The same pattern was observed when we used adjustment of the first dose or adjustment within the first 72 h of treatment to measure the level of dosage optimization. Interestingly, we found a negative association between secondary outcomes and level of adjustment, but it did not reach statistical significance.

The literature on the impact of dose adjustment on clinical outcomes in patients with severe obesity treated for infection is scarce. In one study, high doses of tigecycline (100 mg every 12 h) administered to patients with obesity significantly improved clinical outcomes by reducing mortality, ICU stay, recurrent infections, and septic shock events [10]. However, this retrospective cohort study was limited by the small sample (only 11 patients with obesity), and the authors did not adjust for potential confounding factors. In another study, in a subgroup analysis of patients with severe obesity hospitalized for cellulitis, a low antimicrobial dose (TMP-SMX 1 DS PO twice a day or clindamycin 150–300 mg PO every 6–8 h) was associated with a high rate of clinical failure after discharge [7]. Again, this study was limited by its small sample (46 patients with severe obesity) and by the selection of unusual agents for cellulitis treatment [34]. Finally, inadequate dosing but neither weight nor obesity was associated with clinical failure in another study, and patients weighing ≥120 kg were more likely to receive adequate doses of TMP-SMX upon discharge [35]. Time-dependent killing antimicrobials were also associated with worse outcomes but this association did not remain significant after adjustment for covariates. Most patients in our study (*n* = 60, 55%) received both time-dependent and concentration-dependent killing antimicrobials. Since 2010, β-lactams, especially piperacillin-tazobactam have generally been administered as prolonged infusions in the ICU of our center to improve drug exposures. Prolonged perfusion is an important strategy to optimize PD parameters in β-lactams (increasing the time that concentrations remain above the minimum inhibitory concentration (MIC)), rather than only increasing the dose [36,37]. Besides, MIC values and organisms must be considered when assessing effectiveness and outcomes. In our cohort, the impact of bacterial resistance was limited because we excluded episodes where the pathogen was resistant to the antimicrobial received for more than 48 h.

The PBS was chosen to determine the severity of BSI, because it is simple to calculate, and has been described to better predict outcomes in patients with sepsis (which represented 71% of our cohort) than the Acute Physiology and Chronic Health Evaluation II (APACHE II) [17]. Moreover, in retrospective studies, complex scores such as APACHE II or SOFA are likely to be unhelpful due to missing values. We used the modified SOFA (mSOFA) score to limit the impact on missing variables, due to the retrospective nature of our study [22,23].

This study has several limitations. First, the study is subject to biases and missing data due to its retrospective design. Therefore, serum concentrations were not standardized, and it was impossible to perform pharmacokinetic calculations and measure concentrations to assess the appropriateness of the patients' regimens. Consequently, we decided not to include vancomycin. However, our local guidelines related to adjustments are based on published studies in which PK/PD data were available. We could not assess microbiologic clearance or the duration of BSI because blood samples are not routinely collected after the onset of BSI or they are but at various intervals. To our knowledge, no validated criteria exist to quantify the level of appropriateness. Our classification is subjective, but has been reviewed by two infectious disease specialists (A.C. and L.V.) for its clinical relevance. This classification, although far from being perfect, provides the reader with an order of magnitude regarding adjustment, but further research on this topic is needed. Finally, a posteriori, the study was limited by its small sample size, and it had 38% power to detect a difference of 15% in unfavourable outcomes between patients with and without dose adjustment, coming from a single centre. The small sample size could be explained by the limited proportion of patients hospitalized in our centre with severe obesity and infection treated with antimicrobials requiring adjustment.

However, despite these limitations, this study is the first to evaluate the association between adjustment for obesity and outcomes in patients with severe obesity and BSIs, such as urinary tract infection, pneumonia, cholangitis, and skin and tissue infections. We could assess short- and medium-term outcomes in several types of infections from various sites and of various severities, from mild symptoms to septic shock, thus providing an overview on the need for and impact of dose adjustment for patients with obesity. Most importantly, each prescription was evaluated considering renal function, which may have changed during hospitalization.

#### **5. Conclusions**

In conclusion, after adjustment for confounding factors, we did not find an association between dosing optimization and better outcomes in this cohort of patients with severe obesity and BSIs. However, in the absence of measured concentrations of antimicrobials, links between adjusted doses and outcomes can hardly be made. This study was exploratory and ideally a prospective study with the dosage of antimicrobials would be needed. This would maybe allow identifying a link between the adjustment and the outcomes, which we were unable to demonstrate. We did not find any study investigating the link between antimicrobial adjustment and outcomes like we did, in patients with class III obesity hospitalized for a bloodstream infection. Our study is intended as a first step in a field where knowledge remains extremely limited. Yet, given the wide therapeutic index of most antimicrobials and the trend of their effect on secondary outcomes, mortality, and morbidity associated with BSIs, and PK/PD data, it would be wise to continue to adjust antimicrobials upwards in patients with severe obesity and BSIs, while we wait for further evidence. Prolonged infusions also remain important strategies in optimizing PD as they may have a greater influence than dose increment. Finally, we have shown the positive impact of consultations with infectious disease specialists and an antimicrobial stewardship program based on an expert system in increasing the adherence to antimicrobial dosing adapted to patients with obesity.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2079-6382/9/10/707/s1, Figure S1: Flowchart of included and excluded episodes of bloodstream infections, Table S1: Variables related to the appropriateness according to the period of admission.

**Author Contributions:** Conceptualization, S.S., C.N.A.C., M.-F.L., J.P., A.C. and L.V.; methodology, S.S., C.N.A.C., M.-F.L., J.P. and L.V.; formal analysis, S.S., C.N.A.C. and L.V.; investigation, S.S.; writing—original draft preparation, S.S.; writing—review and editing, S.S., C.N.A.C., J.P., M.-F.L., A.C., L.V.; supervision, M.-F.L. and L.V.; funding acquisition, L.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by The Canadian Medical Protective Association (CMP2519).

**Conflicts of Interest:** L.V. is a shareholder and medical advisor to Lumed Inc. All other authors: none to declare.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

#### **E**ff**ective Treatment for Uncomplicated Urinary Tract Infections with Oral Fosfomycin, Single Center Four Year Retrospective Study**

#### **Miroslav Fajfr 1,2 , Michal Balik 2,3 , Eva Cermakova <sup>4</sup> and Pavel Bostik 1,2,5,\***


Received: 16 July 2020; Accepted: 11 August 2020; Published: 13 August 2020

**Abstract:** Fosfomycin represents a relatively old antibiotic, but it is experiencing a comeback in recent years. According to some studies, the increasing therapeutic use of this drug led to a rapid increase in the levels of resistance in bacteria causing urinary tract infection. In the presented study, levels of resistance to fosfomycin in more than 3500 bacterial isolates before and after fosfomycin introduction into therapeutic use in the Czech Republic and the clinical efficacy of treatment in 300 patients using this drug were assessed. The results show that the resistance levels to fosfomycin in *Escherichia coli* isolates before and after the drug registration were not significantly different (3.4% and 4.4%, respectively). In some other Gram-negative rods, such as otherwise susceptible *Enterobacter*, resistance to fosfomycin increased significantly from 45.6% to 76.6%. Fosfomycin treatment of urinary tract infections showed an excellent seven-day clinical efficacy (79.7%). However, when used to treat recurrent or complicated urinary tract infections, fosfomycin treatment was associated with high levels of infection relapse, leading to relapse in a total of 20.4% of patients during the first two months. This indicates that fosfomycin exhibits good efficacy only for the treatment of uncomplicated urinary tract infections

**Keywords:** fosfomycin; urinary infection; resistance

#### **1. Introduction**

Fosfomycin was introduced for the first time in 1969 as a product of *Streptomyces fradiae*, and it was also isolated from some members of the *Pseudomonas* species [1]. This drug was used for a long time in the treatment of urinary tract infections (UTIs), but the development of newer antibiotics led to a gradual decrease in its use. However, with the development of bacterial resistance to many antibiotics (i.e., β-lactam antibiotics, quinolones, aminglycosides) worldwide, fosfomycin is coming back as a viable alternative. Many countries adopted fosfomycin trometamol in their guidelines for urinary tract infection management not only for uncomplicated infections, but also for infections caused by multidrug-resistant bacteria [2,3]. However, this was associated, in turn, with a rapid increase in the resistance level of bacteria to fosfomycin, according to data from several countries. Data from Spain show an increase in Fosfomycin-resistant *Escherichia coli* extended-spectrum beta-lactamase (ESBL)-producing isolates from 4.4% in 2005 to 11.4% in 2009. However, the overall resistance level of all

*Escherichia coli* isolates (ESBL producers and ESBL non-producers) to fosfomycin in this study remained low at 2.9% in 2019. Another study from China showed that a high percentage of carbapenemase (KPC)-producing *Klebsiella pneumoniae* isolates harbor FosA3 (34%)*,* which would suggest the loss of effectiveness of fosfomycin and the multidrug resistance (MDR) of these bacteria. Data from Poland show the overall susceptibility of *E. coli* isolates to fosfomycin to be 62.2% in complicated UTIs and 77.6% in uncomplicated UTIs [4–6]. These data, thus, indicate a potential problem in the use of fosfomycin in the treatment of nosocomial acquired urinary tract infections (UTIs). *Antibiotics* **2020**, *9*, x FOR PEER REVIEW 2 of 10 increase in the resistance level of bacteria to fosfomycin, according to data from several countries. Data from Spain show an increase in Fosfomycin-resistant *Escherichia coli* extended-spectrum betalactamase (ESBL)-producing isolates from 4.4% in 2005 to 11.4% in 2009. However, the overall resistance level of all *Escherichia coli* isolates (ESBL producers and ESBL non-producers) to fosfomycin in this study remained low at 2.9% in 2019. Another study from China showed that a high percentage of carbapenemase (KPC)-producing *Klebsiella pneumoniae* isolates harbor FosA3 (34%)*,* which would suggest the loss of effectiveness of fosfomycin and the multidrug resistance (MDR) of these bacteria.

In the Czech Republic, fosfomycin trometamol was not licensed for clinical use until October 2014. The resistance of bacteria was, therefore, very low, as described in our previous work and as also reported from other countries [7,8]. In 2015, oral fosfomycin trometamol was implemented into the Czech national UTI treatment guidelines as a drug of second choice for uncomplicated infections of the lower urinary tract. Here, we describe a study of the clinical effects of fosfomycin trometamol use in the University Hospital in Hradec Kralove, Czech Republic, and we evaluate trends in bacterial susceptibility to fosfomycin during the first four years of its use for the treatment of UTIs. Data from Poland show the overall susceptibility of *E. coli* isolates to fosfomycin to be 62.2% in complicated UTIs and 77.6% in uncomplicated UTIs [4–6]. These data, thus, indicate a potential problem in the use of fosfomycin in the treatment of nosocomial acquired urinary tract infections (UTIs). In the Czech Republic, fosfomycin trometamol was not licensed for clinical use until October 2014. The resistance of bacteria was, therefore, very low, as described in our previous work and as also reported from other countries [7,8]. In 2015, oral fosfomycin trometamol was implemented into the Czech national UTI treatment guidelines as a drug of second choice for uncomplicated infections of the lower urinary tract. Here, we describe a study of the clinical effects of fosfomycin trometamol

#### **2. Results** use in the University Hospital in Hradec Kralove, Czech Republic, and we evaluate trends in bacterial susceptibility to fosfomycin during the first four years of its use for the treatment of UTIs.

#### *2.1. Susceptibility of Bacteria Causing UTI to Fosfomycin* **2. Results**

The prevalence of individual bacterial strains isolated from urine samples was similar when comparing samples from two patient cohorts before (Cohort 1) and after (Cohort 2) the introduction of fosfomycin into the treatment of UTIs. The most frequent bacterium isolated was *Escherichia coli*, with levels of 46.1% and 49.7%, respectively. Other bacteria isolated more frequently than in 10% of samples were *Enterococcus faecalis* and *Klebsiella* species. The similar bacterial stratification allowed for a comparison of both cohorts in the next step. The entire spectrum of bacteria isolated is shown in Figure 1. *2.1. Susceptibility of Bacteria Causing UTI to Fosfomycin* The prevalence of individual bacterial strains isolated from urine samples was similar when comparing samples from two patient cohorts before (Cohort 1) and after (Cohort 2) the introduction of fosfomycin into the treatment of UTIs. The most frequent bacterium isolated was *Escherichia coli*, with levels of 46.1% and 49.7%, respectively. Other bacteria isolated more frequently than in 10% of samples were *Enterococcus faecalis* and *Klebsiella* species. The similar bacterial stratification allowed for a comparison of both cohorts in the next step. The entire spectrum of bacteria isolated is shown in Figure 1.

**Figure 1.** Proportional representation of bacteria isolated from urine samples in Cohort 1 (before fosfomycin registration) and Cohort 2 (after fosfomycin registration). Bacteria were stratified by families; Tribus *Proteae* includes *Proteus* spp., *Morganella* spp., and *Providentia* spp. **Figure 1.** Proportional representation of bacteria isolated from urine samples in Cohort 1 (before fosfomycin registration) and Cohort 2 (after fosfomycin registration). Bacteria were stratified by families; Tribus *Proteae* includes *Proteus* spp., *Morganella* spp., and *Providentia* spp.

The general level of resistance (beta-lactamase production, resistance to fluoroquinolones, and multidrug resistance) of Gram-negative bacteria showed a notable increase over time when analyzed in the older (Cohort1) and more recent (Cohort 2) samples (Table 1). Thus, the number of high-risk The general level of resistance (beta-lactamase production, resistance to fluoroquinolones, and multidrug resistance) of Gram-negative bacteria showed a notable increase over time when analyzed in the older (Cohort1) and more recent (Cohort 2) samples (Table 1). Thus, the number of high-risk beta-lactamase-positive (ESBL, AmpC, and K1) bacterial strains rose from 11.7% to 19.8%, and the number of fluoroquinolone-resistant ones rose from 25.9% to 33.8%. *Escherichia* species isolates generally showed lower levels of resistance to antibiotics with no significant increase over time (*p* = 0.23304). The highest level of resistance was detected in *Klebsiella* species isolates, but there was also no significant increase when comparing the two cohorts (*p* = 0.53870). On the contrary, multidrug-resistant (MDR) *Pseudomonas* isolates (defined as resistant to beta-lactams, carbapenems, and fluoroquinolones) showed a markable and statistically significant increase over time from 5.6% to 13.3% (*p* = 0.00026).

**Table 1.** Resistance levels of the most frequent pathogenic bacteria from urine samples before (Cohort 1) and after (Cohort 2) fosfomycin registration. Brackets indicate statistically significant differences (*p* < 0.05) between Cohort 1 and Cohort 2.


BL—high-risk beta-lactamases (ESBL, AmpC, K1); iAmpC—inducibile AmpC beta-lactamases; FQR—fluoroquinolone resistance; MDR—multidrug resistance; NO RES—no resistance; Tribus *Proteae*—*Proteus* spp., *Morganella* spp., and *Providentia* spp.

The analysis of resistance of the three most common Gram-negative rods to fosfomycin and its development over time is shown in Table 2. Eesistance to fosfomycin was generally low in all *Escherichia* isolates, with no statistically significant differences before and after the fosfomycin registration, with levels of 4.4% and 3.4%, respectively. However, both *Klebsiella* and *Enterobacter* isolates showed statistically significant increases in resistance after the introduction of fosfomycin. Thus, in the high-risk beta-lactamase-producing *Klebsiella* isolates, the resistance to fosfomycin increased from 87% to 95.2% (*p* = 0.0467). In the *Enterobacter* isolates with no other resistance detected, fosfomycin resistance increased dramatically from 45.6% to 76.6% (*p* = 0.0105). However, the overall increase in fosfomycin resistance in all *Enterobacter* isolates was also high and statistically significant (55.2% vs. 74.4%; *p* = 0.0468). Taken together, of the more prevalent bacterial isolates from the UTIs that showed other drug resistances (high-risk beta-lactamases production or resistance to fluroquinolones), only the *E. coli* isolates were sufficiently susceptible to fosfomycin both before and after its introduction into the treatment regimes.


**Table 2.** Resistance levels of the most frequent Gram-negative bacteria from urine samples to fosfomycin before (Cohort 1) and after (Cohort 2) fosfomycin registration. Brackets indicate statistically significant differences (*p* < 0.05) between Cohort 1 and Cohort 2.

BL—high-risk beta-lactamases (ESBL, AmpC, K1); iAmpC—inducible AmpC beta-lactamases; FQR—fluoroquinolone resistance; NO—no resistance.

#### *2.2. Clinical E*ffi*cacy of Fosfomycin*

The clinical effects were analyzed in 300 patients treated with fosfomycin trometamol for UTIs (Cohort 3). The basic demographic data of the patients are shown in Table 3. In total, 428 fosfomycin trometamol doses were prescribed, i.e., 1.43 doses on average for a patient, with a maximum of four doses per patient. Altogether, 66.0% (*n* = 198) of patients were administered a single dose and 34.0% (*n* = 102) of patients received multiple doses. Multiple-dose therapy was used mostly in patients with complicated (in 51.0% of these cases) or recurrent (in 43.7% of these cases) infections, but it was relatively rare in patients with uncomplicated infections (in 19.5% of these cases). In 50.0% of patients, bacterial cultivation was performed before fosfomycin treatment, while 50.0% of patients were treated empirically without the culture. The type of UTI was clearly specified in 234 cases. In 67.1% (*n* = 157) of patients, complicated/recurrent UTIs were diagnosed, while 32.9% of patients had an uncomplicated UTI (all females). Nearly 30% of patients (*n* = 88) did not come for a follow-up visit, and the therapeutic effect could not be evaluated.


**Table 3.** Basic demographic data of patients from Cohort 3 (patients evaluated for fosfomycin treatment efficacy). UTI—urinary tract infection.

\* Only patients with available data.

The clinical effects of fosfomycin treatment in the individual groups are shown in Table 4. Among those, where the effects could be evaluated, fosfomycin had an immediate curative effect in 79.7% of cases (*n* = 169); however, in 20.1% of cases (*n* = 43), the drug had no positive effect. The highest fosfomycin immediate cure rate was observed in patients with infections caused by *Escherichia* species (69.8%) and *Enterococcus faecalis* (50.0%). In cases with no detectable antibiotic effect of fosfomycin, the bacterial culture (before antibiotic administration or control culture) was performed in 79.1% of patients. In one-third of these patients, susceptible bacteria were detected (*Escherichia* species or *Enterococcus faecalis*), while the culture was negative in 26.5% of cases (*n* = 9) and 14.7% (*n* = 5) had mixed culture; the remaining cases involved bacteria with no EUCAST susceptibility breakpoints of fosfomycin in the disc diffusion test (*Klebsiella pneumoniae*, *Pseudomonas aeruginosa*, *Streptococcus agalactiae*).

**Table 4.** Results of clinical efficacy evaluation of fosfomycin treatment, case distribution according to UTI type, and clinical follow-up after fosfomycin administration. Statistically significant differences (at *p* < 0.05) are shown in brackets.



The second evaluation criterion was the recurrence level after the administration of fosfomycin up to one year after drug use. Sufficient data could only be obtained from 186 patients, since 114 patients escaped the evidence of the Urology Clinic. The fact, the relatively high number of patients with uncomplicated UTIs that were not present for their follow-up visit may indicate that the antibiotic had a curable effect (according to standard procedures, in the case of successful treatment, no follow-up visit is needed). However, since there was no evidence for it, these patients were excluded from the analysis. From the included cases, 40.9% (*n* = 76) had no relapse of the same UTI, 21.5% (*n* = 43) had a relapse during the first two months after fosfomycin treatment, and 14.5% (*n* = 27) of patients had recurrence in the time period of 2–12 months after fosfomycin administration. From the patients with an early relapse, 57.5% were administered a single dose and 42.5% were administered multiple doses of the drug. In most cases, fosfomycin had no effect or the UTI showed early recurrence in patients with complicated UTIs (complicated UTIs created more than 90% of acute fosfomycin failure cases or cases with early recurrence). However, in patients with uncomplicated UTIs, cure failure or recurrence was very rare (5.6–11.1%). Thus, the statistical evaluation showed a much lower relapse rate in the group with uncomplicated UTIs in comparison to complicated or recurring UTIs for all three analyzed time periods, which was highly significant (*p* = 0.00004).

The dosage impact on immediate effect was not seen, as the same percentage of single-dose therapy (60.4% and 60.3%) was found in cases with no cure effect and with immediate cure effect.

#### **3. Discussion**

In this study, urine samples from patients experiencing urinary tract infections were analyzed with the aim to assess the efficacy of fosfomycin treatment and to identify whether the introduction of this drug into the treatment regime led to any increase in antimicrobial resistance to fosfomycin. Although many patients did not come for their follow-up visits, thereby decreasing the numbers available for analysis, the size of our cohort is comparable to the studies cited throughout this report and elsewhere [9]; therefore it represents, in our opinion, a valid set of data, which can be used for further treatment guidance.

Our study showed a relatively stable proportion of bacterial species isolated from the UTIs with a dominance of *E. coli*, *Enterococcus faecalis*, and *Klebsiella* species. These bacterial species represent the most prevalent isolates from urinary tract infections worldwide [10,11]. Previous studies showed that fosfomycin remains very effective for the treatment of urinary tract infections caused by various bacteria, including *E. coli* or enterococci, in which the susceptibility to the drug was reported around 90% [7,10,12,13]. This was fully confirmed by our data, where the overall susceptibility in both cohorts (pre- and post-fosfomycin introduction) was 96.5% for all *E. coli* isolates and 100.0% for *Enterococcus faecalis* (data not shown). While there is some evidence of similarly high susceptibility to fosfomycin in ESBL-producing bacteria [13], most studies showed significantly lower numbers (around 80%) than in non-beta-lactamase producers [7,10,14]. This overall trend was also reflected in the data presented here. For example, while the resistance level in non-ESBL isolates of *E. coli* was 3.0%, the resistance increased to 9.7% in ESBL-positive isolates.

According to a previously published meta-analysis, the risk of selecting resistant mutants during fosfomycin monotherapy was calculated at 3.4% [15]. This is likely why some countries, such as China, Spain, Turkey, and some regions of India, reported alarmingly increasing fosfomycin resistance levels associated with its increased application. Thus, the fosfomycin resistance levels of non-beta lactamase-producing *E. coli* reached as high as 15% in some of these countries [16]. On the contrary, in some other areas, including Japan and the majority of the European and American countries, fosfomycin resistance was maintained at low levels around 4.5% despite the increase in drug use [12,16,17]. Moreover, reports from Hungary showed inconsistent fosfomycin resistance levels of *Citrobacter*, *Enterobacter*, and *Serratia* isolates over five years of study, with an overall resistance level of 9.3% in outpatient and 13.8% in inpatient settings. The resistance level of Tribus *Proteae* members to fosfomycin from the same country was calculated to be 18.7% in inpatients and 30.3% in outpatients

over 10 years of study [18,19]. A study from Israel conducted during a time period almost identical to that in our study showed a notable increase in overall fosfomycin resistance in Gram-negative bacteria from 20.7% in 2015 to 30.9% in 2016 [20]. Our data show low resistance levels in *Escherichia coli* isolates and high resistance levels in isolates of *Klebsiella* and *Enterobacter.* With the exception of *Enterobacter,* the resistance levels were, however, maintained over time with no demonstrated effect of increased fosfomycin use after its registration. The reason for this relatively stable susceptibility to fosfomycin in our country can be explained by the relatively infrequent prescription of this drug in the Czech Republic overall. In our system, fosfomycin is fully paid for by patients with no contribution from the health insurance system; as such, it is not usually the first drug of choice.

Fosfomycin was reported as a relatively effective drug for the treatment of UTIs. For example, one United States (US) study reported a microbiological cure rate of 59.0% [21] and a similar Chinese study showed an overall 15-day cure rate of 65.07%, with variability in accordance with gender, age, and infection type [22]. Our data show an overall seven-day clinical cure rate of 79.7%, which is slightly higher than the rates mentioned above. One potential explanation for the better results in our study may lie in the differences in the spectrum of bacteria isolated. Thus, in our cohort extended-spectrum beta-lactamase-producing bacteria represented only 14.0% of isolates, while, in the studies above, the numbers of multidrug-resistant isolates were higher. Our data also confirmed the correlation between the treatment efficacy and the type of UTI. Thus, a report from China showed the effectivity of fosfomycin treatment in 97.71% of uncomplicated and 62.69% of complicated UTIs [22], while our data show efficacies of 94.8% and 70.2%, respectively. A high percentage of our patients showed relapses of UTIs in our study. In total, 21.5% of patients treated with fosfomycin experienced a relapse of infection in the first two months after therapy. Fosfomycin therapy in the complicated UTIs was, according to our data, ineffective in 51.3% of cases (i.e., the drug has no effect, or there is early relapse of symptoms within two months). In the treatment of complicated UTIs or UTIs caused by multidrug-resistant microbes, many studies suggested multiple dose regimens of fosfomycin [1,21–26] with three-dose regimens being the most common. Patients in our cohorts were mostly treated with a single-dose regimen; however, about one-third of them received multiple doses. Our data support the approach to apply multiple doses of fosfomycin in complicated UTIs, because early relapses were much more frequent when single-dose regimens were used. The treatment assessment studies mostly showed multiple dose regimens within the range of 1.4 to 2.0 doses per patient in accordance with the NICE guidelines, with administration of single-dose treatments in female and double-dose regimens in male patients. In our study, the average regimen consisted of 1.4 doses per patient, and multiple-dose regimens were used more frequently in the male subgroup (in 48.7% of cases, compared to 29.0% in females), which is fully in accordance with other published data [9,21,27]. Some studies showed a significantly higher failure level in UTIs associated with *Klebsiella* infections than in those caused by *E. coli* [9,27]. Similarly, we found a higher failure rate in infections caused by *Klebsiella* isolates (35.7%) than *E. coli* (17.0%), but the highest treatment failure rate of 44.4% was present in UTIs associated with bacterial mixtures (combinations of *Kl. pneumoniae*, *E. coli*, *Proteus* spp., or *Enterococcus*).

The strength of the presented study is in the comparison of resistance levels in a naïve bacterial population (before oral fosfomycin use) with those in a bacterial population after the introduction and wide use of fosfomycin in treatment regimens. The other important aspect of the presented data lies in the evaluation of oral fosfomycin cure effectivity in urinary tract infections. Although our study had a limitation in the relatively low number of patients (in part due to the fact that many patients did not come for their follow-up visits), the size of our cohort is comparable to those cited in studies throughout this report and elsewhere (e.g., 24). The fact, that a relatively high number of patients with uncomplicated UTIs did not present for their follow-up visit may indicate that the antibiotic had a curable effect. According to standard procedures, in the case of a successful treatment, no follow-up visit is required; thus, the long-term curable effect of oral fosfomycin in uncomplicated urinary tract infections may, in fact, be even better. Nevertheless, since this could not be supported by any evidence, these patients were excluded from the analysis.

#### **4. Materials and Methods**

#### *4.1. Patient Cohorts and Samples*

Urine samples were obtained from patients treated at the Urology Clinic, University Hospital in Hradec Kralove, Czech Republic during the period 2013–2018. Samples were collected from patients with UTI signs including dysuria, flank pain, urinary frequency or urgency, leukocyturia and/or positive culture, and fever. Negative samples and all samples positive for the following microorganisms were excluded from the study: yeast, coagulase-negative staphylococci (with the exception of *Staphylococcus saprophyticus*), and other non-pathogenic bacteria (e.g., lactobacilli, corynebacteria). Moreover, all positive isolates without available fosfomycin susceptibility data were excluded.

The antimicrobial resistance of bacteria to fosfomycin prior to and following the drug's introduction into the UTI treatment guidelines was evaluated in samples from two cohorts of patients. Cohort 1 (before fosfomycin registration) consisted of samples collected in the period 2013–2014 and contained a total of 594 bacterial isolates. Cohort 2 (post fosfomycin introduction) consisted of 2935 bacterial isolates obtained from patients treated at the Urology Clinic in the years 2015–2018. The evaluation of fosfomycin treatment efficacy was performed in 300 patients (Cohort 3) treated at the Urology Clinic in the period 2015–2018 for UTIs. Of these patients, 25.7% (*n* = 77) had uncomplicated UTIs (defined as acute cystitis in women or acute cystitis in young men), 22.0% had undefined UTIs, and 52.3% (*n* = 157) had complicated UTIs (defined as recurrent UTIs, UTIs after urinary tract surgery, UTIs in men with benign prostate hyperplasia or carcinoma, or UTIs associated with lithiasis). Data from the electronic records of patients and microbiological test results were used for the assessment of clinical outcome. The patient data were strictly anonymized; thus, according to the Ethics Committee of the University Hospital, no patient consent was needed.

#### *4.2. Susceptibility Evaluation*

Standard Mueller–Hinton agar (Trios, Czech Republic) and a disc diffusion test with a 200 µg fosfomycin disc with the 50 µg of G-6-P (glucose-6-phosphate) supplement were used according to the EUCAST and CLSI guidelines. For the susceptibility evaluation of *Enterobacteriales*, the EUCAST guideline was applied (EUCAST Clinical breakpoint tables v. 09, valid from 2019-01-01). *Escherichia coli* isolates were considered susceptible to fosfomycin in the case of zone diameters≥24 mm. Similar criteria were adopted to show the resistance levels in two other Gram-negative bacteria (*Klebsiella* species and *Enterobacter* species) for illustration, as used in some previous publications [10,14]. For other bacteria (*E. faecalis*), the CLSI susceptibility criteria (M100 S29, valid from 2019-01-01) were used. The following susceptibility criteria for *E. faecalis* were used: zone diameters ≥16 mm indicated susceptible isolates and zone diameters ≤ 12 mm indicated resistant ones.

#### *4.3. Clinical E*ffi*cacy Evaluation*

The clinical effects of the applied treatment were evaluated by urologists according to the criteria described below. The classification of UTIs was adopted from EAU Urological Infection Guidelines 2020 with three groups presented: uncomplicated, complicated, and recurrent (classification available on EAU website: https://uroweb.org/guideline/urological-infections/#3). The first assessment was the drug's acute effect, meaning the curative effect in the first seven days after drug administration; cases were classified as cured (the symptoms of UTI disappeared and/or negative culture result in control sample), no effect or acute failure (without any positive effect; the UTI symptoms were present despite the treatment), or not assessed (no data available, patients did not come for follow-up). The second criterion was the recurrence of the same clinical unit in the first year after drug administration with a focus on the first two months, which would indicate a relapse and treatment failure. All these data were correlated with the culture results, dose regime, and type of UTI.

#### *4.4. Statistics*

Statistical evaluation was performed using Pearson's chi-squared test with a significance level of *p* = 0.05. In the case of low numbers for the chi-squared test, the modified Fisher's exact test was used. The statistical evaluations were performed using the NCSS 11 Statistical Software (2016) (NCSS, LLC. Kaysville, UT, USA, ncss.com/software/ncss).

#### **5. Conclusions**

Our data confirmed the high susceptibility of Escherichia coli and Enterococcus faecalis to fosfomycin. The results did not show any increase of resistance levels after the registration and introduction of this drug to UTIs treatment regimes. The oral fosfomycin showed a high treatment efficacy in uncomplicated urinary tract infections. However, in complicated or recurrent infections the treatment led to a relatively high recurrence index.

**Author Contributions:** Conceptualization, M.F.; methodology, M.B. and M.F.; validation, M.F.; formal analysis, M.F.; investigation, M.B. and M.F.; resources M.B.; data curation and statistical analysis, E.C.; writing—original draft preparation, M.F.; writing—review and editing, P.B.; supervision, P.B. and M.F.; funding acquisition, P.B. All authors read and agreed to the published version of the manuscript.

**Funding:** This research was funded supported by the Ministry of Health of the Czech Republic DRO plan (UHHK, 00179906) and by the long-term organization development plan by the Ministry of Defense of the Czech Republic.

**Conflicts of Interest:** The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Review* **Transitioning of** *Helicobacter pylori* **Therapy from Trial and Error to Antimicrobial Stewardship**

#### **David Y. Graham**

Department of Medicine, Michael E. DeBakey VA Medical Center and Baylor College of Medicine, RM 3A-318B (111D), 2002 Holcombe Boulevard, Houston, TX 77030, USA; dgraham@bcm.edu; Tel.: +713-795-0232; Fax: +713-795-4471

Received: 15 September 2020; Accepted: 1 October 2020; Published: 3 October 2020

**Abstract:** *Helicobacter pylori* is the only major infection for which antimicrobial therapy is not designed using the principles of antimicrobial stewardship. Traditionally, antimicrobial therapy is a susceptibility-based therapy, achieves high cure rates, and includes surveillance programs to regularly provide updated data regarding resistance, outcomes, and treatment guidelines. Current *H. pylori* therapies identified by trial-and-error, and treatment recommendations and guidelines are based on comparisons among regimens that rarely take into account the prevalence or effect of resistance. The majority of patients currently treated achieve suboptimal results. A paradigm shift is required to abandon current approaches and embrace antimicrobial stewardship, and therefore reliably achieve high cure rates; develop, propagate, and update best practice guidelines; and provide surveillance of local or regional susceptibility/resistance patterns. These also require timely updates to clinicians regarding the current status of resistance, antimicrobial effectiveness, and ways to prevent antimicrobial misuse to extend the useful life of currently available antibiotics. Here, we discuss the differences among current approaches to *H. pylori* therapy and antimicrobial stewardship and identify what is required to achieve the transition. Conceptually, the differences are significant, and the transition will likely need to be both abrupt and complete. Recommendations for therapy during the transition period are given.

**Keywords:** *Helicobacter pylori*; antimicrobial stewardship; therapy; antibiotics; metronidazole; clarithromycin; fluoroquinolones; amoxicillin; proton pump inhibitors; ethical trials

#### **1. Introduction**

The widespread misuse of antibiotics has resulted in increasing antimicrobial resistance which threatens the continued usefulness of currently available antimicrobials. The consequences of critical antibiotics becoming clinically ineffective have resulted in strong pressure to utilize the principles of antimicrobial stewardship for selecting and managing therapy for infectious diseases [1,2]. Management of *Helicobacter pylori* infections has long been within the purview of *Gastroenterology* which has been slow to accept the paradigm shift needed to change the approach to therapy from one utilizing trial-and-error to one based on the principles of antimicrobial stewardship. *H. pylori* was officially recognized as an infectious disease in 2015 [3], and those involved in studying *H. pylori* therapy are only now gradually beginning to accept that *H. pylori* therapy should no longer be exempt from the guidelines and practices governing treatment of other infectious diseases (i.e., the principles of antimicrobial stewardship).

#### **2. Antimicrobial Stewardship in Traditional Infectious Disease Therapy**

The principles of antimicrobial stewardship codified and extended thinking and practices involved in the development and implementation of methods to simultaneously improve antimicrobial therapy, prevent antimicrobial misuse, and reliably achieve high cure rates, while minimizing the risk of developing resistance in order to prolong the useful life of antibiotics. Until now, the approach to *H. pylori* therapy has not routinely involved elements critical to antimicrobial stewardship, such as optimization of the therapy in terms of drugs, dosing, or duration of therapy. As a result, current therapies have subsequently failed to reliably achieve high cure rates without prescribing unnecessary antimicrobials or prolonging therapy beyond the experimentally identified optimal duration. Stewardship elements that are also lacking include the development, propagation, and regular updating of best practice guidelines, as well as susceptibility/resistance surveillance to update clinicians of the current status of resistance and antimicrobial effectiveness, and the availability and potential role of new antimicrobials.

Fundamentally, a successful antimicrobial therapy is a susceptibility-based therapy, as success is predicated on the target organism being susceptible to the antimicrobial agents utilized. Antimicrobial therapies are also optimized so as to reliably achieve excellent results (i.e., high cure rates such as >95%) (Table 1) [4]. Comparisons among therapies are infrequent and, when done, compare proven highly effective regimens using noninferiority methodology [5–7]. As we discuss below, *H. pylori* therapy is most often prescribed empirically and most often fails to reliably achieve high cure rates [8]. Comparisons of empiric therapies are the rule and they focus on which is the better of the two regimens with seemingly little regard to the actual, often poor, cure rates achieved. In addition, consensus and guideline recommendations regarding *H. pylori* therapy rarely involve the principles of antimicrobial stewardship and are based on principles used for other gastrointestinal diseases, such as irritable bowel syndrome, which fail to reflect the marked differences and realities of antimicrobial therapy.


**Table 1.** Definitions of terms to describe outcome of therapy.

From [4], with permission.

#### **3. Original Development of** *H. pylori* **Antimicrobial Therapies**

Soon after the discovery of *H. pylori*, it was shown that, in vitro, the organism was susceptible to a wide variety of commonly used antimicrobials. However, the infection proved to be difficult to cure using those antimicrobials, and it also proved impossible to reliably achieve high cure rates [9] (reviewed in [10]). This prompted a period of trial-and-error during which clinical trials were done with a wide variety of antimicrobials (reviewed in [10]). Eventually Tom Borody in Australia discovered an effective three-drug regimen consisting of bismuth, metronidazole, and tetracycline called bismuth triple therapy regimen [11].

*H. pylori* therapy was initially focused on treatment of peptic ulcer disease which was extremely common and one of the major gastrointestinal causes of morbidity, surgery, and medical costs. The basic approach was to treat the ulcer with a histamine-2 receptor antagonist (H2RA) and the infection with bismuth triple therapy [12]. It was soon recognized that the presence of metronidazole resistance greatly reduced the effectiveness of bismuth triple therapy, but that this could be partially or completely overcome by adding a proton pump inhibitor (PPI) and increasing both the dosage of metronidazole and the duration of therapy (e.g., to 14 days) (reviewed in [13]). This modified regimen is called bismuth quadruple therapy.

Bismuth triple therapy was commercially marketed in 1996 and while widely used with a PPI, the commercialization of bismuth quadruple therapy, in 2006, was delayed until after the patent regarding the addition of a PPI expired.

#### **4. Development of** *H. pylori* **Therapies with Di**ff**erent Antibiotics**

One of the initial attempts to produce an *H. pylori* therapy was with the macrolide, erythromycin. Erythromycin was able to suppress the infection but was unable to cure the infection [14]. Thus, macrolides were considered to be ineffective in vivo, until the introduction of clarithromycin which, when combined with a PPI or H2RA to increase gastric pH, proved effective. The initial dual therapy combination of clarithromycin and a PPI ultimately failed because of emergence of clarithromycin resistance during therapy [15–17]. However, with the addition of amoxicillin, it proved both well tolerated and effective [17]. This three-drug regime is called clarithromycin triple therapy or, alternatively, standard triple therapy. The effectiveness of clarithromycin triple therapy subsequently resulted in the development of additional triple therapies consisting of PPI and amoxicillin plus metronidazole, a fluoroquinolone, or rifabutin.

The original bismuth quadruple therapy was also modified by replacing the metronidazole with amoxicillin or furazolidone. Other recent successful iterations include bismuth and a PPI plus tetracycline and amoxicillin or metronidazole and amoxicillin (reviewed in [13,18,19]). The most recent advance has been with a dual therapy consisting of omeprazole plus amoxicillin. This was originally introduced in 1989 but proved unable to reliably achieve high cure rates and, until recently, was generally considered to be a failure [20–22] (reviewed in [19]). However, the introduction of potassium competitive acid blockers (P-CABs), such as vonoprazan, have reinvigorated interest in dual therapies (see below) [23].

#### **5. The E**ff**ect of Gastroenterology Rather Than Infectious Disease Being Responsible for Development of** *H. pylori* **Therapies**

Within a few years of discovery, the diagnosis and management of *H. pylori* had primarily become the responsibility of gastroenterology. Gastroenterologists found little need, or desire, to incorporate susceptibility testing as a key element required for choosing, defining, optimizing, and evaluating the results of *H. pylori* therapies. Instead, they adopted the trial-and-error approach which focused on comparison of therapies. This approach has remained the standard despite evidence that overall cure rates were often both poor and declining.

The gastroenterologist's approach to the development of *H. pylori* therapies was most likely based on the long experience with common gastroenterology diseases, such as constipation or irritable bowel syndrome. Most diseases in gastroenterology are characterized as follows: (a) the cause is largely unknown, (b) there is a large placebo response to therapy, and (c) the success of most therapies is low requiring a comparator which is often a placebo. The fact that the etiologies were unknown required a focus on the results rather than an explanation for a poor response. This contrasts with infectious disease where the cause of a poor response is discoverable. These differences resulted in two different approaches, one that focused on outcome as a comparison (the what school) and the other that focused on the actual outcome (i.e., cure rate) (the why school). In gastroenterology, the "what school" has dominated planning, analysis, and reporting of clinical trials which can be best characterized as comparisons of arbitrary regimens focusing on the difference, despite the fact that often none of the arms achieved a high cure rate. This approach was abetted by medical journals which expected, even demanded, comparative trials. Eventually, thousands of patients were enrolled in *H. pylori* treatment trials of which many, if not most, achieved poor cure rates in at least one treatment arm. The fact that at least one arm in a trial achieved a poor result was, however, often not a result of chance. The published study design confirmed that based on prior experience, the authors reliably predicted which regime would produce the unacceptably low cure rate. By definition, informed consent requires this information to be shared with the subjects, but it was withheld [24,25]. Informed consent also

requires informing patients about any new information arising during the trial and allowing them to reconsider. This also was not done. These issues with informed consent have continued to plague *H. pylori* comparative trials up to the present.

The development of sequential therapy, probably, best illustrates the differences between gastroenterology and the traditional development of therapy for an infectious disease. Sequential therapy is a regimen in which a dual therapy with a PPI and amoxicillin are given for five days, followed by five days with the PPI, clarithromycin, and metronidazole. It was developed in Italy and was prompted by increasing treatment failures with triple therapy. By 2001, it had been conclusively shown (e.g., studies involving more than 53,000 patients) that, because of increased clarithromycin resistance, clarithromycin triple therapy was no longer able to reliably achieve high cure rates [26,27]. Clarithromycin triple therapy was used as the straw man in multiple studies in which more than 1800 patients participated. The investigators were able to show that, in the mid-2000s, in one region of Italy, sequential therapy yielded higher cure rates than clarithromycin triple therapy [28]. The fact that prior trials in the same population had already proven that one regime was inferior, was also withheld from the consent of subjects in subsequent trials. Despite many iterations of the same trials, sequential therapy was neither optimized nor explored for its weaknesses. When sequential therapy was tested in different geographic areas with different resistance patterns (e.g., resistance to clarithromycin and metronidazole vs. increased resistance to clarithromycin but low resistance to metronidazole) sequential therapy proved ineffective and was abandoned [28,29].

The traditional infectious disease approach focuses on attainment of a prespecified cure rate (e.g., ≥95%) (i.e., the why school) using susceptibility-based therapy and would never have been included in the empiric comparisons described above. Attempts to optimize sequential therapy would have discovered that (a) the duration of therapy of 14 days provided a higher cure rate, (b) that sequential therapy was only effective in the presence of isolated clarithromycin resistance, and (c) that all those with clarithromycin resistance received the clarithromycin with no benefits. As noted previously, comparisons of highly effective susceptibility-based therapies are rare and, when preformed, are generally limited to head-to-head comparisons of proven, highly reliable, optimized regimens using noninferiority methods with both regimens expected to achieve high cure rates [7]. Typically, in science, observation (the what) is typically followed by experiments to understand the phenomenon (the why) (i.e., they are complimentary). *H. pylori* therapy has generally stopped with the what.

#### **6. Meta-Analysis and** *H. pylori* **Therapy**

Although meta-analysis has become the holy grail for analysis of studies in gastroenterology, it was often used inappropriately for assessing *H. pylori* therapy. The main problem has been that the comparisons involved were often fatally flawed. For example, if one compared 14-day sequential therapy with 14-day clarithromycin triple therapies in treatment of adherent patients with susceptible infections, one would expect both to have very similar high cure rates. Different patterns of resistance would be an obvious important difference (e.g., to clarithromycin and to metronidazole which is only present in one of the therapies), whereas relative potency of the PPI used is not (i.e., 40 mg of pantoprazole = 9 mg of omeprazole, whereas 40 mg of esomeprazole = 64 mg omeprazole [30,31]. If the populations did not differ in relation to resistance patterns or PPI relative potency, one would expect both to yield high cure rates. Failure to do so would signify the presence of important differences between the two populations such that they could not be compared as the results would be nongeneralizable and produce flawed conclusions. When meta-analyses compare trials where the data for each population is population specific and not generalizable, they are best described as Shmeta analyses [4,30]. Generalizability is one of the key requirements for valid and ethical research (Table 2) [32].

**Table 2.** Guidelines to implement antimicrobial stewardship for treatment of *H. pylori* infections.


#### Adapted from [32].

Another example of misuse of meta-analyses has been when it was used to provide guidance regarding therapy. For example, the 2017 American College of Gastroenterology guideline used a meta-analysis to show that bismuth quadruple therapy should replace triple therapy. They provided "an updated meta-analysis, which included 12 randomized controlled trials (RCTs) with 2753 patients; the intention-to-treat (ITT) eradication rate was 77.6% with bismuth quadruple therapy vs. 68.9% with clarithromycin triple therapy" [33]. However, because both regimens achieved clinically unacceptably low cure rates, the appropriate conclusion, based on the results presented, would be that neither should be used as an empiric therapy (at least in the regions where the studies were done).

Although *H. pylori* gastritis is an infectious disease of known cause for which reliably high cure rates are possible and there is no placebo response, the current status is that consensus statements and guidelines have been ineffective, and most patients continue to receive largely poor effective therapy [8,34].

#### **7. The Role of Pharmaceutical Companies in Developing** *H. pylori* **Therapy**

Pharma became involved with *H. pylori* when PPIs were new drugs and peptic ulcer was the major disease for which antisecretory drugs were used. H2RA's were proven effective for treatment of peptic ulcers and omeprazole was having a difficult time becoming accepted. *H. pylori* represented an opportunity as it was a new problem without a simple and effective therapy. PPIs also appeared to possibly have a role to play in therapy and, in order to promote omeprazole, AstraZeneca sponsored a series of consensus conferences regarding *H. pylori* (including Maastricht I) [35,36]. This proved to be a win-win for AstraZeneca, as well as for spreading interest and knowledge regarding *H. pylori*. It also helped solidify *H. pylori* as a gastroenterology disease. Other pharmaceutical companies' subsequent involvement primarily was to promote their anti-*H. pylori* therapies. The most recent example has been designed to promote the bismuth quadruple therapy, Pylera®, in Europe. Pharma's goals included ensuring that their regimens, plus their suggested duration of therapy, were included in lectures, consensus conferences, and guidelines. Most regimens are now off-patent which has reduced, but not eliminated, pharmaceutal company influence on the knowledge and recommendations disseminated.

#### **8. The Role of the U.S. Food and Drug Administration**

The original approvals of *H. pylori* therapies by the U.S. Food and Drug Administration (FDA) were obtained during the period when high cure rates could not be reliably achieved (Figure 1) [37–40]. Although it is a common misconception, FDA approval does not carry with it any implication that the regimen has been optimized in terms of doses or duration of therapy or that it will reliably achieve high

cure rates. For example, the cure rates reported in the studies used to obtain approval of clarithromycin triple therapies ranged from 79% to 86%. The cure rates with clarithromycin triple therapies were the following: 77% with omeprazole (for 10 days), lansoprazole (for 10 and 14 days), rabeprazole (for 7 days); 78% with rabeprazole (for 10 days); and 83% with esomeprazole (for 10 days) [17,41–43] (Figure 1) [37–40]. The pivotal study with pantoprazole was not submitted for FDA approval, likely because the per protocol cure rates were relatively low (70% for the clarithromycin and 76% for the metronidazole seven-day triple therapy) [37]. Of interest, whenever two durations of therapy were examined (e.g., 7 and 10 days or 10 and 14 days), the shorter duration was always selected for the marketed version. cure rates. For example, the cure rates reported in the studies used to obtain approval of clarithromycin triple therapies ranged from 79% to 86%. The cure rates with clarithromycin triple therapies were the following: 77% with omeprazole (for 10 days), lansoprazole (for 10 and 14 days), rabeprazole (for 7 days); 78% with rabeprazole (for 10 days); and 83% with esomeprazole (for 10 days) [17,41–43] (Figure 1) [37–40]. The pivotal study with pantoprazole was not submitted for FDA approval, likely because the per protocol cure rates were relatively low (70% for the clarithromycin and 76% for the metronidazole seven-day triple therapy) [37]. Of interest, whenever two durations of therapy were examined (e.g., 7 and 10 days or 10 and 14 days), the shorter duration was always selected for the marketed version.

*Antibiotics* **2020**, *9*, x FOR PEER REVIEW 6 of 16

regimen has been optimized in terms of doses or duration of therapy or that it will reliably achieve high

**Figure 1.** Intention to treat cure rates and standard deviations reported from the clinical trials published for studies designed to obtain FDA approval for triple therapy with PPI, clarithromycin, and amoxicillin, in the United States. From [4] with permission. **Figure 1.** Intention to treat cure rates and standard deviations reported from the clinical trials published for studies designed to obtain FDA approval for triple therapy with PPI, clarithromycin, and amoxicillin, in the United States. From [4] with permission.

The original bismuth triple therapy (Helidac®) was approved using small studies that had been completed independently of the pharmaceutical company [44]. It was marketed for 14 days and could be used directly or with an H2RA. The more recent version as a three-in-one combination product (Pylera®) was studied and subsequently marketed for 10 days to offer a commercial advantage over the traditional product [45]. No comparisons were, or have been, done to address what is the optimum duration of therapy in the presence of metronidazole resistance (see below). The optimum duration must be defined experimentally rather than by marketing efforts of Pharma. The original bismuth triple therapy (Helidac®) was approved using small studies that had been completed independently of the pharmaceutical company [44]. It was marketed for 14 days and could be used directly or with an H2RA. The more recent version as a three-in-one combination product (Pylera®) was studied and subsequently marketed for 10 days to offer a commercial advantage over the traditional product [45]. No comparisons were, or have been, done to address what is the optimum duration of therapy in the presence of metronidazole resistance (see below). The optimum duration must be defined experimentally rather than by marketing efforts of Pharma.

#### **9. Basis for the General Recommendation for a Treatment Duration of 14 Days**

**9. Basis for the general recommendation for a treatment duration of 14 days**  *H. pylori* is one of those organisms, like *Mycobacterium tuberculosis,* that can enter a dormant state (persister state) in which its metabolism slows as does the need for replication. This process allows the organism to survive despite the presence of antibiotics [46–48]. As noted above, although therapy would appear to have eliminated the infection, early experiments found that, *H. pylor*i was only suppressed and would rapidly reappear. Reappearance either denoted emergence of resistance or, if the organism remained susceptible to the antibiotics used, the duration of therapy was inadequate. The traditional response to recurrence without resistance is to lengthen the duration of therapy. With tuberculosis, this may require many months of therapy, with *H. pylori* 14 days appears to be sufficient. *H. pylori* is one of those organisms, like *Mycobacterium tuberculosis,* that can enter a dormant state (persister state) in which its metabolism slows as does the need for replication. This process allows the organism to survive despite the presence of antibiotics [46–48]. As noted above, although therapy would appear to have eliminated the infection, early experiments found that, *H. pylori* was only suppressed and would rapidly reappear. Reappearance either denoted emergence of resistance or, if the organism remained susceptible to the antibiotics used, the duration of therapy was inadequate. The traditional response to recurrence without resistance is to lengthen the duration of therapy. With tuberculosis, this may require many months of therapy, with *H. pylori* 14 days appears to be sufficient. With *H. pylori,* this phenomenon is most often seen with amoxicillin-containing therapies. Because *H.*

*Antibiotics* **2020**, *9*, x; doi: FOR PEER REVIEW www.mdpi.com/journal/antibiotics

With *H. pylori,* this phenomenon is most often seen with amoxicillin-containing therapies. Because *H.* 

*pylori* only replicates within a narrow pH range (near pH 7), strategies to enhance killing would be to maintain an intragastric pH of greater than six, increasing the duration of therapy, or both [46,49,50].

#### **10. Optimization: Duration of Therapy**

The principles of antimicrobial stewardship require that therapies be optimized to achieve the highest cure rates while taking into consideration safety and cost-effectiveness. Because of the general observation regarding amoxicillin-containing regimens that 14-day therapy is generally superior to shorter durations, it has been recommended that the initial trial should be for 14 days and, only if the 14-day therapy proves highly effective, should one consider testing shorter durations [24]. The effectiveness of PPI plus amoxicillin-containing triple therapies is duration dependent. For example, when these triple therapies are given to patients with susceptible infections, the cure rate is typically between 88 and 92% with a 7-day regimen, 90–94% with a 10-day regimen, and 94–98% with a 14-day regimen [51,52]. PPI, amoxicillin, fluoroquinolone triple therapy is an excellent example of a marked delay in recognizing that it was possible to achieve high cure rates with this regimen. A possible bias by clinical investigators toward shorter durations resulted in a large number of studies and meta-analyses with PPI, amoxicillin, fluoroquinolone triple therapy [53] before it was recognized that cure rates ≥95% were obtainable with 14-day therapy [54].

With many regimens, particularly with amoxicillin-containing triple therapies, one can predict the population cure rate based on the prevalence of resistance or vice versa. For example, since amoxicillin resistance is currently very rare, it can generally be ignored. In contrast, clarithromycin resistance is all-or-none and clarithromycin is functionally removed from the regimen making the cure rate dependent on the remaining amoxicillin PPI dual therapy [55]. The cure rate for any population can be estimated as follows: (cure rate with susceptible infections x the proportion with susceptible infections) + (cure rate with resistant infections x the proportion with clarithromycin resistance). An alternate approach would be to use an *H. pylori* treatment nomogram (Figure 2) [55]. The nomogram has the advantage of allowing one to easily visualize the effect of the relation of prevalence of resistance on cure rates with different durations of therapy.

Although 14-day therapy may prove to be optimal, clinicians may still be obliged to use a shorter, government-approved duration which may be less effective. Clinicians may also be confused by the recommendations from consensus conferences which, until recently, recommended 7-day triple therapy or, in other instances, recommended a range of durations such as 7 to 14 days. Optimal duration can never be expressed as a range. In addition, consensus recommendations often fail to include the caveats needed to understand when any specific duration would be recommended. The reasons for this lack of clarity are unclear. Possibilities include not wishing to appear opposed to what is approved locally, or bias related to one or more conference sponsors.

Bismuth quadruple therapy does not contain amoxicillin and the antimicrobials used are relatively acid insensitive, such that the lessons learned with amoxicillin-containing therapies may not apply. As noted above, early studies showed that, with metronidazole susceptible infections, a duration of 4 to 7 days was sufficient to achieve cure rates of >95% [56–59]. However, consensus conferences have almost exclusively recommended 14-day bismuth quadruple therapy [60–63]. This recommendation was based on the fact that, in many areas, metronidazole resistance is common and increasing, and susceptibility testing is rare. Thus, as a general rule, unless proven otherwise, resistance should be considered to be present which requires one to lengthen the duration of therapy and increase the dosage of metronidazole [13]. In areas where metronidazole resistance is rare or, when metronidazole susceptibility has been confirmed, durations shorter than 14 days are effective and 7-day therapy is generally recommended.

**10. Optimization: duration of therapy** 

that cure rates ≥95% were obtainable with 14-day therapy [54].

*pylori* only replicates within a narrow pH range (near pH 7), strategies to enhance killing would be to maintain an intragastric pH of greater than six, increasing the duration of therapy, or both [46,49,50].

The principles of antimicrobial stewardship require that therapies be optimized to achieve the highest cure rates while taking into consideration safety and cost-effectiveness. Because of the general observation regarding amoxicillin-containing regimens that 14-day therapy is generally superior to shorter durations, it has been recommended that the initial trial should be for 14 days and, only if the 14-day therapy proves highly effective, should one consider testing shorter durations [24]. The effectiveness of PPI plus amoxicillin-containing triple therapies is duration dependent. For example, when these triple therapies are given to patients with susceptible infections, the cure rate is typically between 88 and 92% with a 7-day regimen, 90–94% with a 10-day regimen, and 94–98% with a 14 day regimen [51,52]. PPI, amoxicillin, fluoroquinolone triple therapy is an excellent example of a marked delay in recognizing that it was possible to achieve high cure rates with this regimen. A possible bias by clinical investigators toward shorter durations resulted in a large number of studies and meta-analyses with PPI, amoxicillin, fluoroquinolone triple therapy [53] before it was recognized

With many regimens, particularly with amoxicillin-containing triple therapies, one can predict the population cure rate based on the prevalence of resistance or vice versa. For example, since amoxicillin resistance is currently very rare, it can generally be ignored. In contrast, clarithromycin resistance is all-or-none and clarithromycin is functionally removed from the regimen making the cure rate dependent on the remaining amoxicillin PPI dual therapy [55]. The cure rate for any population can be estimated as follows: (cure rate with susceptible infections x the proportion with susceptible infections) + (cure rate with resistant infections x the proportion with clarithromycin resistance). An alternate approach would be to use an *H. pylori* treatment nomogram (Figure 2) [55].

**Figure 2.** *Helicobacter pylori* treatment nomogram illustrating the duration-related difference in cure rate with 7- and 14-day clarithromycin triple therapy and the effect of clarithromycin resistance on the outcome. As shown, 15% clarithromycin resistance results in a decline in the population cure rate to approximately 85% with 14-day therapy and to approximately 75% with 7-day therapy. It also **Figure 2.** *Helicobacter pylori* treatment nomogram illustrating the duration-related difference in cure rate with 7- and 14-day clarithromycin triple therapy and the effect of clarithromycin resistance on the outcome. As shown, 15% clarithromycin resistance results in a decline in the population cure rate to approximately 85% with 14-day therapy and to approximately 75% with 7-day therapy. It also shows the lack of utility of consensus conference recommendations to use 15% resistance as a yes-no guide to therapy.

*Antibiotics* **2020**, *9*, x; doi: FOR PEER REVIEW www.mdpi.com/journal/antibiotics The approval of Pylera® in Europe resulted in an effort to shorten the recommended duration to 10 days to coincide with the approved and marketed duration. In the USA, the duration issue was not a problem, as Pylera® was offered in bottles of capsules which allowed the physician to prescribe, and the pharmacist to dispense, whatever the physician decided was the best duration of therapy for the individual patient. At the same time that tetracycline became very difficult to obtain, the company changed the packaging of Pylera® from bottles to a 10-day dose pack. This requires purchasing two packs to achieve a 14-day therapy for treatment of patients with metronidazole resistant infections, which represents a problem in the USA because the average retail price of Pylera® is \$1110/10-day dose pack.

The optimum duration of bismuth quadruple therapy remains unknown, largely untested, and is impossible to prove without head-to-head comparisons in populations where the pattern of resistance is known. Another problem related to doing comparisons is that whether a strain is considered resistant or susceptible to metronidazole may depend in part on the test used. It is unclear why Etest may overestimate metronidazole resistance as compared with results obtained with agar dilution [64,65]. Etest is widely used because it is easier to perform. Because of Etest's tendency to overestimate the prevalence of metronidazole resistant infections, it has been our practice to always confirm metronidazole resistant results by agar dilution which correlates better with clinical outcome. Efficacy studies relying on Etest to determine metronidazole resistance are more likely to overestimate the therapies' efficacy in the presence of resistant strains. There are a number of studies ostensibly done to test whether 10-day bismuth quadruple therapy is highly effective. These have been studied in populations with an unknown, but generally low, prevalence of metronidazole resistance [66]. Such studies often achieve cure rates between 88% and 92% which appear suboptimal [8,66,67]. Studies are needed that are designed using the principles of antimicrobial stewardship and all treatment components, including the antisecretory activity of the PPI chosen, antimicrobial dosages, frequency of administration, administration in relation to meals, and the duration of therapy. The question "What

is the optimal duration for bismuth quadruple therapy with susceptible infections, with resistant infections, and for populations where the resistant pattern is unknown?" remains unanswered. It is likely that, for susceptible infections, 10 and 14 days are too long and, for resistant infections, both may be too short.

#### **11. Poly-Antimicrobial Therapies**

The discovery of *H. pylori* and the search for effective treatment coincided with the problem of increasing global antimicrobial resistance. The worldwide increase in macrolide resistance resulted in a precipitous decline in *H. pylori* cure rates with clarithromycin-containing therapies. One response was to modify the current empiric regimens by increasing the number of antimicrobials used (i.e., if two antimicrobials were no longer effective, why not add a third or a fourth?) (reviewed in [30]). This led to the use of a variety of empirically administered regimens containing combinations of a PPI, amoxicillin, clarithromycin, and metronidazole named sequential, concomitant, hybrid, and reverse hybrid therapies. As noted above, 10-day sequential therapy consists of a five-day course of a PPI plus amoxicillin followed by five-day course of a PPI, clarithromycin, and metronidazole. Although the 14-day therapy was more effective, the longer duration has not been widely used, as overall, the regimen has been considered to be obsolete [52].

The alternative was to give all four drugs concomitantly as a concomitant therapy, and proceeded or followed by a dual PPI-amoxicillin therapy (as a hybrid or reverse hybrid therapies). Concomitant therapy is representative of the group. It is functionally equivalent to giving both metronidazole and clarithromycin triple therapies simultaneously [68,69], with success being dependent on the infection being susceptible to amoxicillin and clarithromycin or to metronidazole. Treatment failure requires resistance to both metronidazole and clarithromycin. Although effective, the potential for these three antimicrobial-containing therapies to contribute to the global problem of antimicrobial misuse was not considered. The problem is that all subjects receive at least one antimicrobial not required to cure the infection and whose only function is to potentially contribute to the global antimicrobial resistance (Table 3) [4,69].


**Table 3.** Hypothetical scenario of number of unnecessary antibiotics given in relation to antibiotic susceptibility patterns.

Legend: Table showing the number of ineffective or unnecessary antibiotics used by a population of patients similar to those seen in Texas with the *H. pylori* resistance pattern of 20% resistant to clarithromycin, 40% resistant to metronidazole (8% dual resistance) which receives concomitant therapy with a PPI, amoxicillin, clarithromycin, and metronidazole. From [69] with permission.

The quantity of unnecessary antibiotic misuse with these therapies is not trivial. For example, successful therapy with 14-day concomitant therapy containing 1 gm of metronidazole and clarithromycin would produce 14,000 kg of unneeded antibiotic per 1 million successful treatments and 28,000 kg per 1 million treatment failures. Empiric concomitant therapy was recommended by the Maastricht V, Toronto, and American College of Gastroenterology guidelines [33,61,62], but not by the Houston consensus which considered concomitant therapy to be obsolete [60]. The tendency to add antimicrobials has continued with new sequential therapies and even therapies containing four antimicrobials [30,70]

#### **12. Requirements and Impediments for Transition of** *H. pylori* **Therapy to the Principles of Antimicrobial Stewardship**

Although susceptibility testing for human pathogens is available in most hospitals and clinics worldwide, local susceptibility testing for *H. pylori* is almost universally unavailable. In the United States, culture and susceptibility testing for *H. pylori* is currently available from the Mayo Clinic laboratory and a few other major commercial laboratories. However, the details of how to obtain this service remain the responsibility of the individual physician or endoscopy unit. Molecular susceptibility testing of biopsies or stool specimens is also available commercially from a few sources (e.g., American Molecular Laboratories, http://amlaboratories.com/clinical-lab-menu/amhpr-h-pylori-antibiotic-resistance-panel). Consensus statements have typically recommended susceptibility testing only for patients with at least two treatment failures. The reasons why susceptibility testing is lacking are many and include lack of demand, difficulties with reimbursement, lack of a tradition of susceptibility-based therapy to treat *H. pylori,* and lack of surveillance programs to provide local or regional resistance patterns and to guide therapy. This may change in the USA as the Centers for Medicare and Medicaid Services (CMS) has recently published a regulation requiring all hospitals participating in its programs to establish antimicrobial stewardship programs [71]. They require the appointment of a physician and a pharmacist to be responsible for developing plans and procedures to ensure appropriate therapy. Their requirements also include providing susceptibility testing, treatment guides, as well as monitoring of therapy and prescriptions which are included in the U.S. Centers for Disease Control (CDC) guidelines [72]. These include creation and promotion of susceptibility-based treatments, tracking of antibiotic dispensing, and setting targets for improvement (i.e., monitoring and reporting). It is not yet clear whether *H. pylori* infections are included in the mandate.

The current *H. pylori* treatment guidelines have proven to be ineffective, as they have failed to provide recommendations to reliably yield high cure rates and for creation and promotion of susceptibility-based treatments, tracking of antibiotic dispensing, as well as setting targets for improvement [34]. The prevalent A vs. B comparison mentality has been concerned with differences in (a) actual results, (b) rather than whether either achieved acceptable cure rates, or (c) for understanding the reason for the differences (i.e., the data required to reliably achieve high cure rates). As noted above, most meta-analyses have involved studies whose results are only relevant to the individual study included and are not generalizable or useful for any other population [30].

Consensus conferences and guidelines have often failed to provide clinically useful guidance regarding therapy. For example, as noted above, by 2001 empiric clarithromycin triple therapy was proven to no longer achieve clinically acceptable cure rates in most areas. Rather than state outright that clarithromycin should no longer be used empirically, the 2006 Maastricht III conference suggested using a cutoff of 15–20% resistance above which clarithromycin should not be used empirically [73]. They noted, "Clarithromycin resistance is increasing. It is the main risk factor for treatment failure. Treatment should achieve an eradication rate of >80%. The threshold of clarithromycin resistance at which this antibiotic should not be used, or a clarithromycin susceptibility test should be performed is 15–20%." The cut-off was further refined to >15% in the 2012 Maastricht IV confirmed in the 2017 Maastricht V consensus [61,74]. In retrospect, these guidelines were both impractical and toothless as they were impossible to implement because the required susceptibility data was unavailable. Subsequent analyses of worldwide *H. pylori* resistance have confirmed that resistance exceeded 15% for clarithromycin, metronidazole, and levofloxacin in all WHO regions [75].

The more recent Houston consensus recommended that clarithromycin, metronidazole, and levofloxacin should not be used empirically unless proven to be reliably highly effective locally [60]. Post treatment test-of-cure is currently the only method that most clinicians can use to indirectly assess susceptibility/resistance patterns. With few exceptions, routine testing for cure has been recommended for decades. As a surveillance tool, the results, if heeded, provide information whether a regimen does, or does not, reliably achieve high cure rates. In theory, this information would be collected, shared, and used to indicate whether a regimen should be replaced or modified. Overall, it appears that while this

simple measure may be used to assist in the management of individual patients it has been ineffective as a surveillance tool [8,33]

#### **13. Adoption of the Principles of Antimicrobial Stewardship**

While the goal is straightforward, accomplishing adoption of antimicrobial stewardship will not be easy as it requires a major paradigm shift (Table 4) [32].

**Table 4.** American Society of Infectious Diseases criteria for ethical active-controlled superiority studies of antibacterial agents.


#### Adapted from reference [32].

One of the first steps is to develop treatment guidelines based on the principles of antimicrobial stewardship. All recommendations not based on antimicrobial stewardship should be replaced by ones proven to reliably produce high local cure rates. To date, none of the currently used or recommended treatment regimens have been optimized nor do they consider local resistance patterns. In addition to providing new treatment guidelines, surveillance of treatment outcomes must routinely be monitored to assure continuing effectiveness. Clinical trials should focus on achieving high cure rates (e.g., >95%). "Good enough" is not good enough. Comparative trials should be restricted to comparisons of proven highly effective therapies that utilize noninferiority methodology. Studies using a regimen known to have an inferior cure rate as a comparator without truly informed consent are unethical and should not be done and, if done, should not be published [25,32] ( Table 2; Table 4). No regimen should be used empirically unless it has been proven to reliably achieve high cure rates in the target population. Surveillance programs should be implemented to provide early warning if the effectiveness of currently recommended therapies declines, so that new therapies and guidelines can be implemented. Until susceptibility testing becomes widespread, surveillance should consist of routine tests-of-cure and the results should be reported in order to alert clinicians when a regimen no longer should be prescribed empirically. Current ongoing local and regional antimicrobial surveillance programs should include *H. pylori.* Large effective consortia, such as the European Registry on *Helicobacter pylori* Management, should be repurposed from simply collecting treatment results to providing surveillance, susceptibility testing, and up-to-date treatment recommendations. Clearly, we still have a long way to go and many things to do to adopt antimicrobial stewardship for *H. pylori* therapy.

#### **14. Proposal Regarding How to Improve Empiric Therapies While Introducing Antimicrobial Stewardship**

Fundamentally, the goal is to reliably achieve high cure rates in routine clinical practice (Table 5). The principles are as follows: First, to use only regimens proven to achieve high cure rates locally; second, to provide real-time information about whether the goal is achieved by routinely monitoring and reporting outcomes based on test-of-cure data; third, to abandon or modify therapies that fail to reliably achieve the desired high cure rates.

**Table 5.** Reliable achievement of high cure rates with empiric therapies in clinical practice.


Clinically, cure is defined using a noninvasive test such as a negative urea breath test at least four weeks after ending therapy or a negative stool antigen test preferably at least six weeks post therapy. Defining what is the minimal high cure rate is complicated by the fact that these noninvasive tests are only approximately 95% sensitive and specific. This limitation is overcome in clinical trials by requiring two positive or negative tests using different methodology (e.g., histology and UBT). This is not practical in daily practice and the cure rate that best approximates ≥95% needs to be identified experimentally. As noted above, the sensitivity and specificity of the noninvasive tests currently used to determine cure are at best 95%. Currently, only one test, typically the UBT, is used and it is practically impossible to reliably confirm that ≥95% have been cured. As such, we propose the cut-off of ≥93% until the testing required to define cure allows a more precise estimate. For population-based clinical results, the cure rate should be based on modified intention-to-treat results that include only those who have test-of-cure data. Obtaining a test-of-cure should be vigorously attempted on all patients irrespective of the duration of therapy (e.g., even for one day), as well as those on lost to follow-up for weeks or even months.

All locally or regionally approved therapies should be proven to reliably achieve high cure rates. These recommendations should include the following: (a) antibiotic doses and frequency of administration; (b) PPI minimum dosage should be 60 mg omeprazole equivalent (e.g., 60 mg omeprazole, 60 mg lansoprazole, 40 mg esomeprazole or rabeprazole, or vonoprazan 20 mg); (c) twice daily for 14 days; and (d) the duration should be 14 days unless the regimen has been formally optimized to use a different duration (Table 6).

**Table 6.** Elements of empiric regimens used while the principles of antimicrobial stewardship are being introduced.


Therapies that contain unneeded antibiotics should not be prescribed (e.g., those with three antibiotics, such as concomitant or sequential therapies, or vonoprazan triple therapies) [76]. The test-of-cure result should also be used as part of ongoing surveillance. Ideally, the results should be reported to a central site so that data from an area/region can be pooled and shared.

Regimens that fail to achieve the prespecified endpoint in a prespecified percentage (e.g., 10%) should be removed from the list of approved empiric therapies, although they may remain on the list of approved susceptibility-based therapies.

Although all the approved therapies are expected to achieve high cure rates (i.e., they have been quasi-optimized), each should be formally optimized, as they most likely canfurther improved. Optimization of therapies should be one of the first goals in the introduction of antimicrobial stewardship. As noted above, any modification of an approved therapy should first be confirmed as reliably highly effective using pilot studies without a comparator that also include susceptibility testing. A head-to-head comparison using noninferiority methodology should only be considered after a new or modified regimen has been proven to achieve a high cure rate.

**Funding:** This research received no external funding.

**Conflicts of Interest:** Graham is a consultant for RedHill Biopharma and Phathom Pharmaceuticals regarding novel *H. pylori* therapies and has received research support for culture of *Helicobacter pylori* and is the PI of an international study on the use of anti-mycobacterial therapy for Crohn's disease.

**Support:** Graham is supported in part by the Office of Research and Development Medical Research Service Department of Veterans Affairs, Washington DC and Public Health Service grants DK56338 which funds the Texas Medical Center Digestive Diseases Center. The contents are solely the responsibility of the author and do not necessarily represent the official views of the VA or NIH.

#### **References**


with Clarithromycin, Metronidazole, or Amoxicillin in Patients with Peptic Ulcer Disease: Results of Two Double-Blind, Randomized Studies. *Helicobacter* **2003**, *8*, 626–642. [CrossRef]


© 2020 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Review*

#### **Update on Acute Bone and Joint Infections in Paediatrics: A Narrative Review on the Most Recent Evidence-Based Recommendations and Appropriate Antinfective Therapy**

#### **Giovanni Autore, Luca Bernardi and Susanna Esposito \***

Pediatric Clinic, Pietro Barilla Children's Hospital, Department of Medicine and Surgery, University of Parma, 43126 Parma, Italy; giovanniautore@gmail.com (G.A.); bernardi.luca91@gmail.com (L.B.)

**\*** Correspondence: susanna.esposito@unimi.it; Tel.: +39-0521-903524

Received: 7 July 2020; Accepted: 4 August 2020; Published: 6 August 2020

**Abstract:** Acute bone and joint infections (BJIs) in children may clinically occur as osteomyelitis (OM) or septic arthritis (SA). In clinical practice, one-third of cases present a combination of both conditions. BJIs are usually caused by the haematogenous dissemination of septic emboli carried to the terminal blood vessels of bone and joints from distant infectious processes during transient bacteraemia. Early diagnosis is the cornerstone for the successful management of BJI, but it is still a challenge for paediatricians, particularly due to its nonspecific clinical presentation and to the poor specificity of the laboratory and imaging first-line tests that are available in emergency departments. Moreover, microbiological diagnosis is often difficult to achieve with common blood cultures, and further investigations require invasive procedures. The aim of this narrative review is to provide the most recent evidence-based recommendations on appropriate antinfective therapy in BJI in children. We conducted a review of recent literature by examining the MEDLINE (Medical Literature Analysis and Retrieval System Online) database using the search engines PubMed and Google Scholar. The keywords used were "osteomyelitis", OR "bone infection", OR "septic arthritis", AND "p(a)ediatric" OR "children". When BJI diagnosis is clinically suspected or radiologically confirmed, empiric antibiotic therapy should be started as soon as possible. The choice of empiric antimicrobial therapy is based on the most likely causative pathogens according to patient age, immunisation status, underlying disease, and other clinical and epidemiological considerations, including the local prevalence of virulent pathogens, antibiotic bioavailability and bone penetration. Empiric antibiotic treatment consists of a short intravenous cycle based on anti-staphylococcal penicillin or a cephalosporin in children aged over 3 months with the addition of gentamicin in infants aged under 3 months. An oral regimen may be an option depending on the bioavailability of antibiotic chosen and clinical and laboratory data. Strict clinical and laboratory follow-up should be scheduled for the following 3–5 weeks. Further studies on the optimal therapeutic approach are needed in order to understand the best first-line regimen, the utility of biomarkers for the definition of therapy duration and treatment of complications.

**Keywords:** antibiotic; bone infection; joint infection; osteomyelitis; pediatric infectious disease; septic arthritis

#### **1. Introduction**

Acute bone and joint infections (BJIs) in children may clinically occur as osteomyelitis (OM) or septic arthritis (SA). BJIs generally present clinically within 2 weeks of disease onset [1]. In clinical practice, one-third of cases present a combination of both conditions, and this combination may

occur in as many as 75% of cases in newborns [2]. BJIs are usually caused by the haematogenous dissemination of septic emboli carried to the terminal blood vessels of bone and joints from distant infectious processes during transient bacteraemia. Less common infection routes are direct inoculation due to open fractures or invasive procedures and extension from contiguous infections, such as cellulitis and sinusitis. BJI can be classified as acute, subacute and chronic according to its duration: <2 weeks, <3 months and >3 months from onset, respectively. Chronic infections are relatively rare conditions in paediatric patients, could be caused by the establishment of biofilm, and different surgical approaches must be considered [3].

The mean annual incidence of BJI in high-income countries is approximately 8 per 100,000 children [4,5]. Despite the high variability between different reports, an increasing trend has been observed over the last few decades, probably due to increased diagnostic effectiveness. Gafur et al. observed that the annualised per capita incidence of OM increased 2.8-fold in the same paediatric hospital within two decades [4]. Children aged ≤ 5 years showed a higher prevalence, accounting for half of all cases [6]. Although uncommon, BJI in children should not be underestimated because local and systemic complications may result in life-threatening conditions and severe disabilities. If not promptly diagnosed and treated, the infection may extend to soft tissues, causing pyomyositis (especially in young infants) and sepsis [7]. Local progression may result in subperiosteal or intraosseous abscesses, pathological fractures, and abnormal bone growth due to the involvement of the epiphysis [8–10]. Venous thrombosis and septic embolism may also occur, more commonly in children aged >8 years [11,12].

Early diagnosis is the cornerstone for the successful management of BJI, but it is still a challenge for paediatricians, particularly due to its nonspecific clinical presentation and to the poor specificity of the laboratory and imaging first-line tests that are available in emergency departments. Moreover, microbiological diagnosis is often difficult to achieve with common blood cultures, and further investigations require invasive procedures. The indications and effectiveness of available diagnostic tools are still debated. In addition, the resistance pattern of aetiological agents and poor bone penetration of antibiotics represent a challenge for an appropriate therapeutic approach. Therefore, the aim of this narrative review is to provide the most recent evidence-based recommendations on appropriate antinfective therapy in BJI in children.

#### **2. Aetiology and Pathogenesis**

BJI commonly occurs in primarily healthy children without clear predisposing conditions. However, a higher prevalence is described in patients affected by immunodeficiencies and haemoglobinopathies, particularly sickle cell disease (SCD) and chronic granulomatous disease (CGD) [13,14]. Previous experimental studies on animal models have suggested that minor trauma may increase the susceptibility of bone and joint tissues to bacterial seeding [15]. More recent clinical studies question this association because children affected by BJIs show the same rate of previous minor trauma observed in the general paediatric population [16].

More than 80% of the cases of OM occur in the metaphysis of long tubular bones, and the most common localisations are the femur and tibia, followed by the pelvis and calcaneus among nontubular bones [17,18]. A nationwide survey conducted in the USA revealed that osteomyelitis in the pelvis, upper arm, hand and forearm was associated with a higher risk of septic arthritis and bacteraemia or septicaemia [19]. The knee and hip are the most commonly involved joints, accounting for more than half of SA cases [18]. In a large multicentre study recently conducted in Spain, the involvement of the hip in children affected by combined osteomyelitis and septic arthritis was the main negative prognostic factor associated with a higher risk of complications and sequelae [20].

The choice of proper empiric treatment is the main issue in the management of paediatric BJI. First-line antibiotics should cover the most likely aetiologies according to the patient's age and local prevalence of community-acquired and nosocomial pathogens. In addition, they should have an appropriate bone penetration. According to the largest studies in recent literature, methicillin-sensitive

*Staphylococcus aureus* (MSSA) may still be considered the most common pathogen in Europe, with a prevalence ranging from 30% to 63% of confirmed cases [5]. However, the emerging role of*Kingella kingae* has been confirmed by several studies reporting its isolation in up to 53% of all cases [5]. In addition, *Streptococcus pneumoniae* and Group A β-haemolytic *Streptococcus pyogenes* (GABHS) maintain a relevant role in patients aged ≥ 6 months. On the other hand, the prevalence of *Haemophilus influenzae* type b has drastically decreased in the post-vaccination era, accounting for less than 1% of cases [21].

The aetiology, however, largely depends on the patient's age (Table 1). A recent multicentre study conducted in France on 71 patients aged under 3 months reported that *Streptococcus agalactiae* is the main community-acquired pathogen accounting for 45% of these cases, followed by *S. aureus* (22% of all cases) which was the most frequent microorganism in infants aged over 2 months, and *Escherichia coli* (18%) [22]. These findings were confirmed by Juchler et al. in infants aged 0–6 months, reporting *S. aureus*, *S. agalactiae*, and *E. coli* as the most frequently isolated pathogens, accounting for 31%, 15%, and 8% of these cases, respectively [23]. In this age group, both studies observed that *Klebsiella pneumoniae* and *Candida albicans* are particularly frequent nosocomial pathogens, and they each account for 7% of cases [22]. Among children aged between 6 months and 5 years, *K. kingae* is the most frequent cause of osteoarticular infections, accounting for half of these cases [23]. Less frequent pathogens in this age group are *S. pneumoniae* and GABHS, which were isolated in 29% and 7% of cases, respectively [23]. In the post-vaccination era, *S. pneumoniae* still represents a causative agent of septic arthritis, accounting for over 4% of all cases [4,17]. The detection of *K. kingae* requires specific PCR assays. *K. kingae* seems to be very specific for the age group under 5 years. In fact, a large review of 566 osteoarticular infections caused by this pathogen reported that 80% of these cases occurred in children aged under 4 years [24]. Moreover, Ferroni et al. observed a higher prevalence of *K. kingae* among children affected by SA than OM [25]. A prospective case-control study revealed that oropharyngeal carriage of *K. kingae* is strongly associated with haematogenous BJI in children aged under 4 years. Using a specific polymerase chain reaction (PCR) assay, the authors identified the microorganism in the oropharynx of 71% of previously confirmed cases of osteoarticular infections and only in 6% of age-matched healthy controls, reporting an odds ratio of approximately 38.3 (95% CI, 18.5–79.1) [26].


**Table 1.** Most common pathogens causing bone and joint infection in children and recommended first-line intravenous (IV) empiric treatment of different age groups.

MRSA, methicillin-resistant *S. aureus*.

BJIs in children aged over 5 years are most frequently caused by *S. aureus*, which accounts for up to 61% of these cases [23]. *K. kingae* is isolated in less than 13% of cases in this age group, followed by GABHS [20,23]. Once considered mainly nosocomial, the prevalence of particularly virulent strains of community-acquired *S. aureus* (CA-SA) is increasing, causing severe forms of BJI. Approximately 70–90% of confirmed cases caused by CA-SA involve methicillin-sensitive strains (MSSA), but there has been an increase in cases of BJI from community-acquired methicillin-resistant

*S. aureus* (CA-MRSA) [27,28]. Studying the regional prevalence of CA-MRSA is mandatory because, according to the recent guidelines published by the European Society for Paediatric Infectious Diseases (ESPID), a local prevalence over 10% should induce clinicians to choose a different empiric treatment from conventional first-line drugs. Another virulence factor causing severe forms of staphylococcal infections with issues of management is Panton–Valentine leukocidin (PVL). PVL is a bicomponent, pore-forming toxin produced by some strains of *S. aureus* (PVL-SA) that kills leukocytes. A multicentre European study conducted in 7 countries reported that PVL-SA reached an incidence up to 18%, and PVL was produced more frequently by MSSA strains in Europe, whereas it was more common among MRSA strains in the United States [29]. In the same study, MRSA isolates represented 6% of all cases caused by CA-SA [29]. Nationwide studies conducted in Spain and the UK observed rates of CA-MRSA ranging from 2.5% to 3% of all community-acquired staphylococcal infections, while PVL-SA has been isolated in 10% of these cases in Italy [20,30]. CA-MRSA seems to be less prevalent in European populations than in the USA, where its prevalence reached 30% in some regions within the last two decades [31].

#### **3. Clinical Presentation**

Onset symptoms of BJI are usually nonspecific in children. Children with osteomyelitis usually present acutely with fever and constitutional symptoms, such as irritability and decreased activity. Once the infection progresses, focal symptoms and signs of bone inflammation may occur. According to a large systematic review conducted over a population of 12,000 children with BJI, the most common onset features included localised pain (81%); focal warmth, swelling and point tenderness (70%), fever (62%) and limitation of function (50%) [17]. Clinical suspicion may be difficult in newborns and toddlers because they often lack focal findings and may continue to feed well. Although osteomyelitis is rare in this age group, it should be considered for the differential diagnosis in infants with skin infections, urinary tract anomalies, prematurity, and neonatal sepsis. Mediamolle et al. reported that 94% of infants aged under 3 months showed pain, and 87% of them had functional limitations, but only 52% were febrile [22]. BJI should also be suspected in older children presenting with fever without a source, bacteraemia, and abnormal radiological findings in the evaluation of trauma. A recent history of infections is uncommon, and this makes it difficult to identify the initial source of haematogenous dissemination. Ferroni et al. were able to tentatively identify the portal of entry in only 55% of cases, of which 55% were upper respiratory tract infections; 15%, skin trauma; 11%, gastro-enteritis; 8%, varicella; and 2%, congenital infections [25].

Special populations of children at higher risk of BJI may present with atypical features. Patients with haemoglobinopathy (i.e., sickle cell disease) may present multifocal infections that are difficult to distinguish from vaso-occlusive crisis [13]. Children with chronic granulomatous disease may have BJIs caused by uncommon bacterial or fungal pathogens that usually remain asymptomatic even in advanced stages [14].

When OM is suspected, the differential diagnosis should include traumatic fracture, cellulitis or pyomyositis, rheumatic fever, thrombophlebitis, leukaemia, tumours, sickle cell infarction, tuberculosis, scurvy, and other bone inflammatory processes such as hypophosphatasia and chronic recurrent multifocal osteomyelitis (CRMO). Sepsis should also be ruled out in neonates. Moreover, some clinical signs of SA may also occur in cases of transient synovitis, viral arthritis, reactive arthritis, juvenile idiopathic arthritis, Henoch–Schoenlein purpura, and Perthes disease.

#### **4. Diagnosis**

#### *4.1. Blood Examination*

Initial blood tests for children with suspected BJI include the complete blood count (CBC), erythrocyte sedimentation rate (ESR), and level of C-reactive protein (CRP) [32]. Elevated ESR and CRP show a high sensitivity at disease onset but a low specificity. Dartnell et al. observed that CRP

was elevated in 81% of patients at the time of presentation, with a peak on the second day from onset; instead, ESR peaked 3–5 days after onset [17]. The same study showed that the WBC count was elevated in only 36% of children at the time of diagnosis [17]. However, the CBC is helpful in the differential diagnosis of children with bone pain (e.g., leukaemia). A prospective study of 265 children with osteomyelitis and septic arthritis reported a mean CRP value of approximately 87 mg/L and a mean ESR value of approximately 51 mm/h at the time of presentation [33]. The combination of elevated CRP and ESR showed the best sensitivity to suspect BJI in children [33]. Several studies observed that CRP and ESR are higher and remain abnormal for a longer period in patients with MRSA infection [34]. In addition, MRSA is associated with greater elevations in CRP, ESR and WBC levels [18]. At this time, the role of procalcitonin is unclear, and its effectiveness compared to CRP is debated [32]. Blood culture should always be obtained at the same time as initial blood tests, as well as in afebrile patients, if the clinical suspicion of BJI is well-founded [32].

#### *4.2. Imaging*

The initial imaging study should be the radiograph of the suspected area in order to exclude other causes of pain [32]. However, radiographs are usually normal at the beginning of BJIs, and other advanced techniques are often required. Exceptions may be presented by newborns, who more commonly show abnormal radiographs at the onset [35].

The main X-ray features that suggest a BJI are periosteal reaction, periosteal elevation (suggesting a periosteal abscess), lytic lesions or sclerosis, and narrowing of the intervertebral disc space. At the onset, these alterations are often undetectable, and the timing of radiographic changes depends on the involved bones and the age of the patient. In long bones, cortical thickening and periosteal reaction/elevation are shown only 10 to 21 days after the onset of symptoms (7 to 10 days in newborns) [36]. Lytic sclerosis usually occurs only after more than a month. Due to the indolent course of the discitis that delays the onset of symptoms, X-ray signs of this category of BJI are often evident at the time of presentation [37].

Indications for additional imaging studies are confirmation of the diagnosis in clinically suspected BJI with normal radiographs, further evaluation of detected lesion and its extension (i.e., the involvement of epiphysis and adjacent soft tissues), surgical planning, and guidance for percutaneous procedures.

Magnetic resonance imaging (MRI) can be considered the gold standard imaging method for the diagnosis of BJI and to evaluate the involvement of surrounding soft tissues or joints [32]. The sensitivity and specificity of this technique range from 80% to 100% and from 50% to 100%, respectively [17]. The variability observed between the reported specificity rates may also depend on radiologist-specific experience. The use of intravenous gadolinium is not routinely required, but it is useful to detect intramedullary or muscular abscesses or necrosis, although it should be avoided in patients with renal insufficiency due to the risk of nephrogenic systemic fibrosis [38,39].

MRI can highlight multiple alterations that suggest the diagnosis of BJI. Areas of active inflammation show a decreased signal in T1-weighted images and an increased signal in T2-weighted images [40]. Fat-suppression sequences, including short-tau inversion recovery (STIR), decrease the signal from fat and are more sensitive for the detection of bone marrow oedema. In discitis, MRI easily detects the reduction of disc space, the increased T2-weighted signal in the adjacent vertebral endplates, and bone oedema in the vertebral body [38].

MRI is superior to other imaging methods, particularly to identify early infections affecting the bone marrow before the involvement of the cortical bone, to detect pelvic OM and discitis that are usually undetected by X-rays, to evaluate the involvement of the growth plates, joint structures, and soft tissues (e.g., pyomyositis, muscular abscesses) and to exclude deep venous thrombosis associated with BJI [41–43]. Furthermore, MRI is usually required in presurgical planning and surgical follow-up when drainage is indicated. MRI is also preferred because it prevents children from exposure to ionising radiation.

The main disadvantages of MRI are the longer scan time than CT and the need for sedation in young children. Furthermore, MRI is not always easily available, and it is more expensive. Because of its high sensitivity, the diagnosis of BJI is unlikely if the MRI is negative [44]. False-positive results can occur in patients with primary infections in adjacent soft tissues.

Regarding other imaging methods, computerised tomography (CT) scans are not generally recommended because they are less sensitive than MRI and expose children to ionising radiation. It should be considered in diagnosis only when MRI is not feasible [32]. However, CT scans may play a role when there is important bone destruction on radiographs to assess the extent of bone damage for a surgical approach [41]. It should also be useful in chronic OM when inflammation is too weak to be detected by MRI. In these cases, a CT scan can be performed without sedation and takes less time than MRI. BJI can be detected by the evidence of increased bone marrow density, new periosteal bone formation with periosteal purulence and irregular erosion of bone surfaces.

Bone scintigraphy is used to identify multifocal OM or when localised signs of bone involvement are too poor. It may be more accessible than MRI, and sedation is required less frequently. Technetium radionuclide scanning (99mTc) has high sensitivity but lower specificity compared to MRI; furthermore, scintigraphy has proven to be scarcely sensitive (53%) for OM caused by MRSA [44,45]. 99mTc scanning is triphasic, consisting of the flow phase (2 to 5 s after injection), blood pool phase (5 to 10 min after injection), and delayed phase (2 to 4 h after injection). BJI causes focal absorption in the third phase, and signal intensity is related to the level of osteoblastic activity. Localisation of a lesion near a growth plate can complicate the interpretation. Using 99mTc-methylene diphosphonate (99mTc-MDP), early evidence of infection can be detected 24 h after onset. Hsu et al. observed that specificity may increase with Gallium scanning and In-labelled leukocytes, although these techniques are more complex and add higher radiation exposure [46]. The disadvantages of scintigraphy are the lack of information on the size of pus collections that could be drained (i.e., in cases of intramedullary abscess), exposure to ionising radiation, and false-negative results that may occur if the blood flow to the periosteum is interrupted (i.e., in cases of subperiosteal abscess) [44].

Ultrasonography (US) is not useful for the diagnosis of BJI. With US, it is possible to identify the fluid collections in soft tissues associated with bone infections, and US can be a support for percutaneous diagnostic and therapeutic drainage [47]. Table 2 summarises indications and features of imaging methods in bone and joint infection in pediatric age.


**Table 2.** Indications and features of imaging methods in bone and joint infection in pediatric age.

MRI, magnetic resonance imaging; CT, computed tomography; US, ultrasonography.

#### *4.3. Microbiological Diagnosis*

With the increasing prevalence of antibiotic-resistant microorganisms and emerging pathogens, it is important to collect as many microbiological specimens as possible, and different microbiological tests are often required (Table 3). A microbiological diagnosis is achieved in barely more than half of all cases. In their systematic review, Dartnell et al. reported that microbiological diagnosis is achieved in approximately 50% of all cases of BJI [17]. In the context of a paediatric emergency department, Akinkugbe et al. reported microbiological isolation in only 38% of their cases [30]. On the other hand, a significantly higher success rate, approximately 70%, was reported in a population of patients aged under 3 months [22]. Calvo et al. also observed that the rate of microbiological diagnosis reached 61% in the combined form of OM and SA [20].


#### **Table 3.** Indications for microbiological tests.

Blood culture has the lowest sensitivity, but it is also the most accessible technique. Reported rates of positive blood cultures are highly variable in the literature. McNeil et al. estimated a general sensitivity of approximately 46%, and other studies observed even lower rates for SA [20,48]. Juchler et al. observed that sensitivity may be increased by performing PCR assays on negative cultures; in this way, the authors reported an increase of +4.5% [23]. However, according to the ESPID guidelines, blood culture should always be analysed in cases where there is clinical suspicion (including afebrile patients); the culture should be performed at the same time as the initial laboratory evaluations and should be repeated at fever peaks [32].

When joint involvement is clinically suspected, synovial fluid can be obtained for microbiological analysis. For easily accessible joints (e.g., knee), arthrocentesis may be performed under conscious sedation in the emergency room. Less accessible joints, such as the ankle and hip, need an interventional radiologist. The sensitivity of synovial fluid culture is higher than that observed for blood culture and can be higher than 50% [48]. Arthrocentesis should always be performed when the involvement of accessible joints is clinically evident, and it should also be considered in complicated or nonresponsive cases affecting proximal joints. In order to increase the diagnostic yield of joint aspirate, synovial fluid can be inoculated in blood culture vials.

Bone samples can be obtained with a minimally invasive percutaneous needle biopsy, especially when subperiosteal abscesses occur, or with surgical biopsy in the operating room. The reported sensitivity rates for these invasive techniques may reach 82%, but they expose patients to higher risks [48]. Bone biopsy should be performed in complicated cases with negative blood culture that do not respond to empiric treatment, but it may also be considered when a bone abscess is easily accessible [32]. In fact, for most uncomplicated BJIs, invasive biopsy does not affect the clinical outcomes [49,50]. Nevertheless, surgical biopsy should be performed in every patient who undergoes surgical treatment. In their large prospective study, Ferroni et al. performed arthrocentesis for every SA and bone biopsy only when a subperiosteal abscess occurred; in this way, the authors reported high rates of microbiological isolation, reaching 40% for synovial fluid and 87% for bone samples [25].

Nucleic acid amplification methods, such as conventional and real-time polymerase chain reaction (PCR), improved the detection of pathogens even after the administration of antibiotics. Synovial fluid PCR may remain diagnostic up to 6 days after the first antibiotic dose, and similar results have been observed for bone samples [51]. Specific PCR analysis may also be the only way to identify *K. kingae* and its isolation can be enhancement by inoculation of samples in blood culture vials [51,52].

#### **5. Antinfective Treatment**

Empiric antinfective treatment should be started as soon as BJI is clinically suspected. The choice of empiric antimicrobial therapy is based on the most likely causative pathogens according to patient age, immunisation status, underlying disease, and other clinical and epidemiological considerations, including the local prevalence of MRSA. In addition, antibiotic bioavailability and bone penetration should be considered [53]. Then, management is guided by the results of the antibiograms obtained from the microbiological investigations performed before starting antimicrobial therapy [36].

In neonates younger than two months, empirical treatment should be oxacillin or cefazolin and gentamicin to cover *S. agalactiae* and other gram-negative organisms that are common causes of BJI in this age group [22,54,55]. In children aged 3 months and over, anti-staphylococcal penicillin or a cephalosporin such as cefazolin or cefuroxime should be used to target MSSA, *S. pneumoniae*, GABHS and *K. kingae* [32]. Among the anti-staphylococcal penicillins, the use of flucloxacillin should be preferred because it is well tolerated and shows high bone penetration, even if it is difficult to use for the type of formulation.

In areas with a local prevalence of MRSA higher than 10%, the administration of empirical therapies active against these pathogens is indicated [32,49]. In these cases, the first-choice drugs are clindamycin, vancomycin or linezolid [32,56]. Peltola et al. suggested the empirical use of clindamycin in areas where the prevalence of MRSA is over 10%, and the clindamycin resistance rate is under 10% or vancomycin if the prevalence of MRSA is over 10% and the clindamycin resistance rate is over 10%, with linezolid as the second-line choice [49]. Dalbavancin, even as a single dose, appeared effective in children with BJI due to MRSA [57,58]. Compared to other available antibiotics that are active against MRSA, the advantages of dalbavancin include a lower potential for drug interactions and the possibility of fewer required doses due to a longer half-life [59]. Another second-line drug after the failure of previous antibiotics may be daptomycin [60,61]. In complicated severe cases, when the involvement of PVL SA is suspected, antibiotic therapy should aim to inhibit toxin production. In these cases, inhibitors of protein synthesis, such as clindamycin, linezolid, and rifampicin, are the first choice [62,63]. Among less common pathogens, *Salmonella* spp. is a frequent cause of BJI in developing countries and in patients with sickle cell anaemia, and it should be treated with a third-generation cephalosporin or fluoroquinolone [5]. Candida spp. is mainly isolated in spondylodiscitis and requires prolonged antifungal treatment and surgical debridement [5].

The total duration of antibiotic treatment is widely debated in the literature. Classically, BJIs are treated with long courses of intravenous therapy and prolonged hospitalisation, with OM usually treated for 3–6 weeks and SA for 2–4 weeks. Peltola et al. have shown that even 10 days of treatment is sufficient for SA [64]. Moreover, a recent paper from France has shown that 15 days of treatment is sufficient in most of the cases [65]. Another prospective French study on 70 cases reported no failures of treatment with an intravenous regimen prolonged up to 8 days [25]. A retrospective study was conducted in Spain on 607 children with a mean duration of intravenous therapy of 12.9 days and reported good outcomes [20]. A multicentre randomised trial was conducted in Finland on 252 children randomly assigned to two therapeutic groups. The treatment involved a common short cycle of 2–4 days of intravenous antibiotics for both groups, followed by oral therapy with clindamycin or a high-dose first-generation cephalosporin for 20 days in the first group or 30 days in the second group. The authors observed no significant differences between the two groups, suggesting the effectiveness of shorter treatment regimens [66]. However, a limitation of this study was the absence of cases due to MRSA or PVL-SA. When spondylodiscitis occurs, it is still recommended to carry out intravenous therapy for 1–3 weeks [66]. A similar observational study conducted in the United States showed excellent outcomes with an early transition to oral antibiotics within 4 days; the researchers reported no significant difference in the treatment failure rate compared to that with longer intravenous

regimens [67]. Only in case of patients with infection due to PVL-SA prolonged antimicrobial treatment and multiple surgical procedures are recommended since these infections are often complicated with abscesses and venous thrombosis [64].

The timing for switching from an intravenous to an oral regimen is still debated. Clinical criteria are apyrexia, compliance with oral therapy, pain reduction, and both general and local clinical improvement. Clinical conditions should be in accordance with the reduction in inflammatory markers such as CRP, ESR, and WBC count. Different cut-offs have been proposed for the evaluation of laboratory markers. Some authors prefer to wait until the complete normalisation of CRP before switching the antibiotic regimen [68]. Faust et al. considered acceptable a CRP value under 20 mg/L or at least a decrease of 2/3 of its maximum peak [36]. The ESPID guidelines recommend switching to oral therapy only when the patient presents an improvement in clinical conditions without fever for at least 24 h and a decrease of 30–50% from the CRP maximum peak is observed. However, the guidelines suggested prolonging the intravenous regimen if drug-resistant or more virulent pathogens are isolated [35].

In most observational studies and randomised clinical trials, oral therapy consists of high-dose cephalosporin or clindamycin [59,69–71]. Trials conducted by Peltola et al. showed a failure rate under 1% at follow-up [64]. Trimethoprim/sulfamethoxazole (TMP/SMX) has been successfully used in oral treatment of BJI in children [72–74]. The duration of oral therapy in uncomplicated BJIs is frequently approximately 3–4 weeks with rigorous monitoring of inflammatory markers and drug tolerability [32]. In this way, therapy can be continued at home, allowing patient discharge and subsequent outpatient follow-up.

#### **6. Conclusions**

The clinical presentation of BJI in children may be nonspecific and paucisymptomatic, especially in newborns and immunocompromised patients. Indirect functional signs of OM or SA should be carefully evaluated. All children with negative or inconclusive initial radiographic examination should undergo further highly sensitive imaging studies, such as MRI or bone scintigraphy. MRI is the gold-standard imaging method. It should always be performed in a diagnostic dilemma when the initial radiograph is negative. Contrast enhancement is not routinely required.

Empiric antibiotic therapy should be started as soon as possible. The choice of empiric antimicrobial therapy is based on the most likely causative pathogens according to patient age, immunisation status, underlying disease, and other clinical and epidemiological considerations, including the local prevalence of virulent pathogens, antibiotic bioavailability and bone penetration.

Despite the high success rate reported with empirical therapies, aetiological diagnosis is highly recommended. Blood culture should be obtained in every patient (even if he/she is afebrile) at the initial evaluation and repeated at the fever peak. Synovial fluid samples should also be obtained in the case of SA if antibiotics have already been administered. Bioptic samples are not routinely required in uncomplicated BJI. Instead, minimally invasive percutaneous bone biopsy and surgical biopsy should be considered in complicated infections and when surgery is indicated.

Multidisciplinary management is necessary to achieve an early diagnosis. Paediatricians should consult an experienced radiologist (or an interventional radiologist if percutaneous procedures are indicated) and an orthopaedic surgeon. Microbiological laboratories should also be directly consulted if the involvement of pathogens that are difficult to isolate is suspected (e.g., *K. kingae*).

Empiric antibiotic treatment consists of a short intravenous cycle based on anti-staphylococcal penicillin or a cephalosporin in children aged over 3 months with the addition of gentamicin in infants aged under 3 months. An oral regimen may be an option depending on the bioavailability of the antibiotic chosen and clinical and laboratory data. Further studies on the optimal therapeutic approach are needed in order to understand the best first-line regimen, the utility of biomarkers for the definition of therapy duration and treatment of complications.

#### **7. Methods**

We conducted a review of recent literature by examining the MEDLINE (Medical Literature Analysis and Retrieval System Online) database using the search engines PubMed and Google Scholar. The keywords used were "osteomyelitis", OR "bone infection", OR "septic arthritis", AND "p(a)ediatric" OR "children". We included clinical trials, observational studies, reviews and meta-analyses on acute haematogenous osteomyelitis and septic arthritis in children. The exclusion criteria were patients older than 18 years, non-acute or non-haematogenous infections, case series with fewer than 20 patients, articles published before 2005, and non-English language articles.

**Author Contributions:** G.A. and L.B. co-wrote the manuscript; S.E. supervised the project and made substantial scientific contributions. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

#### **Opportunities and Challenges for Improving Anti-Microbial Stewardship in Low- and Middle-Income Countries; Lessons Learnt from the Maternal Sepsis Intervention in Western Uganda**

**Louise Ackers 1,\* , Gavin Ackers-Johnson <sup>1</sup> , Maaike Seekles <sup>1</sup> , Joe Odur <sup>2</sup> and Samuel Opio <sup>3</sup>**


Received: 29 April 2020; Accepted: 4 June 2020; Published: 9 June 2020

**Abstract:** This paper presents findings from an action-research intervention designed to identify ways of improving antimicrobial stewardship in a Ugandan Regional Referral Hospital. Building on an existing health partnership and extensive action-research on maternal health, it focused on maternal sepsis. Sepsis is one of the main causes of maternal mortality in Uganda and surgical site infection, a major contributing factor. Post-natal wards also consume the largest volume of antibiotics. The findings from the Maternal Sepsis Intervention demonstrate the potential for remarkable changes in health worker behaviour through multi-disciplinary engagement. Nurses and midwives create the connective tissue linking pharmacy, laboratory scientists and junior doctors to support an evidence-based response to prescribing. These multi-disciplinary 'huddles' form a necessary, but insufficient, grounding for active clinical pharmacy. The impact on antimicrobial stewardship and maternal mortality and morbidity is ultimately limited by very poor and inconsistent access to antibiotics and supplies. Insufficient and predictable stock-outs undermine behaviour change frustrating health workers' ability to exercise their knowledge and skill for the benefit of their patients. This escalates healthcare costs and contributes to anti-microbial resistance.

**Keywords:** antimicrobial stewardship; pharmacy; sepsis; wound management; culture and sensitivity testing; resistance patterns; low-and middle-income countries; Uganda

#### **1. Introduction**

A recent review of research on antibiotic stewardship [1] found limited evidence of effective and feasible stewardship interventions in low- and middle-income countries (LMICs) and, where examples of effective interventions were identified, emphasised the essential need for contextualised. This paper reports on a recent, highly contextualized, facility-level intervention in a Regional Referral Hospital (RRH) in Uganda, known as the Maternal Sepsis Intervention (MSI). Funding for this action-research intervention came from the Commonwealth Partnerships for Antimicrobial Stewardship (CwPAMS) [2]. The funding body stipulated a focus on antimicrobial use (stewardship) and a project completion within 15 months with a budget of £60,000. The intervention was necessarily aligned with the Ugandan National Action Plan on Anti-Microbial Resistance or 'NAP' [3]. The NAP was launched in 2018 in an attempt to 'slow down and contain' [3] (p. 3) anti-microbial resistance (AMR). It sets out five Strategic Objectives focused on Awareness-Raising; Infection Prevention; Optimal Access and Use of Antimicrobials; Surveillance and Research.

The CwPAMS objectives resonate most directly with Strategic Objective 3 of the NAP with a primary emphasis on Antimicrobial Stewardship (AMS). The NAP describes the use of antimicrobial agents as 'the major modifiable driver of AMR'. According to the Plan, achieving optimal antimicrobial use 'will require strengthening technical and regulatory frameworks, ensuring availability of appropriate medicines and changing behaviour amongst prescribers, dispensers and consumers' [3] (p. 14).

This articulation of the funding body's objectives with the NAP on AMR framed the design of the MSI. Building on strong pre-existing relationships especially in the field of maternal and new-born health, the project partners decided to focus the intervention on the Post-Natal and Gynaecology (PNG) ward in a RRH in Western Uganda. Fort Portal Regional Referral Hospital (FPRRH) has the second highest maternal mortality rate in the country. The most recent Ministry of Health report indicates a maternal mortality rate of 632/100,000 almost double the reported national average [4]. Sepsis competes with haemorrhage as the leading causes of maternal mortality [4–6] and Reinhart et al. [7] describe sepsis as a 'Global Health Priority'.

Surgical site infection (SSI) following caesarean section contributes significantly to maternal mortality and morbidity [8] and to antimicrobial consumption. As a component of hospital acquired infection, it is also largely preventable. The decision to focus on the PNG ward reflected the opportunity to assess the potential for preventive intervention through improved infection prevention control (IPC) to reduce post-caesarean section SSIs. This focus also enabled us to address stewardship practices on a ward associated with the highest levels of antibiotic consumption in the hospital.

The intervention built on the long-established Kabarole Health Partnership which involves a UK and Ugandan registered NGO (Knowledge For Change (K4C)) as the key operational partner together with the University of Salford; Kabarole Health District, FPRRH and the Pharmaceutical society of Uganda. The Health Partnership model has been actively developed through the Tropical Health and Education Trust (THET) as a more democratic and grounded approach to foreign engagement in global health.

Substantial pre-existing research conducted in partnership with Knowledge For Change (K4C) has established the principle of co-presence to the achievement of effective knowledge mobilisation and behaviour change in health partnerships [9]. In practice, the mechanism involves the deployment of UK professionals working alongside Ugandan staff employed by K4C and local health workers in what can best be described as 'knowledge mobilisation clusters.' Long term continuity of engagement is the hallmark of K4C's approach. Understanding the contextual dynamics of AMS is critical to behaviour change at individual and organisational levels. Capturing the effectiveness of this approach – based on continuous and active co-working—requires a longitudinal ethnographic methodology with in-built reflexivity.

The MSI is reported in full in Ackers et al. [10]. This paper focuses on the mechanism that has supported the emergence of clinical pharmacy at FPRRH, and could form the basis of highly effective AMS. The development of this 'mechanism' has taken place over the past year. It has evolved in an iterative fashion as part of a continuous, exploratory, journey supported by on-going ethnographic co-researching. This type of approach does not lend itself to a linear, before-and-after, hypothesis testing structure. The paper instead charts the evolution of the MSI and the data collected along the way. As data presents new theories, this then creates new opportunities for data collection.

The intervention team started with a focus on SSI wounds, which led to an initial emphasis on wound care. Nurses and midwives are the custodians of wounds in Ugandan public facilities. The quality of wound care was found to be grossly inadequate at project inception; wound dressing was infrequent, and health workers were avoiding this task. Improvements in wound care led by nurses and midwives created the opportunity for swabbing and laboratory testing. Active engagement between nurses, midwives and laboratory scientists then created the evidence base, stimulating the opportunity for highly effective and impactful clinical pharmacy and multi-disciplinary team working. The first part of the paper tracks this process. Ultimately the effectiveness of this team in achieving

optimal AMS is limited by access to antibiotics and IPC supplies. The second part of the paper presents data evidencing the dynamics of supply chain failures in the Ugandan public health system.

#### **2. Methods**

The approach can best be described as a multi-method ethnography, commencing with observational work on the ground. Observational work was undertaken on a co-researching basis with a lead role played by Ugandan staff employed through K4C, supported by repeated and extended site visits by the Principal Investigator and virtual co-presence over a 15-month period. The team were joined by the Ugandan lead and attended Hospital IPC meetings on two occasions. Observations, complemented by on-going WhatsApp and Skype conversations were recorded in notebooks, minutes, reports and emails, and entered into NVIVO for storage and analysis.

This observational research generated theory inductively which, in turn, stimulated the search for other sources of data. Although we had anticipated accessing facility data on antibiotic consumption we could not have known or understood the complexity of this process and the challenges of even defining consumption in a public hospital setting prior to the start of the project. In such situations and given the essentially inductive quality of ethnographic research, where context is 'everything' [11], a simple a priori (deductive) hypothesis setting is inappropriate. In that respect, a process of conceptualisation, theory generation and data collection took place simultaneously. Every attempt to record or collate data stimulated intense on-going discussions about the recording processes and the nuances of its interpretation. In most cases it led us to new lines of enquiry (theories) and approaches to data collection. Much of the data, as is normal in this context, was not collated and had to be manually and painstakingly searched for from casefiles or record books. The very poor quality of documentation in patient files and subsequent records management is a critical dimension of context with implications for AMR [8]. Data collection became a process of exploration, involving forms of local capacity-building along the way on methods of organising and storing hospital records and entering them into excel spreadsheets. In this context (as in many others), much of the facility-based data could not be interpreted at face value as facts; but rather, as artefacts reflecting their (social) construction.

Facility data has been collected from a wide range of sources. Firstly, data on drug orders and supplies from National Medical Stores (NMS), was obtained through an on-line national pharmacy data base, known as the Rx system, the use of which was functionalised through the project. This was supplemented by data from paper-based records (the Dispensing Log) of supplies distributed from the central hospital stores to the wards over a 4-month period, from December 2019 to April 2020. Further, the Infectious Diseases Institute (IDI) supported hospital laboratory have proved key partners both in the intervention itself, with laboratory results providing a critical stimulus to multi-disciplinary team working, but also in generating research data. This commenced prior to the project as part of Ackers-Johnson's microbiology doctorate [12] and has continued throughout, generating valuable data on resistance patterns. The laboratory provided data on test results of samples taken from the PNG ward in 2019. This complemented a data set generated from 142 cases of suspected sepsis between January 2019 and February 2020 that were identified through a manual search of paper-based patient records.

In January 2020, a phase of qualitative interviewing took place to capture perceptions of the impact and effectiveness of the intervention. Twenty-five interviews were conducted with all cadres involved in the MSI, including 50% of the nurses, midwives, intern doctors, laboratory technicians and pharmacists working on the PNG ward, two hospital managers and three UK volunteers. The interviews were transcribed and thematically analysed using NVivo 12. Ethical approval for the work was gained from the University of Salford, Makerere University and the Ugandan National Council of science and technology (HS249ES).

#### **3. Results and Discussion**

The MSI built on on-going under-pinning research including three PhD studies. One of these (Ackers-Johnson) involved active co-researching on emerging AMR patterns with microbiologists in the hospital laboratory. The team was aware that the laboratory was struggling to obtain adequate samples from the hospital wards for testing and that maternal sepsis was one of several priorities for their research.

In common with all projects funded by THET, the CwPAMS funding stream identified a knowledge transfer mechanism based on the harnessing of UK (National Health Service) health worker expertise as the basis for behaviour change interventions. The role and contribution of professional volunteer engagement has been researched extensively by the authors, with an emphasis both on the impacts on LMICs [9,13] and, in a study financed by Health Education England, the benefits to the NHS [13]. Our approach to the MSI was informed by this research and resulted in the decision to deploy professional volunteers in co-working, mentoring roles for the duration of the intervention. The aim was to have continual presence on the ground with UK volunteers working alongside locally recruited staff (through K4C) and health workers in the hospital. One of the volunteers recruited was a member of the Ugandan diaspora working in the NHS. This volunteer knew the region, spoke the local language and specialised in wound care and SSIs. The importance of creating the conditions for serendipitous opportunities to influence action-research interventions has been reported elsewhere [14–16].

#### *3.1. The Maternal Sepsis Intervention and Wound Management as the Focus for Change*

The MSI proposal made no specific reference to wound care; wounds were something to be swabbed in order to test resistance patterns, and we had not anticipated the value of wound care to AMR work and patient outcomes. The early decision to focus on post c-section wounds turned out to be pivotal; it encouraged a very grounded approach focusing on multi-disciplinary team working at the patient's bedside. The lack of effective wound care was found to be contributing to extended patient stays and inappropriate use of antibiotics. More immediately, the ward had become associated with the stench of infection; nurses and midwives were reluctant to spend time with patients with badly infected wounds. They considered the work to be unpleasant and, in the absence of hand washing and protective clothing, staff feared the risk to their own health. The initial engagement in wound cleaning and dressing by the K4C midwives, the UK volunteers and a pioneering local midwife stimulated an holistic investment in IPC measures. This very quickly delivered major and very tangible results: Critically, it created a safer and more comfortable environment for wound swabbing. Working closely with our hospital laboratory colleagues, the project began to see a transformation of practice from a situation where no wounds were being swabbed (or other samples taken) to one where wounds were being dressed (and seen) twice daily; all patients with suspected infections were being identified, having samples taken and sent to the laboratory for culture and sensitivity testing. Table 1 presents the results of data collected from case files of all suspected sepsis cases in the 12 months commencing 1st January 2019. They show the lack of swabbing and culture and sensitivity testing on the wards prior to project commencement. The implementation aspect of the project began to impact in July, after a short initial transition period with few cases swabbed. After 22nd July 2019, nearly all suspected sepsis cases have been identified and samples sent to the laboratory for testing.

Laboratory results from these tests were present in the files of 67 of the 74 (90.5%) patients who had had a swab taken. For four patients, the results had gone missing from the file; for two patients the test was not completed because the IDI hospital laboratory was closed over Christmas and New Year and one patient's lab test was not completed because the patient had discharged herself against medical advice. Although this emphasises the importance of improving record-keeping, this level of documentation represents a remarkable achievement in the context.


**Table 1.** Volume and Proportion of Suspected Sepsis Cases Sent for Laboratory Testing.

The impact of the focus on wound care and culture and sensitivity testing is explained by a local midwife. She had taken a particular interest in the use of sugar in wound care prior to the project (in Sudan) and had previously worked alongside K4C staff and British nursing students on the labour ward, so relationships were strong. She describes the impact the project has had on her personally and on the ward and patients. She notes that, prior to the project, empirical prescribing of antibiotics lacked the desired effectiveness, and this lack of effectiveness was compounded by prolonged prescribing of the same antibiotics. Importantly, she also specifically recognises the role that clinical pharmacists are now playing:

*You came in at a critical time [and] brought new skills. Before there was no culture and sensitivity testing. Some of us knew about it but had never used it—even the doctors. When you came in it is me who benefitted most; I was carrying a very heavy burden and you helped me. You came as a combined team. We have not lost any women from sepsis since the project started and Dorothy (a Ugandan midwife employed by K4C) came. I had worked with her on labour ward with your students. Even the laboratory has started to respond—the burden was lifted, and everyone started getting involved.*

*We did use culture and sensitivity tests in Mulago (National Referral Hospital) but with not much emphasis and sometimes you have your interests on other things and we left it to the doctors. Here much of the things are now done by nurses*/*midwives—like doing culture and sensitivity tests. We knew culture and sensitivity would get results. Now I try to do the septic patients first. Before we noticed some were not getting better and we did not pay much attention to how this woman has been on this treatment for so long and you just gave her more antibiotics. [* . . . *] now [the pharmacist] comes on the ward daily and looks around and helps us as sometimes the intern doctors are busy and lack supervision. Before we used the same medicines—same—same—we just gave what was prescribed.*

In addition to describing the importance that swabbing and testing has made to progress, the midwife alludes to a major change in team-working and task-shifting with midwives and nurses now playing a very central role in these processes. This has been critical to the effectiveness of the MSI. Midwives and nurses are barely mentioned in the NAP. In the Ugandan context their active engagement and empowerment is absolutely essential to AMS, not least because they are most often the only cadres continually present on the ground. The presence of senior doctors is at best sporadic with rotating and largely unsupervised intern doctors providing most medical input [10,17,18]. This evidence adds weight to Brink et al.'s proposal for new nurse-led models of AMS in Africa [19].

The following section examines how the presence of laboratory results has created the opportunity for improved antimicrobial stewardship through clinical pharmacy engagement.

#### *3.2. AMS Performance Indicators: The Engagement of Clinical Pharmacy at FPRRH*

One of the key AMS performance indicators identified by the pharmacy team at FPRRH is the 'Review of Pharmacotherapy' by pharmacists. In practice we are concerned here with the extent to which pharmacists are directly engaged in multi-disciplinary decision-making following the receipt of laboratory test results showing resistance patterns. The data collated from patient notes showed that pharmacists reviewed the pharmacotherapy in 91.8% of cases where test results showed a bacterial growth.

We must not underestimate the impact of the introduction and embedding of culture and sensitivity testing to the team-building process. Test results trigger team-based activity; they engage all staff irrespective of cadres. Having the laboratory results provides a focus for interests to coalesce around; they stimulate team discussion and active pharmacy engagement and create the environment for genuinely patient-centred care. Having the evidence-base for rational prescribing undermines entrenched disciplinary hierarchies. It was clear from the discussion in a hospital-wide Infection Prevent Control (IPC) Committee that the tension between doctors and pharmacists persists in other areas of the hospital and had been evident, on occasion, on the PNG ward:

*Sometimes you find a pharmacist has changed a prescription and then on the ward round the doctor changes it back to a drug the patient is resistant to. (Midwife)*

A pharmacy intern responded to this comment:

*I think it should be teamwork here and respect for each other. If we advise and then the prescription is changed the clinicians come and undermine that decision without listening to the pharmacist. We can see that post-natal is taking the lead in consulting with pharmacy, but other units are relying on empirical usage designed for health centres and not for hospitals. If you are rigid on the usage you will not use the pharmacists*/*laboratory's advice.*

Evidence of a transformational increase in direct clinical pharmacy engagement on the wards is supported by qualitative findings. One midwife refers to the impact of laboratory results on these hierarchies:

*Sometimes there can be ego—that the doctor or pharmacist thinks, 'I am the overall boss so I can't be directed on what to do', but with the data that goes down.*

Every respondent identified the improvement in teamwork and identified this as the source of change on the wards:

*We are now working hand-in-hand with the pharmacists, the laboratory and the doctors—in-charge nurse.*

An intern doctor also describes how useful he has found the expertise of the pharmacy team:

*It's changed a lot now; the senior pharmacist comes regularly, and you may find there are 2 microorganisms sensitive to di*ff*erent antibiotics. Now I don't have the time to walk to the pharmacy and those people have studied medicines. I have textbook knowledge; if someone has a UTI (Urinary tract infection) I give x. I did study this but as time goes on you get used to giving certain medicines quite often and you are not so equipped to understand how one medicine interacts with another one or if a patient has TB or is HIV positive or how to combine drugs—so a pharmacist being available on the ward has really brought in great improvement.*

A midwife shows her appreciation of the teamworking environment that has developed. She refers to the presence of pharmacy on the ward and the lengths the pharmacists have gone to, to try to secure appropriate antibiotics:

*It has greatly improved because right now we have pharmacists who come on a daily basis or if not, every day we can't go 3 days without seeing one who can guide us on the mothers and which drugs to take. They interact with the doctors; if you don't interact there is that collision. Right now, there is no tension. They say, 'what if we do this' and there is a discussion. We never used to have any pharmacists coming on the wards, so it was majorly the doctors dealing with the prescriptions. We have managed to reduce the irrational use of antibiotics.*

The use of language in this response by the midwife illustrates the growing status of midwives and nurses in the multi-disciplinary teams and their ability to talk confidently about 'rational' prescribing; a concept they would not have been aware of prior to the MSI. It also illustrates the 'boundary spanning' role nurses and midwives are playing on the wards mediating professional hierarchies and tensions. Pharmacists were also very aware of their role in mediating these boundaries, and take care not to 'clash' with doctors:

*[Pharmacy] don't see all the cases. We try to pick cases where we feel pharmacists can have an input and that way, we don't clash so much with the doctors. We work within our mandates so there are no clashes. Intern doctors are also using this as a chance to learn about AMR.*

The MSI has achieved optimal pharmacy engagement (in an RRH context) on the PNG ward. The impact of laboratory testing has played an important role in empowering pharmacy. This is evident in the new policy, initiated by pharmacy with strong support from the laboratory, of only permitting use of high-end antibiotics when laboratory test results are available. A midwife respondent makes the point that there is a limited role for pharmacy on the ward in the absence of laboratory results:

*Pharmacy will tell you there is no point in them coming unless there are cultures. Clinicians are not allowed to change antibiotics now without cultures.*

Discussions have taken place in the hospital's IPC committee to extend this policy to all wards, illustrating the wider impact of the MSI on the hospital as a whole. The laboratory respondent welcomes this achievement, which also represents the growing recognition of the pharmacy presence in the hospital:

*The policy of only prescribing high end antibiotics to patients who have had culture and sensitivity testing has really worked; these antibiotics are being guarded jealously now. In fact, (the pharmacy team) are very strict on that. I really feel this could work on other wards. It is only working on post-natal ward at present because they have laboratory reports.*

#### *3.3. Creating the Evidence Base and Momentum for a Hospital Antibiogram*

Another important aspect of this wider impact can be seen in the role that the MSI has played in creating the evidence base for a hospital antibiogram. An antibiogram is a collection of data, based on laboratory testing of the pathogens in a specific facility that summarises patterns of resistance to different antimicrobial agents (or antibiotics). Although international and national trends in resistance patterns can be identified, regional and facility-specific patterns enable even closer targeting of antibiotics.

Whilst we have seen the benefits of culture and sensitivity testing in terms of trying to identify the optimal antibiotic for individual patients; in cases of suspected sepsis, health workers cannot wait for the test results, but must immediately start the patient on an antibiotic, whilst awaiting the testing process. This is known as 'empirical prescribing'. In FPRRH (as in many other facilities), the prescribing decision, usually made by a junior doctor, will be based on their usual practice, perhaps with reference to the formulary—and is very much tempered by their perception of what is available in stores. Where a hospital antibiogram exists, this initial empirical prescribing can be informed by local evidence and has a much higher chance of success. The presence of an antibiogram with associated awareness raising and sensitisation amongst all staff, and especially medical interns, would have major impacts on empirical prescribing across the hospital. Prior to the MSI, FPRRH did not have the volume of laboratory results to create the necessary evidence base for a hospital antibiogram. A member of the pharmacy team describes how this has changed:

*If you go to maternity, you will notice a very big change. The ward sends the biggest volume of swabs now to the laboratory because those people [midwives] are aware.*

The laboratory scientist confirms this:

*On the basis of the increased swabbing we hope to be in a position to have an antibiogram. This will be very informative—the sample size is now very adequate, but we want to enter this information into a comprehensive database which has di*ff*erent parameters—length of stay–age-sex-ward–so that when you are doing the analysis it is very comprehensive. The antibiogram will be good for the clinicians to guide prescribing—it will be good for the patients.*

The results presented above demonstrate the ability to make considerable progress in AMS at a RRH.

#### *3.4. Access to Antibiotics in FPRRH: Supply Dynamics*

Ultimately, the model that has evolved on PNG ward has the potential to significantly reduce infection, improve prescribing practice, reduce antibiotic consumption and overcome some of the effects of AMR on patient outcomes. The major challenge facing the MSI model is access to the right antibiotics and antimicrobials at the right time. The laboratory scientist is clear about this:

*Antibiotic stock-outs remain a serious constraint; in many cases patients can only be given the right antibiotics if they pay and many of them can't pay. We have to be very clear, antibiotic stock-outs are a key factor fueling AMR. If we look at the scenario where we have investigations done and antibiotics are available, and the outcomes are good, but we have done the investigations and the antibiotics are out of stock we won't have a good outcome.*

There is a bigger concern here too; if supplies are not available and the ward staff are unable to respond effectively to laboratory results this can be predicted to have a major impact on staff motivation and the behaviour change gains. Problems of access critically restrict pharmacy's ability to engage in rational prescribing; prescribing the drug most likely to work according to laboratory results. This in turn leads to over-consumption of poorly performing antibiotics and poor patient outcomes. The following section describes the supply chain system at FPRRH and presents data on consumption. Critical problems include:


#### *3.5. The Impact of the MSI Project on the 2020*/*2021 Procurement Plan*

In Uganda, the funds for procurement of drugs and supplies in the public sector are highly centralised and inadequate. NMS procures and distributes supplies to health facilities based on a centrally allocated Annual Supplies Budget. Each hospital is required to produce an Annual Procurement Plan. Once agreed, this Plan is fixed and cannot be varied over the year reducing the opportunity for flexibility and responsiveness to the hospital laboratory results and any changes indicated by a future antibiogram. This budget is held by NMS. With the exception of private wards, it is not possible for a RRH to source supplies from elsewhere. NMS deliveries take place bi-monthly.

During the annual procurement process, the hospital may only order those items authorised by the 'Essential Medicines and Health Supplies List for Uganda'. However, not all essential drugs feature in the NMS catalogue. A hospital pharmacist describes the situation as follows:

*As much as we may desire a certain antibiotic, we can't plan for it if it is not present in the catalogue. A case in point is Amikacin and Moxifloxacin.*

Out of the nine antibiotics tested against *Acinetobacter* samples in the laboratory, only two—doxycycline and amikacin—showed greater levels of susceptibility than resistance (for details of these test results, see Supplement 1). Given the much higher success rate of amikacin, it is paramount that the antibiotic can be obtained for cases of severe *Acinetobacter* infections where other avenues have failed.

Within the constraints described above, the MSI has influenced procurement planning for 2020/21. Figure 1 evidences significant changes in antibiotic ordering and consumption arising directly from the intervention, where antibiotics in red denote project-related increases and those in blue denote decreases.
