**About the Special Issue Editors**

**Zeno Stanga** studied medicine at the University of Bern and obtained board certification in internal and general medicine in 1995. He trained in general internal medicine, nutritional medicine and metabolism in several hospitals in Switzerland. After research fellowships in Nottingham (U.K.) and Charleston (USA), he founded a nutritional support team in 2001 and leads the nutritional medicine unit at the University Hospital of Bern. He is currently associate professor for internal and nutritional medicine at the Medical Faculty of the University of Bern. He is a member of the appointment commission and active in several working groups of the medical faculty. Professor Stanga was responsible for the Swiss Certificate of Advanced Studies in Clinical Nutrition and acted on the board of directors of the Swiss Society of Clinical Nutrition for 20 years. He has published over 150 scientific papers and is associate editor of the ESPEN Blue Book.

**Philipp Schuetz** was born in Switzerland and studied Medicine at the University of Basel, Switzerland, and the University Kremlin Bicetre in Paris, France. He is a board-certified internist and endocrinologist with special interest in clinical nutrition. He is the head of internal medicine and emergency medicine Kantonsspital Aarau and part of the faculty of medicine at the university in Basel, Switzerland. He has published more than 300 studies and research articles in high-impact journals including *Lancet, JAMA, Annals of Internal Medicine*, and many others. Prof. Schuetz obtained a research professorship from the Swiss National Science Foundation (SNF) and was principal investigator of the EFFORT trial, the largest-yet randomized controlled trial examining the clinical effects of clinical nutrition in medical ward patients.

### *Editorial* **Nutritional Management and Outcomes in Malnourished Medical Inpatients in 2020: The Evidence Is Growing!**

**Philipp Schuetz 1,2,\* and Zeno Stanga <sup>3</sup>**


Received: 16 December 2019; Accepted: 18 December 2019; Published: 20 December 2019

Access to adequate food is a fundamental human right. There is no doubt that nutrition is essential in maintaining health and preventing or treating disease. Particularly when patients are affected by disease-related malnutrition, their risk of adverse clinical outcomes increases significantly and optimizing nutritional support becomes mandatory [1,2].

There is ongoing debate about what constitutes an optimal nutritional care process in terms of screening, assessment, and use of nutritional support in different patient populations. Issues include dose and quality of proteins and total energy, route of delivery, and whether or how nutritional support needs to be adjusted for specific medical and metabolic conditions. As we move toward personalized medicine, which is based on patients' individual needs, an understanding of these different factors is important. Well-planned clinical studies of high methodological quality are needed to develop the best approach to providing individualized nutritional support [3].

Historically, much of the evidence regarding effects of nutritional support has come from small interventional trials and observational studies with cross-sectional or cohort-study designs, whereas there was an important lack of large-scale randomized interventional research, which is needed to establish causal effects rather than just statistical associations [4,5]. As a consequence, the medical community has struggled to design efficient, evidence-based approaches for the prevention and treatment of malnutrition [4].

Recently, however, the results of several high-quality trials have provided important new insights that advance nutritional science significantly and translate nutrition research into practice [6]. Regarding prevention of cardiovascular disease through nutrition, the PREDIMED (Prevención con Dieta Mediterránea) trial provided strong evidence that a Mediterranean diet supplemented with extra-virgin olive oil or with mixed nuts reduces the risk of cardiovascular and metabolic disease by about 30% over five years [7]. Regarding the use of clinical nutrition in patients at nutritional risk or with established malnutrition, two trials found nutritional support to be highly effective. First, the multicenter, randomized, placebo-controlled NOURISH trial (Nutrition effect On Unplanned ReadmIssions and Survival in Hospitalized patients) including 652 older adults affected by malnutrition found that a high-protein oral nutritional supplement containing beta-hydroxy-beta-methylbutyrate was associated with a significant reduction in 90 day mortality, with a number needed to treat (NNT) of 20 [8]. Second, the EFFORT (Effect of early nutritional support on Frailty, Functional Outcomes and Recovery of malnourished medical inpatients Trial) including 2028 medical inpatients at nutritional risk in eight Swiss hospitals showed that protocol-guided individualized nutritional support designed to achieve protein and energy targets results in significantly lower rates of severe complications (NNT = 25) and mortality (NNT = 37) compared to regular hospital food [9,10]. Moreover, functional

decline was significantly lower, and quality of life as well as activities of daily living significantly improved. A recent meta-analysis including these two trials and several other trials also came to the conclusion that nutritional support in the malnourished medical inpatient population reduces the risk for both, mortality and hospital readmission by about 25% [6].

Evidence-based medicine is an approach used to optimize decision-making by emphasizing evidence from properly designed and well-conducted research, typically randomized trials and meta-analyses from such trials. With the growing number of high-quality trials such as the ones mentioned above, we are increasingly able to practice "evidence-based clinical nutrition" and to adapt nutrition to the individual patient's needs. This Special Issue of the *Journal of Clinical Medicine* (JCM) focuses on a topic that is critical for hospitals today: "Nutritional Management and Outcomes in Malnourished Medical Inpatients". This special edition presents a number of reviews and original research articles in the field of nutritional management and clinical outcomes in malnourished medical inpatients. Twenty-six important articles illustrate the different facets of this complex and timely topic.

The articles included cover the process of nutritional care, including screening tools to identify nutritional risk (Nutritional risk screening and assessment [11]), patient muscle mass assessment including bioimpedance analysis (Clinical value of muscle mass assessment in clinical conditions associated with malnutrition [12]; Decreased bioelectrical impedance phase angle in hospitalized children and adolescents with newly diagnosed type 1 diabetes: a case-control study [13]), nutritional biomarkers (Nutritional laboratory markers in malnutrition [14]), nutritional therapy planning (Indirect calorimetry in clinical practice [15]; Micronutrient deficiencies in medical and surgical inpatients [16]), use of nutritional support overall (Efficacy and efficiency of nutritional support teams [17]; Challenges and perspectives in nutritional counselling and nursing: a narrative review [18]) and in specific patient populations (e.g., medical patients, critical care patients, geriatric patients, oncologic patients, patients after allogenic stem cell transplantation, patients with dysphagia or eating disorders, as well as the nutritional challenges associated with metabolic disorders) (Nutritional management of medical inpatients [19]; Medical nutrition therapy in critically ill patients treated on intensive and intermediate care units: a literature review [20]; Metabolic and nutritional characteristics of long-stay critically ill patients [21]; Protein intake, nutritional status, and outcomes in intensive care unit survivors: a single-center cohort study [22]; Early supplemental parenteral nutrition in critically ill children: an update [23]; Management of malnutrition in older patients—current approaches, evidence, and open questions [24]; Nutrition in cancer patients [25]; Management of dehydration in patients suffering swallowing difficulties [26]; Nutrition in gastrointestinal diseases: liver, pancreatic, and inflammatory bowel diseases [27]; Nutritional management and outcomes in malnourished medical inpatients: anorexia nervosa [28]; Nutritional challenges in metabolic syndrome [29]; Nutritional challenges in patients with advanced liver cirrhosis [30]). Potential complications of nutritional interventions, such as refeeding syndrome (Management of refeeding syndrome in medical inpatients [31]), and treatment challenges posed by gastric motility disorders are discussed (Gastroparesis and dumping syndrome: current concepts and management [32]). Economic aspects of nutritional management (Economic challenges in nutritional management [33]) and specific considerations such as pharmaceutical/therapeutic aspects of artificial nutrition are additionally reviewed (Management of glucose control in non-critically ill, hospitalized patients receiving parenteral and/or enteral nutrition: a systematic review [34]; Pharmaceutical aspects of artificial nutrition [35]).

Last but not least, the call for political commitment in the treatment of malnutrition is of key importance. Health care institutions and associations must be mobilized to take action against malnutrition by expanding information and public-awareness-raising campaigns, adopting supportive policies, as well as allocating resources (Hospital malnutrition, a call for political action: a public health and nutritionDay perspective [36]). Stakeholders including political organizations, nutritional social networks, and researchers will be valuable promoters of this important political priority in the future.

Understanding the optimal use of nutritional therapy is highly complex. Timing, route of delivery, and the amount and type of nutrients all play important roles and potentially affect patient outcomes. Recent trials provide important information to strengthen the evidence regarding the use of nutritional therapy in specific patient populations, but there are still important questions to be addressed by robust clinical trials in the future. It is now important to incorporate these recent findings into clinical practice in order to ensure that our patients receive high-quality, safe, and optimal care.

**Author Contributions:** P.S. and Z.S. wrote this article and take full responsibility for its content. All authors have read and agreed to the published version of the manuscript.

**Funding:** The article processing charge was funded by the Research Fund of the Department of Diabetes, Endocrinology, Nutritional Medicine, and Metabolism and in part by Nestlé Health Science.

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

#### **References**


© 2019 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* **Hospital Malnutrition, a Call for Political Action: A Public Health and NutritionDay Perspective**

**Michael Hiesmayr 1,\*, Silvia Tarantino 1, Sigrid Moick 1, Alessandro Laviano 5, Isabella Sulz 2, Mohamed Mouhieddine 1, Christian Schuh 2, Dorothee Volkert 6, Judit Simon <sup>3</sup> and Karin Schindler <sup>4</sup>**


Received: 29 September 2019; Accepted: 14 November 2019; Published: 22 November 2019

**Abstract:** Disease-related malnutrition (DRM) is prevalent in hospitals and is associated with increased care needs, prolonged hospital stay, delayed rehabilitation and death. Nutrition care process related activities such as screening, assessment and treatment has been advocated by scientific societies and patient organizations but implementation is variable. We analysed the cross-sectional nutritionDay database for prevalence of nutrition risk factors, care processes and outcome for medical, surgical, long-term care and other patients (*n* = 153,470). In 59,126 medical patients included between 2006 and 2015 the prevalence of recent weight loss (45%), history of decreased eating (48%) and low actual eating (53%) was more prevalent than low BMI (8%). Each of these risk factors was associated with a large increase in 30 days hospital mortality. A similar pattern is found in all four patient groups. Nutrition care processes increase slightly with the presence of risk factors but are never done in more than 50% of the patients. Only a third of patients not eating in hospital receive oral nutritional supplements or artificial nutrition. We suggest that political action should be taken to raise awareness and formal education on all aspects related to DRM for all stakeholders, to create and support responsibilities within hospitals, and to create adequate reimbursement schemes. Collection of routine and benchmarking data is crucial to tackle DRM.

**Keywords:** malnutrition; hospital; nutrition care; continuity of care; mortality; process indicators; benchmarking; disease related malnutrition.

#### **1. Introduction**

Disease-related malnutrition (DRM) is highly prevalent in hospitalized patients and associated with complications and poor outcome [1–3]. Malnutrition in hospitals originates from imbalances, either deficiencies or excesses, in nutrients intake compared with body needs. Nutritional status and needs may be modified acutely or chronically by the disease process itself. Further deterioration may occur due to hospitalization, thus making the population of hospitalized patients very different from

the general population. Prolonged nutrients imbalance is associated with change in body mass index (BMI). Association between body mass index (BMI) and mortality is U-shaped in the general population and J-shaped in patients, especially with chronic diseases, meaning that mortality is higher if BMI is low and lower in patients with increased lean body mass and even obesity [4,5]. This observation called "obesity paradox" underscores the importance of a "good" nutrition status for patients with illnesses for short and long-term outcomes.

Malnutrition in hospitalized patients often addresses an evident poor nutritional status (low BMI and low muscle mass) whereas being at risk of malnutrition is derived from a set of risk factors typically associated with a loss in lean body mass persisting over a certain period of time [6]. Nutrition care in hospitals is a treatment for patients with malnutrition and a preventive intervention for patients at risk of malnutrition. Nevertheless, nutrition care is still an underrated field when compared to medical diagnostics procedures, or pharmacological and technological interventions in hospitals.

There are several reasons why nutrition care has received so little attention in acute care hospitals. There is a common knowledge deficit reflected in a lack of proper education in university curricula for healthcare professionals (doctors, nurses, care assistants) combined with patients and relatives giving low value to nutrition as part of a successful therapy of the primary disease. Since no immediate effects of nutrition can be expected during a short hospital length of stay, the attention for nutrition related issues is often low. Standard nutrition care processes such as screening, assessment, planning and monitoring together with documentation and continuity of care are not regular parts of care on all hospital wards [7]. Moreover, food provision and related tasks are not considered part of healthcare responsibilities. Food provision is not part of the medical budget but usually of the administrative budget of a hospital where cost reduction is not considered to influence directly patient care. Inadequate management of food provision might affect food quality, presentation [8] and composition and, subsequently, patient care. Food costs are to be added to the overall malnutrition related costs. Patients with malnutrition usually stay longer in hospitals, are more often re-hospitalized or transferred to long-term care [9–12]. Each of these, together with the reimbursement schemes, do create additional costs for the healthcare system. Insights on improved hospital nutrition care processes and reduced healthcare costs come from the Swiss EFFORT study.

In patients with risk of malnutrition, nutrition intake was improved with enriched meals or oral nutritional supplements, food intake was monitored and was associated with better outcomes and reduced healthcare costs [11].

These multiple barriers and the lack of proper attention to DRM and nutrition care in hospitals has been formally addressed at a political level by a resolution of the European Council in 2003 but was not followed by national regulatory actions [3].This lack of action led to several independent initiatives. The European Society for Clinical Nutrition and Metabolism (ESPEN) (www.espen.org) together with the Medical University of Vienna developed a tailored action to tackle malnutrition in hospitals and healthcare institutions [4]. The resulting project "nutritionDay" aimed to generate more awareness on DRM with a yearly one-day data collection on patient's malnutrition risk factors, outcomes and quality indicators of nutrition care (www.nutritionday.org). In the nutritionDay analysis, several nutrition related risk factors, such as low BMI, recent weight loss and reduced food intake (in the week before nutritionDay or on nutritionDay itself), were found to be independently associated with death within 30 days in hospital [4]. In an another analysis that led to the development of the PANDORA score for prediction of death in hospital within 30 days after nutritionDay, decreased food intake was identified among the seven most important risk factors as it contributed 3–12 points out of a maximum of 75 points to the final score [5]. nutritionDay rapidly became a worldwide benchmarking tool for monitoring and improving nutrition care in hospital wards [13–15]. Patient related risk factors for decreased eating have a similar pattern in all world regions [16] and nutrition care processes implementation appears highly variable [7].

A multi-stakeholder initiative to promote screening for risk of disease-related malnutrition/undernutrition and implement nutritional care across Europe "Optimal Nutritional Care for All" has started an annual meeting with national nutrition societies and political representatives in 2008 [17]. The last meeting in Sintra, Portugal in 2018 involved representatives from 18 countries that had already joined the initiative and had the motto: "Optimal Nutrition Care across Europe: Fair Access and Shared Decision Making". The decision was taken together with the European Patient Forum [18] to translate relevant guidelines into a lay version to increase the possibility for patients to take informed responsibility [19].

The human right for proper nutrition care was acknowledged in the Cartagena declaration [20] and signed on 3rd May 2019 by all presidents of the Latin American Federation of Nutritional Therapy, Clinical Nutrition (FELANPE). The declaration includes 13 principles, such as patient empowerment, dignity, ethical principles, justice and equity and urges the United Nations and the Human Rights Council to recognize the Right to Nutrition Care as a human right in line with the "Sustainable Development Goals" [21].

The worldwide Global Leadership Initiative on Malnutrition (GLIM) composed by the major nutrition societies aims at defining universally accepted criteria for DRM [22]. This process is still ongoing with planned steps of validation and regular updating. Current recommendations from scientific societies emphasize that it is mandatory to identify the malnourished as well as those at risk early during hospitalization to trigger proper treatment or a set of preventive measures [19]. The recommended three step process consists of screening, assessing and developing a nutrition care plan. A further important and less appreciated step is that nutrition care needs to be monitored and adapted to the patient's changing condition. Finally, proper documentation and communication of a comprehensive care plan to the next sector, "extramural health care", and the patients themselves are essential to ensure continuity of care. To be efficient, responsibilities need to be cleared delineated and all relevant stakeholder (Table 1) involved.


**Table 1.** List and role of important stakeholders.

a. Patients and relatives

b. Extramural medical services/family medicine/primary health care centres

c. Extramural care services/mobile nursing

d. Services for disabled and dependent persons

e. Local food producers

f. Medical food producing industries


**Table 1.** *Cont.*

b. Public procurement of food supply and services

This study aims to determine in the medical patients of the nutritionDay database 2006–2018, first, the prevalence of simple nutrition related risk factors and their association with outcome and, second, to determine the routine use of recommended nutrition care procedures such as screening, nutrition intake monitoring and documentation in patients with and without risk factors.

Based on the findings we will propose several options for political action to tackle malnutrition in hospitals.

#### **2. Experimental Section**

The nutritionDay audit is a cross-sectional international data collection in hospitalized patients with 30 days in hospital outcome assessment. The nutritionDay project was approved by the Ethical Committee of the Medical University of Vienna (EK407/2005) and it has been amended annually. In accordance with national regulations, the project was also submitted to national or local ethical committees in each participating country. This trial was registered at clinicaltrials.gov as NCT02820246. We analysed the complete nutritionDay database 2006–2015 for the association between nutrition related risk factors and 30 days hospital mortality. We assessed the 2016–2018 nutritionDay database for nutrition care process indicators newly introduced in 2016.

We included all adult patients with the exception of women before or after giving birth. Patients were divided into four groups: group surgical (patients admitted to a surgical unit or waiting for surgery or after surgery on any medical/other ward), group medical (all patients admitted to general medical, cardio, gastro, hepatology, nephrology, infectiology or oncology), group long-term (geriatric or long-term care wards) and others (all other specialties such as ear-nose-throat, gynecology, obstetrics, trauma, orthopedics, etc.). Risk factors were age, gender, BMI, weight change in the last three months, decreased eating during the previous week and on nutritionDay, fluid status (as intravascular or tissue fluid overload or depletion as observed by a clinician), reduced mobility and whether patients have been admitted to an intensive care unit at any time before nutritionDay during this index admission. In addition we included in the multivariate analysis the diagnostic categories derived from the 17 ICD 10 top categories (brain and nerves, eye and ear, nose and throat, heart and circulation, lung, liver, gastrointestinal tract, kidney/urinary tract/female genital tract, endocrine system, skeleton/bone/muscle, blood/bone marrow, skin, ischemia, cancer, infection, pregnancy, others) and six comorbidities (diabetes, stroke, COPD, myocardial infarction, cardiac failure and others), each one used as a unique variable as condition being present versus not present. All factors were used as categorical variables and included a missing category according to the STROBE guidelines [23]. Reference categories were either the "normal" category or the group including the median or the largest subgroup as appropriate. Age was

divided into eight groups spanning 10 years with 60–70 years as reference; BMI was divided into 6 WHO groups with normal BMI 18.5–25 used as a reference [24].

Sensitivity analysis included only patients from units fulfilling quality criteria, such as >60% recruitment of admitted patients and outcome reporting for >80% of included patients. In addition a second sensitivity analysis was done for the multivariate modelling of the association between risk indicators and outcome after exclusion of all cases with missing values. The sensitivity study population includes 82,993 patients compared with 153,470 in the full study population.

Descriptive statistics report frequencies and median with interquartile range. Comparison of proportions were done with the χ<sup>2</sup> test and corrected for multiple testing. All associations between mortality and risk factors were done with logistic regression based on general linear models with units as clusters and weighting of patients to compensate for time-based bias from cross-sectional data acquisition [25]. In short, patients with a longer observed length of stay had less weight because they were more likely to be included in the study sample than short stay patients. We used univariate analysis for all risk factors and included all significant factors in the multivariate analysis. The effect of risk factors was also analysed within all four patient groups separately in the multivariate model (STATA 15.1, Statacorp, College Station, TX, USA). To show the extent of use of selected nutrition care processes, they are shown as percentages within different risk categories. Significant differences to each reference group were tested with a proportions test, comparisons were considered significant when *p* < 0.005 as we accounted for multiple tests for each reference category.

#### **3. Results**

Medical patients in the nutritionDay database (*n* = 59,126/153,470) (39%) represent the second largest group after surgical patients (*n* = 63,289/153,470) (41%). Patients were admitted in 61 countries and 19 countries recruited more than 1000 patients.

Medical patients were four years older than surgical patients (65.2 years SD 17.2 versus 61.2 years SD 18.0). The proportion of female was 48.3 in medical and 48.1 in surgical patients. Weight (71.3 kg SD 19.4 versus 71.9 kg SD 18.1) and height (166.1 cm SD 10.3 versus 166.5 cm SD 10.2) were similar in both groups. BMI was nearly identical (25.7 SD 6.3 versus 25.8 SD 5.8) (*p* = 0.02). Based on WHO categories (www.who.int) both groups have a similar proportion of obese with BMI > 30 (17.7% versus 17.4%) but the proportion with low BMI < 18.5 was significantly higher in medical than in surgical patients (7.5% versus 6%) (*p* < 0.0001).

#### *3.1. Prevalence of Nutrition Risk Factors*

Nutrition risk factors such as weight loss during the last three months (26,790, 45%), not eating normally in the previous week (28,950, 49%) and did not eat all food served on nutritionDay (30,965, 52%) were highly prevalent in the medical patients of the cohort 2006–2015 (Figure 1) and in the three other patient groups (Table 2).

**Figure 1.** Prevalence of risk factors and association with odds ratio for death in hospital within 30 days after nutritionDay in medical patients. Prevalence is indicated by dots. Each dot represents 1% of the total population. All risk indicators are collected on one single day, the nutritionDay 2006–2015. Odds ratio are indicated with 95% confidence intervals and colours according to risk indicator categories Graph of Community–Hospital–Continuum from Magdalena Maierhofer's architectural diploma thesis: A Hospital is not a Tree (2016).

**Table 2.** Demographic characteristics and prevalence of nutrition risk factors of patients in the nutritionDay cohort 2006–2015, in the four studied groups.



**Table 2.** *Cont.*

\* indicates the reference categories.

There is some overlap between risk categories. Two third of patients (17,497/26,790) with weight loss reported not eating normal in the previous week and a similar proportion did not eat all served on nutritionDay (16,681/26,790).

Nearly 40% of patients (9551/24,679) that did eat normally in the previous week were eating less than all meal served on nutritionDay indicating a new nutrition risk associated with hospitalization.

In most patients (80%) all four nutrition risk factors, low BMI, weight loss, reduced eating during the previous week and on nutritionDay could be evaluated. 91% had no more than one risk factor missing. Only 32,216/140,418 (23%) of patients had no single nutrition related risk factor, 31% had one risk factor, 28% two risk factors and 16% had three risk factors.

#### *3.2. Nutrition Care*

#### 3.2.1. Food provision

Oral diet, either normal hospital food or special diet, was mostly used in medical patients (Table 3). Two third of oral diets were given as hospital food and one third as special diet. Oral nutritional supplements were given to 9.3% of patients whereas enteral nutrition was used in 7.1% and parenteral nutrition in 3.5% (Table 3). Surprisingly, the use of enteral or parenteral nutrition only increased by a factor of three between patients who reported having eaten their full meal and those who had eaten nothing. Two-thirds of patients reporting eating nothing were on oral diet. The use of enteral and parenteral nutrition was not differentiating much in patients who have eaten nothing regardless of being allowed to eat or not.



Percentages are indicated within each eating category. Multiple entries in the nutrition type for one patient are possible. Patients who ate nothing on nutritionDay were divided into those who did not eat even if they were allowed to (nothing\_a) and those who did not eat because they were told not to eat by the doctor (nothing\_na) as, for instance, before planned surgery or diagnostic tests. Combined enteral and parenteral nutrition was used in less than 2.5% of patients in any patient category and is not shown.

#### 3.2.2. Process indicators

A total of 1415 units reported in 2016–2018 about the screening tool utilised (Figure 2). The majority, used a formal tool such as NRS-2002 (Nutrition Risk Screening 2002), MUST (Malnutrition Universal Screening Tool), MST (Malnutrition Screening Tool), or a local tool. Only 10% of the units did not have a routine screening nor fixed screening criteria. Patients identified as malnourished were overall 3340/28,100 (11.9%), with the highest proportion identified with an informal tool 722/4793 (15.1%) or NRS-2002 1169/8453 (13.8%) followed by MUST 134/1210 (11.1%), MST 269/2269 (11.9%) and visual experience 197/1666 (11.8%) whereas the proportion of identified malnourished patients was lower with no implemented routine 204/2154 (9.5%). The proportion of patients identified at nutritional risk was overall 4944/28,100 (17.6%). Of those, the highest proportion, 22.4% was identified with an unspecified tool while only 14–16% of patients were identified at risk when no routine, only visual appearance or MST was used. Only 28.9% of malnourished patients had a BMI below 18.5, 49.8% a normal BMI, 21.4% were overweight or obese. Being identified as malnourished was associated with unintentional weight loss within the last three months in 85% of patients, with reduced nutrient intake before admission in 57% of patients and with not eating a full lunch in hospital in 65% of patients. Not being identified as malnourished or at risk of malnutrition was associated with unintentional weight loss within the last three months in 47% of patients, with reduced nutrient intake before admission in 28% of patients and with not eating a full lunch in hospital in 48% of patients. The proportion of patients with two or more nutrition risk factors was 50% in the 2016–2018 cohort.

**Figure 2.** Proportion of different methods/approaches used for malnutrition screening in 1415 units from 46 countries in the nutritionDay cohort 2016–2018. NRS-2002 (nutrition risk screening 2002); MUST (Malnutrition Universal Screening Tool); MST (Malnutrition Screening Tool); SNAQ (Short Nutritional Assessment Questionnaire).

Nutrition intake monitoring was more frequent in patients with unintentional weight loss than in patients with stable weight (52% versus 41%) (Figure 3). Still, nearly 50% of patients with a history of weight loss did not have their nutrition intake monitored while in hospital. Similarly, history of poor nutrition intake before admission triggered more frequent monitoring compared with history of

normal eating but intake was never monitored in more than 50% of patients with a history of poor nutritional intake. Actual poor food intake did not have any effect on monitoring of food intake.

**Figure 3.** Nutrition care process indicators versus three nutrition associated risk factors. Daily nutrition intake monitoring (**left**) and nutrition expert consulted (**right**). Bars indicate percentage answering "yes", significant differences to each reference group are shown with \* *p* < 0.005 and \*\* *p* < 0.00001, 'n.s.' indicates no significant difference. Missing values were <7.5% in all subcategories. Colour coding similar to Figure 1 (the empty bar is the reference).

A nutrition expert was consulted more frequently when an unintentional weight loss was reported by patients or when food intake was reported as reduced before hospital admission (Figure 3). A nutrition expert was never consulted in more than 46% of patients even when risk factors were reported. Documentation of malnutrition in patient's chart followed a similar pattern as consulting an expert and was always below 41%.

Low BMI was associated with a more frequent monitoring of intake (58% versus 47%) (*p* < 0.0001) and malnutrition reporting in patient record (52% versus 36%) (*p* < 0.0001) compared with normal BMI. In obese patients (BMI 30–35) we observed the lowest intake monitoring (40%) and malnutrition reporting (27%).

#### *3.3. Outcome*

Thirty days after nutritionDay we collected outcome data and found that 73% of the study patients had been discharged home, 11.8% had been transferred to another health care facility, hospital, long-term care or rehabilitation; 9.4% were still in the same hospital and 5.5% had died (Table 4). Similar results were observed in the sensitivity analysis. In medical patients, death was observed more than twice as frequently as in surgical patients (4.6% versus 1.7%, *p* < 0.0001) and was similar to the groups geriatrics and long-term care (4.6% versus 4.8%, *p* < 0.0001).

The group "other", which includes patients from neurology, psychiatry and non-operative patients in gynecology and ENT, has a lower mortality (2.6% versus 4.6%) (*p* < 0.0001) a lower proportion discharged home *p* > 0.0001) at day 30 after nutritionDay than medical patients (Table 4).


**Table 4.** Outcomes in hospital within 30 days after nutritionDay in the four patient groups.

Nutrition risk factors were associated with death in hospital within 30 days in a univariate analysis with odds ratios 2.6 CI 95 [2.2–3.0] for weight loss, 5.5 CI 95 [4.6–6.6] for eating less than a quarter of their meals in the previous week and 7.6 CI 95 [6.1–9.6] if eating nothing on nutritionDay despite being allowed to eat. The prevalence of individual risk categories for weight loss, decreased eating last week and on nutritionDay were associated with significantly worse outcomes (Figure 1), while increasing BMI was associated with decreasing odds ratios for death.

Other risk factors such as increasing age and male gender were also associated with poor outcome. The largest univariate association with hospital death was observed in patients with reduced mobility or those who were bedridden. Mortality within 30 days after nutritionDay increased with the number of nutrition risk factors present from 0.9% for no risk factor, 1.68% for one risk factor, 3.55% for two risk factors, 8.13% for three risk factors and 13% for four risk factors.

#### Multivariate Outcome Analysis

All risk factors that were significantly associated with 30 days hospital death in the univariate analysis were also significant in the multivariate analysis (Figures 4 and 5) including all affected organs and comorbidities that could be estimated. Most estimates indicated a reduced strength of the association of the individual risk factors compared with univariate analysis, which indicates confounding. In the multivariate analysis, history of decreased eating before hospital admission was associated with higher odds ratio for death in surgical patients than in medical patients, weight loss was associated with similar odds ratio. History of a stay in intensive care during the hospitalization before nutritionDay was found to be significantly associated with death only in medical patients. "Eating less than normal in the previous week" is associated with death 30 days after nDay in medical, surgical and long-term care patients but not in the "other" group. Weight loss is only associated with poor outcome in medical and surgical patients. Higher BMI was associated with better outcome in medical, surgical and "other" patients, an observation called "reverse epidemiology" and low BMI with worse outcome but not in long-term care patients. Higher age is in all groups associated with worse outcome. Abnormal fluid status, identified as either overloaded or dehydrated, based on clinical judgment, was a robust risk indicator in all patient groups. Decreased food intake on nutritionDay was always associated with worse outcome. A clear difference between eating nothing while not being allowed (nil by mouth) or being allowed is only found in the surgical patient group whereas in all three other groups there was no clear difference. A very robust risk indicator is "reduced mobility" in all patient groups (Figure 4). Having had an ICU stay before nutritionDay was only associated with worse outcome in medical patients.

**Figure 4.** Multivariate analysis of association between demographic and nutrition related risk factors and death in hospital within 30 days after nutritionDay for medical, surgical, long-term care and other patient groups with the general linear model for logistic regression with wards as clusters and weighting of individual patients for sampling probability [25] and including all diagnostic categories and comorbidities (see Figure 5). Odds ratios (OR) with 95% confidence intervals indicated by horizontal line. Reference categories are indicated by an open symbol. Missing values are included in the model as individual categories.

**Figure 5.** Multivariate analysis of association between organ related disease categories from ICD 10 as well as comorbidities and death in hospital within 30 days after nutritionDay for medical, surgical, long-term care and other patient groups with the general linear model for logistic regression with wards as clusters and weighting of individual patients for sampling probability [25] and including demographic and nutrition related risk factors (see Figure 4). Odds ratios (OR) with 95% confidence intervals indicated by horizontal line. Multiple entries are possible. Missing values are included in the model as individual categories.

The association between affected organs as derived from ICD 10 categories is very different for medical and surgical patients (Figure 5). No comorbidities had a negative prognostic effect in the multivariate analysis.

#### **4. Discussion**

This analysis of medical/surgical patients from the nutritionDay cohort 2006–2015 showed that individual nutrition risk indicators such as low BMI, weight loss in last three months, poor eating before admission and low nutrient intake in hospital are highly prevalent in hospitalized patients and are associated with poor hospital outcome within 30 days after nutritionDay. Each additional risk indicator observed was associated with a nearly two-times higher mortality in the overall population.

All four risk indicators were independently associated with death in the hospital within 30 days after nutritionDay in the multivariate analysis (Figure 4) in the two largest groups medical and surgical patients. There is no clear indication that a single risk indicator can be ignored because the presence of each risk indicator is associated with a higher rate of dying by nearly a factor of two for these two groups. Surprisingly recent weight-loss and lower BMI was not associated with worse prognosis in long-term care whereas only weight-loss was not associated with worse outcome for the group of non-medical and non-surgical patients including neurological and psychiatric disorders.

Three risk indicators emerge similarly for all four patient groups, poor eating in hospital, reduced mobility and altered fluid status. All three had also being found as individual risk indicators in the PANDORA risk scoring system whereas recent weight loss and history of poor eating were not included [5]. These robust factors may also serve as indicators that may trigger more consistent observation and initiation of treatments that have proven to efficacious [26].

In recent years (2016–2018) healthcare teams identified only half of the patients as having nutrition related risk factors such as being malnourished or at risk of malnutrition in comparison to the nutritionDay indicators 8701/28,013 (31%) versus 21,070/28,100 (75%) such as unintentional weight loss, low BMI or decreased food intake. Thus, it appears that a large group of patients with nutrition related risk factors is not identified in standard care practice. This observed gap may originate from the different type of data collected from patients during routine clinical practice and the targeted data collection of the nutritionDay audit.

The efficiency of interventions may depend on the size of risk of the individual patient. It has recently been demonstrated that patients identified at high risk of malnutrition by the Nutritional Risk Screening (NRS-2002) risk system benefit from an integrated intervention [26]. It is unknown, however, whether the large group of patients with moderate risk, which account for almost half of the hospitalized patients on any given day, would also benefit from some targeted interventions. Interventions effective for the moderate risk group may be less complex and costly, and potentially integrated in to routine care processes, and could lead to better outcomes and reduced hospital length of stay and better patient outcomes.

#### *4.1. Nutrition Care*

Nutrition care processes such as screening, prescribing nutrition treatment, monitoring daily nutrition intake and recording of malnutrition status are applied in less than 50% of patients. These observations may again arise from identification issues, from lack of a systematic screening processes which then lead to a weak targeting even in the presence of nutrition risk factors. Educational gaps of healthcare professionals in the field of nutrition, as well as the importance given to nutrition care might also be considered crucial to fight DRM. Recent results from a large randomized trial [26] from a Swiss team could demonstrate that a systematic approach to patients with nutrition risk factors is associated with a decreased short and long-term mortality.

Our data suggest that a systematic care for patients with nutrition related risk factors in hospitals is currently missing. Reception and dissemination of current recommendations from major clinical nutrition societies such as ESPEN and ASPEN requires further efforts to reach all the involved stakeholders, not only those directly involved in nutrition care but also those in management and decision making.

#### *4.2. Political Action Derived from Observations*

A structured approach to tackle the challenges of disease related malnutrition needs to include all involved stakeholders to implement universal screening for risk factors, apply systematically supportive nutrition care, monitor the effect in individual patients and take measures to promote continuity of care after discharge from acute care hospitals.

#### *4.3. Limitations of the Study*

A limitation of such a cross-sectional international data collection may be that clinical observations and their evaluation are not matching the same clinical pattern. Participants are provided with a document explaining all collected variables in English and with questionnaires which were translated and checked by clinicians in order to achieve as much as possible consistency between answers. Great care was taken to utilize a simple and self-explanatory wording. This was also essential because the patient questionnaires are intended to be answered by patients independently of their education level. For such factors as fluid status, overloaded or dehydrated, we had to accept the data as clinically evaluated.

Such large databases with participating centres from more than 60 countries carry also the risk of missing data and non-homogeneous data reporting. Over the years we found stable proportions of missing data in the individual risk indicators. Surprisingly, this missingness does not appear to be at random as missing data were often associated with a poor outcome and not with an intermediate outcome from the various strata of a variable. We have addressed this issue by including missing categories in all variables as recommended by the STROBE statement [23] and have performed in addition a sensitivity analysis with complete cases without missing data that yielded quite similar results.

In addition, large databases may identify risk indicators with a minimal clinical relevance despite a large univariate effect. We have, thus, selected to present multivariate analysis where spurious associations are less likely.

A further limitation is that cross-sectional observational data from a convenience sample are not appropriate to determine a causal association between risk indicators and outcome. Nevertheless only observational studies can identify risk indicators because risk indicators cannot be randomised (Table 5).


**Table 5.** Problem areas and suggested options for political action.

#### **5. Conclusions**

In summary the analysis and dissemination of the nutritionDay data helps to increase awareness of nutrition related issues in hospitals and provides discussion for potential political actions. There are many options for political action. The key points appear to be the raised awareness about all the described aspects related to DRM to all stakeholders, as well as the proposal of easy and standardized nutrition care processes, defined responsibilities within hospitals, and the establishment of adequate reimbursement schemes. Collection of data is crucial to allow monitoring of DRM as well as food provision at all levels and to allow benchmarking and discussion within the teams. DRM is unlikely to disappear from hospitals because disease acts as a strong driver towards malnutrition. Appropriate nutrition care should be continued after discharge to prevent further deterioration in patient's autonomy, quality of life and poor outcome. A systematic approach to DRM in hospitals and

an adequate continuity of care may lead to better outcomes [11,12,26]. nutritionDay in this regard has served and it will keep serving as a tool to monitor changes in clinical practice and associated outcome.

**Author Contributions:** Conceptualization: M.H., S.T., D.V., J.S. and K.S.; data curation: M.M. and C.S.; formal analysis: M.H. and I.S.; funding acquisition: M.H., S.T., D.V. and J.S.; investigation: S.M.; methodology: M.H., J.S. and K.S.; project administration: S.T., S.M. and D.V.; resources: M.H. and D.V.; software: M.M. and C.S.; supervision: A.L. and J.S.; validation: M.H. and I.S.; visualization: M.H., S.T. and I.S.; writing—original draft: M.H., S.T., D.V., J.S. and K.S.; writing—review and editing: M.H., S.T., A.L., I.S., D.V., J.S. and K.S.

**Funding:** This research was funded by The European Society for Clinical Nutrition and Metabolism (ESPEN) and supported by the Medical University of Vienna.

**Acknowledgments:** We acknowledge the nutritionDay national coordinators and the Austrian Society for Clinical Nutrition (AKE) for logistic support. We acknowledge all national coordinators contributing to nutritionDay: Jean-Charles Preiser and Brigitte Croix (Belgium), M. Cristina Gonzalez (Brazil), Luiza Kent-Smith (Canada), Angélica María Pérez Cano and Olga Lucía Pinzón Espitia (Colombia), Dina Ljubas (Croatia), Evelyn Frias Toral (Ecuador), Elina Ioannou (Cyprus), Frantisek Novak (Czech Republic), José Gutiérrez (El Salvador), Ulla Siljamäki-Ojansuu (Finland), Stéphane Schneider (France), Elke-Tatjana Schütz (Germany), Meropi Kontogianni and Kalliopi-Anna Poulia (Greece), Laszlo Harsányi (Hungary), Hartono (Indonesia), Pierre Singer (Israel), Marcello Maggio and Pietro Vecchiarelli (Italy), Hiroyoshi Takemoto (Japan), Laila Meija and Ilze Jagmane (Latvia), Gintautas Kekstas and Edita Gaveliene (Lithuania), Cora Jonkers-Schuitema (The Netherlands), Hugo Nilssen (Norway), Kinga Kupczyk (Poland), Paula Alves and Luis Matos (Portugal) Ioana Grintescu and Liliana Mirea (Romania), Vanessa Kotze (South Africa), Katja Kogovšek (Slovenia), Rosa Burgos (Spain), Barbara Hürlimann (Switzerland), Sonqsri Keawtanom (Thailand), Sadik Kilicturgay (Turkey), Sergii Dubrov and Olesia Gavrilenko (Ukraine), Kate Hall (United Kingdom) and Gail Gewirtz (USA).

**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**


© 2019 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* **Nutrition in Gastrointestinal Disease: Liver, Pancreatic, and Inflammatory Bowel Disease**

#### **Lena J. Storck 1,\*, Reinhard Imoberdorf <sup>1</sup> and Peter E. Ballmer 1,2**


Received: 13 June 2019; Accepted: 15 July 2019; Published: 25 July 2019

**Abstract:** Liver, pancreatic, and inflammatory bowel diseases are often associated with nutritional difficulties and necessitate an adequate nutritional therapy in order to support the medical treatment. As most patients with non-alcoholic fatty liver disease are overweight or obese, guidelines recommend weight loss and physical activity to improve liver enzymes and avoid liver cirrhosis. In contrast, patients with alcoholic steatohepatitis or liver cirrhosis have a substantial risk for protein depletion, trace elements deficiency, and thus malnutrition. Patients with chronic pancreatitis and patients with inflammatory bowel disease have a similar risk for malnutrition. Therefore, it clearly is important to screen these patients for malnutrition with established tools and initiate adequate nutritional therapy. If energy and protein intake are insufficient with regular meals, oral nutritional supplements or artificial nutrition, i.e., tube feeding or parenteral nutrition, should be used to avoid or treat malnutrition. However, the oral route should be preferred over enteral or parenteral nutrition. Acute liver failure and acute pancreatitis are emergencies, which require close monitoring for the treatment of metabolic disturbances. In most patients, energy and protein requirements are increased. In acute pancreatitis, the former recommendation of fasting is obsolete. Each disease is discussed in this manuscript and special recommendations are given according to the pathophysiology and clinical routine.

**Keywords:** monitoring; malnutrition; micronutrient deficiency; inflammation; oral nutritional supplements; artificial nutrition

#### **1. Introduction**

Gastrointestinal diseases are often associated with nutritional deficiencies. The complications range from digestive problems to nutrient absorption disorders and necessitate an adequate nutritional therapy in order to support the medical treatment. In this article, we focus on diseases of three important organs: Liver, pancreas, and intestine.

The liver is the main metabolic organ of the human organism because of its multiple functions; e.g., the liver controls the glucose homeostasis by regulation of glycogen synthesis, glycogenolysis, and gluconeogenesis. In addition, important body proteins and lipids are metabolized and synthesized in liver cells. Next to these metabolic pathways, the liver has an exocrine function by secretion of bile acids and detoxification of ammonia by urea production and glutamine synthesis. Taken these extensive functions into account, it is obvious that liver diseases and thus restrictions in liver functions have far-reaching consequences also for clinical nutrition. An important and by dietary habits increasing liver disease is non-alcoholic fatty liver disease (NAFLD), when the fat content of hepatocytes increases [1,2]. NAFLD aggregates benign hepatosteatosis, progressive and inflammatory non-alcoholic steatohepatitis (NASH) [3]. Insulin resistance, hyperinsulinemia, elevated plasma free fatty acids, fatty liver, hepatocyte injury, liver inflammation, oxidative stress, mitochondrial

dysfunction, imbalanced pro-inflammatory cytokines, and fibrosis characterize NAFLD [4]. The other main cause for the development of steatosis is alcohol abuse. Chronic high alcohol consumption causes reduced fatty-acid oxidation as well as increased triglyceride synthesis and deposition, and thereby supports the development of alcoholic steatohepatitis (ASH). Usually, steatohepatitis can be reversible if the relevant noxious agent is eliminated. Inadequate or lack of treatment over many years increases the risk for the development of a liver cirrhosis. Thus, 15–20% of patients with NAFLD will develop liver cirrhosis [5]. Next to these liver diseases that develop over many years, acute liver failure may occur usually as a medical emergency.

Another important organ, which can develop inflammatory processes, is the pancreas. This organ plays a central role in digestion due to its exocrine and endocrine function. A total of 98% of pancreatic tissue is part of the exocrine function, whereby pancreatic enzymes are secreted for food digestion. The endocrine part is located in the Langerhans cells, secreting insulin and glucagon for regulation of glucose homeostasis. Acute pancreatitis is a severe disease, causing self-digestion of the pancreas due to prematurely activated digestive enzymes. Patients with chronic pancreatitis have recurrent inflammatory episodes that replace the pancreatic parenchyma by fibrous connective tissue resulting in a progressive loss of exocrine and endocrine function. Characteristic complications and symptoms are pain, pseudocysts, and pancreatic duct stenosis [6].

The largest digestive organ is the intestine, which regulates the absorption of nutrients. Two important diseases of the intestine are Crohn's disease (CD) and ulcerative colitis (UC) that are characterized by periods of remission and inflammatory flare-ups [7]. In CD, the small and large intestine as well as the mouth, esophagus, stomach, and the anus can be affected with typical ulcerations that occur discontinuously. In contrast, UC mostly affects the colon and rectum and usually shows a continuous pattern of the mucosa. Despite multiple differences, these two diseases share similar symptoms such as abdominal pain, diarrhea, and malnutrition. Hence, these diseases were grouped in the term inflammatory bowel disease (IBD) [7,8].

Nutritional care is clearly important in the treatment of patients with IBD, pancreas or liver disease. Nutritional management includes prevention and/or treatment of malnutrition and micronutrient deficiencies as well as specific recommendations for each condition.

#### **2. Liver Diseases**

#### *2.1. Non-Alcoholic Fatty Liver Disease (NAFLD)*

So far, there are no approved pharmacological therapies for NAFLD [3]. One reason is that the pathophysiology of NAFLD is not yet fully understood despite enormous advance in this field of research. Since overnutrition is the key problem of NAFLD and most of the patients are overweight or obese, weight loss is an obvious therapeutic possibility, hence, intensive lifestyle interventions are well studied [4,5]. The guideline on clinical nutrition in liver disease recommend a 7–10% weight loss to improve steatosis and liver enzymes [3,9]. To improve fibrosis, a weight loss of more than 10% is necessary [10]. Lifestyle interventions including diet and physical activity should be the first-line treatment and only if all efforts fail, bariatric surgery should be proposed [11,12].

In addition, the composition of the diet may also have an effect on liver fat. Low-carbohydrate diet may be helpful with weight loss, but over a long time, a low-carbohydrate diet stimulated NAFLD pathogenesis in an animal model [4]. It is speculated that the carbohydrate's composition is crucial [4]. For example, fructose can easily induce metabolic complications in the liver and in contrast, fiber might be helpful to maintain blood glucose and thus prevent NAFLD. Furthermore, a fat-rich diet induces hepatic steatosis, but only saturated fatty acids are detrimental for the liver metabolism. Monounsaturated fatty acids might be beneficial and polyunsaturated fatty acids might even be a treatment option for NAFLD [4].

The evidence suggest that increased oxidative stress and changes in several molecular factors like pro- and anti-inflammatory cytokines are mainly involved in the progression of NAFLD [13]. Therefore, antioxidants like vitamin C or polyphenols (e.g., resveratrol, curcumin, quercetin, anthocyanin, green tea polyphenols) might have beneficial effects to improve NAFLD [4]. Kitade et al. (2017) showed that the dietary administration of the carotenoids β-crypoxanthin and astaxanthin not only prevents but also reverses NASH progression in mice by regulating M1/M2 macrophage/Kupffer cell polarization [13]. NAFLD induced by diets high in sucrose/fructose or fat can be prevented or improved by soy protein (β-Conglycinin) by decreasing the expression and function of the two nuclear receptors SREBP-1c and PPAR γ2. Fish oil with ω-3 fatty acids inhibits SREBP-1c activity, but controversially increases PPAR γ2 expression [13,14]. In summary, research results for the effects of micronutrients on NAFLD showed positive effects on some factors in liver metabolism. However, it remains unclear whether the use of antioxidants and ω-3 fatty acids improve liver disease, thus, further investigations are needed and no recommendation can be given [4]. In contrast and only in non-diabetic adults, vitamin E (800 IU α-tocopherol daily) can improve liver enzymes and histologic pathology [11,15].

The Mediterranean Diet, which is characterized by a high content of antioxidants and fiber, a balanced lipid profile and a low content of simple sugar, seems to be the optimal diet for the management of NAFLD. This diet is a natural multi-ingredient supplement that may exert its related health benefits by the synergistic and/or complementary action of each food compound [16]. A Mediterranean diet low in carbohydrates mobilizes more liver fat compared to a low fat diet with a similar weight loss [17].

#### *2.2. Alcoholic Steatohepatitis (ASH), Liver Cirrhosis, and Acute Liver Failure*

The nutritional recommendations for patients with ASH and liver cirrhosis are fundamentally different compared to the recommendations for patients with NAFLD, because these patients have a high risk for protein depletion, trace elements deficiency, and malnutrition. Twenty percent of patients with a compensated and 60% of patients with a decompensated liver disease are malnourished [18–20].

Therefore, screening for malnutrition is highly recommended on a regular basis in patients with ASH and liver cirrhosis, but the nutritional assessment can be difficult in patients with cirrhosis especially if there is associated fluid retention and/or obesity. Patients with cirrhosis may have a combination of loss of skeletal muscle and gain of adipose tissue, culminating in the condition of "sarcopenic obesity". In addition, patients had a loss or deficiency of several other nutrients such as vitamin D and zinc [21]. The Nutritional Risk Screening-2002 and the Malnutrition Universal Screening Tool are validated and well-known tools to screen patients at risk for malnutrition [22]. Specifically in patients with liver disease, the Royal Free Hospital Nutrition Prioritizing Tool (RFH-NPT) was developed and when compared to Nutritional Risk Screening, it was more sensitive to identify malnourished liver patients. Next to the important variables unplanned weight loss and reduced dietary intake, the RFH-NPT has additional score points for complications like fluid overload and diuretics [23,24]. In addition, the presence or absence of sarcopenia may be assessed with radiological methods, because sarcopenia is a strong predictor of mortality and morbidity in patients with liver disease [25,26]. Next to radiological methods, handgrip strength is a simple, objective, and practical method to assess sarcopenia [21].

In case of malnutrition, patients need extensive nutritional counseling and therapy. The treatment is a challenge as the nutritional problems are multifactorial [11]. Usually, the resting energy expenditure is increased [27,28] and patients are required to consume 35–40 kcal/kg body weight [23]. Non-malnourished patients should eat 1.2 g protein/kg body weight per day to cover the protein needs, whereas the optimal intake of malnourished and/or sarcopenic patients is 1.5 g protein/kg body weight [11,23]. If energy and protein intake are not adequate with regular meals, oral nutritional supplements (ONS), or artificial nutrition (tube feeding or parenteral) should be used to avoid or treat malnutrition. However, the oral route should be preferred over enteral or parenteral nutrition. The standard formulas should be used, preferably formulas with high energy density (≥1.5 kcal/mL). In case of tube feeding, a percutaneous endoscopic gastrostomy placement is not recommended and can only be used in exceptional cases because of a higher risk of complications, e.g., infections, ascites, or oesophageal varices [22].

Patients with ASH or liver cirrhosis have poor hepatic glycogen stores due to the impaired synthetic capacity of hepatic cells, hence, an overnight fast in these patients is equivalent to a nearly 72 hours fast in healthy persons. As a result, metabolism shifts to fatty acids as a dominant substrate for oxidation. Some tissues dependent on glucose will need neoglucogenesis from amino acids as fatty acids cannot be used for this process. This leads to mobilization of amino acids from the skeletal muscles so that the adequate amount of glucose can be produced. Repeated and frequent fasting results in recurrent proteolysis resulting in muscle loss in human cirrhotic patients [21,29]. Therefore, fasting periods should be avoided and patients with severe liver disease, who have to fast for more than twelve hours, should receive intravenous glucose (2–3 g/kg body weight). When the fasting period lasts longer than 72 hours, total parenteral nutrition may be required [11]. At home, periods of starvation should also be kept short and, therefore, the consumption of three to five meals a day and a late evening snack are recommended [30]. One complication of severe liver disease is hepatic encephalopathy, a disorder of the central nervous system with a wide spectrum, ranging from psychomotor impairments to coma [31]. In case of hepatic encephalopathy, protein intake should no longer be restricted in cirrhotic patients as it increases protein catabolism and thus promotes malnutrition [32]. Advanced cirrhosis patients could benefit from branched-chain amino acids (0.25 g/kg body weight) in order to improve event-free survival and quality of life [33].

Acute liver failure may occur as a medical emergency. In multi-organ failure, a severe derangement of the whole metabolism can occur in these patients due to loss of hepatocellular functions. The metabolic condition is characterized by impaired hepatic glucose production and lactate clearance as well as protein catabolism associated with hyperaminoacidemia and hyperammonemia [11,34,35]. In general, patients with acute liver failure should be treated in the same way as other critically ill patients. Therefore, the patients need to be monitored regularly and if necessary, macronutrients, vitamins, and trace elements should be supplemented [11]. On the one hand, the energy requirement of patients with acute liver disease is increased [36], but on the other hand, a hypercaloric diet induces hyperglycemia and hyperlipidemia. Therefore, it is important to achieve an iso-energetic diet in order to avoid malnutrition and complications. For metabolic monitoring, the following target values should be aimed at: Blood glucose 8–10 mmol/L, serum lactate <5 mmol/L, triglycerides <3 mmol/L, and ammonia <100 mmol/L. The protein requirement is set at 0.8–1.2 g/kg body weight [11]. A summary of nutritional recommendations in liver disease are presented in Figure 1.




**Figure 1.** Summary of nutritional recommendations in liver disease.

#### **3. Pancreatic Diseases**

#### *3.1. Acute Pancreatitis*

The degree of the inflammatory response of the pancreas plays an important role for the assessment of clinical nutrition. Normally, patients with a mild or moderately severe pancreatitis do not need a specific nutritional? intervention, but every patient with acute pancreatitis should be screened for malnutrition with the regular tools and in case of malnutrition, receive an adequate nutritional therapy [37]. The energy and protein requirements are usually not increased and the patients are allowed to eat normal food independently of lipase and amylase activity [37,38]. In contrast to former medical opinions, fasting after an inflammatory episode has no positive effect on the clinical course or prognosis of acute pancreatitis. During the course of pancreatitis, exocrine secretion is blocked and therefore a stimulation of the exocrine function by food intake or artificial nutrition is not expected [6,37]. For those patients who can tolerate an oral diet, an initial low-fat solid diet is preferred [39,40]. This early approach to oral feeding may reduce the length of hospital stay in these patients [41].

In contrast, patients with severe necrotizing pancreatitis need an adequate clinical nutritional? strategy. Initial short-time fasting could be beneficial for patients with ileus or nausea and vomiting, but within 24–48 hours, enteral nutrition should be started. Early enteral nutrition has a more preventive then a nutritive effect, because early enteral nutrition decreased mortality and complications [42–44]. Parenteral nutrition instead could result in an intestinal villous atrophy within a few days, which then facilitates bacterial translocation and may result in severe infections. The administration of enteral nutrition counteracts translocation, and therefore enteral nutrition should be administered whenever possible to prevent intestinal atrophy [6,45].

Due to inflammation and pain, patients with severe acute pancreatitis have an increased energy and protein requirement [46]. Energy requirement is set at 25–30 kcal/kg body weight, glucose intake at 2–4 g/kg body weight, and protein at 1.2–1.5 g/kg body weight. Infusion rate for lipids should be 0.8–1.5 g/kg body weight and the infusion should be monitored regularly since triglycerides in plasma should be <12 mmol/L. Trace elements are supplemented in normal concentrations and high-dose vitamin supplements are not required [37]. In addition, administration of pre- and probiotics as well as immunonutrition cannot be recommended since studies so far have not shown a clear beneficial effect [47,48]. To reach the nutritional goal, enteral nutrition can be completed with oral food intake or parenteral nutrition if necessary. In case of total parenteral nutrition over a long time period, administration of 0.2–0.5 g glutamine/kg body weight could protect against infections and reduce mortality [49]. In addition, monitoring of volume substitution is very important and can decrease mortality [6]. An early return to normal food intake should be pursued [6]. After an episode of acute pancreatitis, approximately 20% of the patients develop the common complications diabetes and exocrine pancreatic insufficiency [50]. Therefore, patients should be regularly screened for these complications in order to prevent nutritional status.

#### *3.2. Chronic Pancreatitis*

Patients with chronic pancreatitis have a high risk for malnutrition due to maldigestion from pancreatic exocrine insufficiency in combination with inflammation and increased energy metabolism. Pancreatic exocrine insufficiency is the main pancreatic cause of malnutrition in these patients [51]. In addition, due to diarrhea and steatorrhea, patients with chronic pancreatitis often have a deficiency of the fat-soluble vitamins A, D, E, and K. Malnutrition is associated with an increased complication rate and increased mortality [52]. Therefore, malnutrition should be avoided using nutritional counseling and if necessary artificial nutrition. Energy requirement is set at 25–30 kcal/kg body weight and protein requirement at 1.5 g protein/kg body weight. Since alcohol is an important cause for chronic pancreatitis, patients should avoid alcohol completely [6,37,53].

In the case of pancreatic exocrine insufficiency, patients should be supplemented with pancreas enzymes. The exocrine insufficiency is diagnosed if the patients have either steatorrhea (>15 g/day fecal fats) or suffer from manifest maldigestion, respective malabsorption. For a first indication, patients can be asked whether they see undigested food in their feces or if the feces are difficult to wash away. Further examination could be a pancreatic function test and the evaluation of maldigestion-related symptoms like diarrhea or flatulence, poor nutritional status, and fecal elastase-1 concentration [51,52]. In patients with an exocrine insufficiency, low levels of circulation fat-soluble vitamins, proteins like albumin, lipoproteins, apolipoproteins, and mineral trace elements like magnesium, zinc, or calcium, can occur [54–57]. Since most of these abnormalities are related to pancreatic exocrine insufficiency, a laboratory analysis might be helpful in order to diagnose the insufficiency. However, other factors like toxic habits or deficient food intake may play a relevant role, too [51]. Pancreas enzymes should be taken during the meals and the dosage is based on lipase activity. A range of 20.000 to 40.000 lipase units per main meal should be administered as an initial dose and 10.000 to 20.000 units for the digestion of smaller in-between snacks [6,58]. It is not necessary to avoid high fat intake if the exocrine pancreas function is compensated [37]. The replacement therapy should promote digestion, but a complete normalization of digestion is usually not achieved. However, there are several options for a failure of normalization. First, the pH value in the stomach inactivates the pancreatic enzymes. Therefore, the addition of a proton-pump inhibitor before breakfast and dinner is recommended in cases of an unsatisfactory clinical response to the standard dose of pancreatic enzymes [59]. Second, the enzyme dose should be increased if needed to normalize digestion and the nutritional status of the patients. If these two strategies fail, another cause for maldigestion like bacterial overgrowth should be evaluated [51]. A summary of nutritional recommendations in pancreatic disease is shown in Figure 2.



**Figure 2.** Summary of nutritional recommendations in pancreatic disease.

#### **4. Inflammatory Bowel Diseases (IBD)**

IBD is a heterogeneous and multifactorial disorder resulting from a complex interplay between genetic variation, intestinal microbiota, the host immune system, and environmental factors such as diet, drugs, breastfeeding, and smoking [60–62]. The relationship between dietary nutrients and intestinal homeostasis is complex and influenced by several interactions between host immune system, the intestinal barrier, and the gut microbiota [60,63]. Patients with IBD have a high risk for malnutrition, which may be the result of reduced oral intake, increased nutrient requirement, increased gastrointestinal losses of nutrients, and occasionally from drug–nutrient interactions [7,64]. In pediatric patients, malnutrition is the main cause for growth retardation [65,66]. In CD, malnutrition is a great problem compared to UC because any part of the gastrointestinal tract can be affected. Therefore, the risk for malnutrition remains even when the disease is quiescent. UC is normally restricted to the colon and hence shows few malabsorptive problems except in active disease [64]. Due to the high risk, patients with IBD should be screened for malnutrition using the established tools at the time of diagnosis and thereafter on a regular basis [67,68]. Malnourished patients should receive an adequate nutritional therapy because otherwise it worsens the prognosis, complication rates, mortality, and quality of life [67,69,70]. The energy requirement of patients with IBD are normally not increased as well as the protein requirement in remission. In contrast, protein requirement is increased in active IBD and therefore, the protein intake should be 1.2–1.5 g/kg body weight [67,71,72].

In addition, patients with IBD have a high risk for micronutrient deficiencies due to losses from diarrhea and/or inadequate dietary intake. The most common micronutrient deficiencies are iron, calcium, selenium, zinc, and magnesium depletion. Vitamin deficiencies include all vitamins and in particular B12, folic acid, and vitamins A, D, and K [73,74]. For example, selenium, zinc, and magnesium depletions are caused by an inadequate dietary intake and chronic loss because of diarrhea. Symptoms associated with deficiencies include bone health impairment, fatigue, poor wound healing, and cartilage degeneration [73,74]. An example for the influence of medication is cholestyramine that can interfere with absorption of fat-soluble vitamins, iron, and B12 vitamin. Main side effect is steatorrhea due to impair absorption of fats [75]. Therefore, laboratory values of patients should be checked on a regular basis and possible deficits should be appropriately corrected.

The most frequent extraintestinal manifestation of IBD is iron-deficiency and anemia, which occurs more frequently in CD and which should be supplemented with iron. Anemia is usually associated with other important symptoms like fatigue, sleeping disorders, restless legs syndrome, or attention deficit [75]. Patients with mild anemia can receive oral iron, when they are tolerant for oral iron and when the disease is inactive. Intravenous iron should be considered in patients with active IBD, with previous intolerance to oral iron, with hemoglobin below 100 g/L, and in patients who need erythropoiesis-stimulating agents [67,76,77]. Furthermore, patients may have deficiencies of calcium, vitamin D, folate, vitamin B12, and zinc [64]. When more than 20 cm of the distal ileum is resected, vitamin B12 must be administered to patients [67].

Low levels of calcium and vitamin D are common in patients with IBD, especially in those with duodenal and jejunal disease [61,73,74]. Calcium deficiency is linked to vitamin D deficiency, which is related to inadequate daily intake, inflammation status, diarrhea, and glucocorticoid therapy. The prevalence among IBD patients is up to 70% in CD patients and up to 40% in UC patients. Nonetheless, it is not established if the vitamin D deficiency is a cause or a consequence of IBD. However, there are suggestions that in genetically predisposed individuals, vitamin D deficiency may be a contributing factor in the development of IBD [78]. Beneficial effects of vitamin D in IBD are supported by pre-clinical studies mainly in mouse models, where the active form of vitamin D has shown to regulate gastrointestinal microbiota function and promote anti-inflammatory response [61].

When oral nutritional intake is insufficient during active disease, ONS are the first step. If oral feeding is not sufficient, tube feeding is superior to parenteral feeding. Parenteral nutrition is indicated in IBD (i) when oral or tube feeding is not sufficiently possible, (ii) when there is an obstructed bowel, where there is no possibility of placement of a feeding tube beyond the obstruction or where this has failed, or (iii) when other complications occur such as an anastomotic leakage or a high-output intestinal fistula [67]. During active disease, specific formulations or substrates, e.g., glutamine, ω-3 fatty acids are not recommended, neither is the use of probiotics. Probiotic therapy using Escherichia coli Nissle 1917 or VSL#3 can be considered for use in patients with mild to moderate UC for the induction of remission [79].

Enteral exclusive nutrition has been extensively used for induction of remission in pediatric CD, in which avoidance of steroids is critical for childhood growth. Several recent pediatric studies have demonstrated and confirmed that enteral exclusive nutrition can induce remission in 60–86% of children [80–82] and is associated with higher remission rates, better growth, and longer steroids-free periods [60]. However, the benefit was lost when partial enteral nutrition was used with access to a free diet [83]. Nonetheless, a recent study has shown that partial enteral nutrition can be effective for induction of remission in children and young adults in combination with a diet, which is based on components hypothesized to affect the microbiome or intestinal permeability [84]. In remission, ONS or artificial nutrition are only recommended if malnutrition cannot be treated sufficiently by dietary counseling. In addition, specific diets or supplementations with ω-3 fatty acids are not

recommended for maintenance of remission. A systematic review has not supported the hypothesis that supplementation of ω-3 fatty acids can induce and maintain remission in IBD [85]. However, several studies have demonstrated that different genotypes can be associated with the variable response to nutritional intervention with ω-3 fatty acids [86]. Probiotic therapy can be considered in UC but not CD for maintenance of remission [67]. In addition, clinical trials show that curcumin supplementation might be effective for the induction and maintenance of remission in UC patients [87,88]. Curcumin suppresses cytokine production by macrophages and intestinal epithelial cells via the inhibition of NF-kB activation [89,90] and thus mitigates induced colitis in mouse models [91,92]. A summary of nutritional recommendations in IBD is presented in Figure 3.


**Figure 3.** Summary of nutritional recommendations in inflammatory bowel disease.

#### **5. Future Perspectives**

Clinical nutrition is clearly important in the treatment of patients with liver, pancreatic, and inflammatory bowel disease. However, there are many gaps of knowledge in the pathophysiology of these diseases despite an enormous research effort and thus the effects from clinical nutrition cannot be at a maximum so far.

A major gap in knowledge is associated with the evolution of NAFLD in children and adolescents. Weight loss is important for histological improvement and the benefits of weight loss will extend beyond those expected from drug treatment of high-risk NASH. It is postulated that only public health strategies will have the opportunity to improve the burden of obesity-related diseases in the future [93]. In addition, due to our aging population, it is expected that the burden of liver disease due to cirrhosis from NASH will increase over the next two decades unless effective preventive and therapeutic interventions are implemented as part of a public health strategy [93].

Furthermore, knowledge about the human microbiome constantly increases. Trillions of microbial cells, which form a symbiotic relationship with the host, play a crucial role in the development of diseases, when the balance of the microbiome becomes disrupted [94]. Increase intestinal permeability and dysbiosis are common characteristics linking the liver to a number of gastrointestinal diseases [95]. For example, alcohol consumption and endogenous alcohol production by gut bacteria in obese individuals can disrupt the tight junctions of the intestinal epithelial barrier, resulting in increased gut permeability. The bacterial endotoxin in the portal circulation leads to liver inflammation and fibrosis through activation of toll-like receptor 4 [95]. Although several mechanisms by which the microbiome might affect liver disease have been proposed, more work is needed to fully understand these relationship [93].

IBD is also associated with a dysbiotic microbiome. However, it is unclear if this dysbiosis plays a role in the pathophysiology or is a result of the disease. Due to the importance of the microbiome in IBD, therapies manipulating the microbiome have gained popularity. While there are some promising trials demonstrating the efficacy of antibiotic combinations in treating IBD, more controlled trials are required [94]. In addition, a more precise understanding of the complex interrelation between dietary nutrients, host immunity, and the microbiome is necessary to increase the effectiveness of dietary interventions used to treat IBD [63]. Moreover, an in-depth knowledge of the genetic background is important for personalized nutritional management, which might lead to a maximum efficacy in therapy [63].

New trials in acute pancreatitis showed that approximately one-third of patients will develop prediabetes or diabetes within five years of an index episode of acute pancreatitis and 24–40% of the patients develop an exocrine pancreatic insufficiency, but the mechanisms and risk factors remain to be specified [96–99]. Furthermore, further evidence is required to determine the optimal panel of laboratory markers for nutritional evaluation in chronic pancreatitis, and the utility, reliability, and accuracy of these markers in diagnosing pancreatic exocrine insufficiency [51].

**Author Contributions:** L.J.S., R.I. and P.E.B. contributed to the conceptualization of the manuscript. L.J.S. wrote the original draft. R.I. and P.E.B. reviewed and edited the original draft. All authors contributed to the final version of the manuscript.

**Conflicts of Interest:** There are no conflicts of interest to declare.

#### **References**


© 2019 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* **Economic Challenges in Nutritional Management**

### **Emilie Reber 1,\*,**†**, Kristina Norman 2,3,**†**, Olga Endrich 4, Philipp Schuetz 5, Andreas Frei <sup>6</sup> and Zeno Stanga <sup>1</sup>**


Received: 31 May 2019; Accepted: 4 July 2019; Published: 10 July 2019

**Abstract:** Disease-related malnutrition (DRM) is a highly prevalent independent risk and cost factor with significant influence on mortality, morbidity, length of hospital stay (LOS), functional impairment and quality of life. The aim of our research was to estimate the economic impact of the introduction of routinely performed nutritional screening (NS) in a tertiary hospital, with subsequent nutritional interventions (NI) in patients with potential or manifest DRM. Economic impact analysis of natural detection of inpatients at risk and estimation of the change in economic activity after the implementation of a systematic NS were performed. The reference population for natural detection of DRM is about 20,000 inpatients per year. Based on current data, DRM prevalence is estimated at 20%, so 4000 patients with potential and manifest DRM should be detected. The NI costs were estimated at CHF 0.693 million, with savings of CHF 1.582 million (LOS reduction) and CHF 0.806 million in additional revenue (SwissDRG system). Thus, the introduction of routine NS generates additional costs of CHF 1.181 million that are compensated by additional savings of CHF 2.043 million and an excess in additional revenue of CHF 2.071 million. NS with subsequent adequate nutritional intervention shows an economic potential for hospitals.

**Keywords:** economic challenges; nutritional management; malnutrition

#### **1. Introduction**

Disease-related malnutrition (DRM) is a debilitating, important and frequently occurring problem with an estimated prevalence of 20–50% on hospital admission [1–5]. The consequences of DRM are well known: Increases in morbidity, complication and mortality rates, resource use for inpatient treatment, prolonged hospital length of stay (LOS), decreased quality of life (QoL) and decreased body function. Many studies have shown the positive effects of nutritional interventions (NI), mainly in the reduction of complication rates, LOS, or rates of non-elective re-hospitalizations [3,6–10]. The recently published study by Schuetz et al.—a multicenter study with 2088 medical patients at nutritional risk, from eight Swiss hospitals—showed a significant reduction in serious complications and mortality as well as an improvement in physical function and quality of life after 30 days [11]. The study by Deutz et al. also showed a significant reduction in mortality after 90 days [12]. Available studies on

cost-effectiveness clearly show that NI can save money and that the costs of NI are more than offset by the savings [7,9,13].

Since 2012, hospitals in Switzerland are remunerated using per-case rates for inpatients through the Swiss Diagnosis Related Groups (SwissDRG) system. In the SwissDRG it is possible to take DRM into account by coding it as a principal or secondary diagnosis. This requires optimal nutritional management from the hospitals, with an initial screening and adequate therapy in the course of treatment. Coding DRM may have the effect that a certain percentage of the patients will be assigned to a DRG with a higher cost-weight, generating additional revenue for the hospital (like a bonus system). Aeberhard et al. investigated the financial effects of coding DRM in the SwissDRG system, including all inpatients from the years 2013 to 2016 in our university hospital. During the observation period, 3.2% of the patients were coded with DRM. In 8.3% of these cases, the coding led to the attribution of a DRG with an increased cost-weight. This resulted in total additional revenue of circa CHF 3.5 million, which was offset by costs of CHF 2.8 million for assessment and treatment of DRM [1]. Thus, the costs of screening and treatment of DRM were already overcompensated for as a result of the DRM coding and changes in SwissDRG attribution. These results have been confirmed in similar studies [14,15].

As regularly performed nutritional risk screening is not established at the Bern University Hospital, these results reflect the natural detection (ND) of DRM. If nutritional screening (NS) is performed routinely on all patients, the detection rate of DRM will likely increase, leading to earlier detection and treatment. Performing routine screening will generate further costs. Besides that, the same costs for each patient at nutritional risk occurs when the NS is performed, as in the natural detection of DRM. However, due to an increased detection rate, the number of patients with treated and coded DRM will rise, with NI costs increasing as a consequence, but with additional revenues generated [16,17]. If DRM is detected and adequately treated, three types of consequences will occur:


Previous economic analyses have always compared either the costs of NI with savings in the costs of basic nutritional treatments [3,7,13] or with the additional revenues due to DRG changes [14,16,17]. No study is known to have compared costs with a combination of both savings and additional revenue.

The objective of this study was to estimate the economic impact of the introduction of a systematically performed NS on all patients in our hospital, with subsequent NI in all patients at nutritional risk.

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

The data for this economic impact analysis were collected from the electronic patient record system of the Bern University Hospital in Bern, Switzerland, between 1 January and 31 December 2018. This is a cost-minimization analysis of the short-term economic consequences of inpatient hospitalization for the year 2018. Included were direct medical costs for NS and NI, cost savings due to NI, and additional revenues due to DRM coding and DRG changes. Using a population-based approach, the economic effects of a systematically performed NS were projected and compared with the existing situation reflecting the natural detection of DRM and routine nutritional management. To this end, a flow chart was created, and the patient flows and resource use were assessed. The costs of ND were subtracted from those of a systematic NS. Furthermore, the number of additional staff needed to perform routine NS was estimated. Data were obtained through analyses of the literature, the use of secondary statistics, and data collection at the Bern University Hospital. In cases of unclear or controversial data, assumptions were made. The flow chart is presented in Figure 1.

**Figure 1.** Nutritional screening (NS) and natural detection (ND). DRM: Disease-related malnutrition; p DRM: potential DRM; m DRM: manifest DRM; ONS: oral nutritional supplements; EN: enteral nutrition; PN: parenteral nutrition.

In a systematic NS, all patients are screened on admission to the hospital using a validated tool—in our case, the Nutritional Risk Screening 2002 (NRS 2002) [18]. Patients with an NRS score of 2 points or less are likely to be unaffected. Patients with an NRS score of 3 are considered at risk of DRM, and those with a score >3 points are considered to have manifest DRM. Patients with potential DRM undergo NS once per week. Patients with potential DRM will receive oral nutritional supplements (ONS) in addition to the customary nutrition and will also undergo NS once per week. Patients with manifest DRM will receive specific NI such as ONS, enteral nutrition (EN), or parenteral nutrition (PN) as needed and indicated in addition to the customary hospital nutrition. Their nutritional status will be regularly observed so that in these patients no additional NS will be needed [19].

#### **3. Input Data**

In Table 1 the input data are summarized.



DRM: disease-related malnutrition; p DRM: potential DRM; m DRM: manifest DRM; ONS: oral nutritional supplements; EN: enteral nutrition; PN: parenteral nutrition; LOS: length of hospital stay; SwissDRG: Swiss diagnosis-related group, cw: cost-weight.

#### *3.1. Target Population*

For a systematic NS with subsequent NI, patients with an expected LOS of >3 days were considered eligible. In 2018, this amounted to 21,819 patients (information from Bern University Hospital). The average LOS of these patients was 10.01 days. This number also included healthy newborns. Therefore, it was assumed that the target population for a systematic NS at the Bern University Hospital would amount to approximately 20,000 patients yearly. Even though Aeberhard et al. [1] included all inpatients, it is unlikely that DRM was detected in patients with a LOS ≤3. Therefore, the reference population for ND patients with DRM was also assumed to be about 20,000 patients per year.

#### *3.2. Detection Rate*

Detection rates of 19% and 25.6% were obtained from two German studies [16,17]. The latter, however, also included patients with NRS ≥2, and the rate of patients with NRS >3 was computed at 20.7%. A Dutch study found that NS could increase the detection rate of DRM from 50% to 80% given a DRM prevalence of 32% [20]. Based on these data, the detection rate was estimated at 20%. For natural detection, we doubled the rate of 3.2% from Aeberhard et al. [1], using 6.4%.

#### *3.3. Proportions with Potential and Manifest DRM*

In patients with NS, the proportion of potential versus manifest DRM was estimated at 45%:55% based on [17]. In the naturally detected patients, these proportions were assumed to be 25%:75% as in [1].

#### *3.4. Number of Screenings*

Patients potentially having DRM are screened on admission and then once weekly. Patients with manifest DRM are only screened once on admission [19].

The number of screenings therefore depends on the proportions of patients with potential and manifest DRM and their expected LOS. Two studies showed that LOS in patients with DRM is longer than in well-nourished patients (11 vs. 7 days [16], 14 vs. 7.6 days [17]). Thus, it was assumed that LOS in patients without DRM was one week or less, and that LOS detected in DRM patients detected by NS was 12–14 days (1–2 weeks). Therefore, it was assumed that patients without DRM receive one NS, patients with potential DRM receive two NS, and patients with manifest DRM receive one.

#### *3.5. Types and Proportions of NI*

We distinguish between three types of NI: ONS, EN, and PN. In the study by Aeberhard et al., 59% of ND patients received ONS, 29% received EN, and 12% received PN [1]. All patients with potential DRM received ONS. Of the 59% of patients prescribed ONS, 25% were patients with potential DRM and 34% were patients with manifest DRM. The proportions of potential and manifest DRM differ in patients with NS. This implies that 45% with potential DRM will be prescribed ONS. In addition, of the 55% with manifest DRM, 25% will receive ONS, 21% will receive EN, and 9% will receive PN.

#### *3.6. Reduction of LOS*

Available data suggest that hospital LOS can be reduced by NI. However, it is difficult to quantify the amount. It seems that this effect is more likely to occur in patients with manifest DRM than with potential DRM and that it is more pronounced in connection with severe DRM [5,20–22]. Based on the findings of Elia et al. (−13.8% in relation to 22.5 hospital days) [3], Sriram et al. (−10% in relation to 6 days) [5] and Bally et al. (0%, calculated −3.2% of 13 days, not significant) [6], it was assumed here that NS will reduce LOS in patients with manifest DRM by 10%. This results in a reduction in LOS of 1.2 days. The same assumption is also made for ND patients.

#### *3.7. DRG Changes*

In ND, coding of a case of DRM led to a DRG change in 8.3% of all coded cases [1]. In patients with NS this proportion was substantially higher, amounting to 27% [16] and 15% [17]. Therefore, it was conservatively estimated to be 15%.

#### *3.8. Increase in Cost-Weight*

The average amount of additional revenue per case with a DRG change amounted to CHF 7564 [1]. Given a base rate of CHF 10,900, this corresponds to an average increase in the cost-weight of 0.694. However, in patients with NS, the average increase was clearly smaller (0.44) [16,17].

#### *3.9. Ethics*

This study was conducted in accordance with the ethical guidelines of the 1957 Declaration of Helsinki and approved by the Bernese Cantonal Ethics committee (BASEC ID 2017-00480), Bern, Switzerland.

#### **4. Cost and Savings**

#### *4.1. Cost of NS*

The costs of NS were calculated based on the time needed in minutes multiplied by the cost per minute of nursing staff. According to Wenger et al., the time needed to administer NS is about 5 min [19]. The costs per minute were calculated from an hourly rate (gross wage including employers' contributions to social insurance, information University Hospital Bern) and amounted to CHF 3.93.

#### *4.2. Costs per Patient with NI*

Therapy costs per patient for NI patients with DRM included daily personnel and materials costs multiplied by the duration of NI in days. In the case of EN and PN, one-time costs per therapy were added (Table 2). These values were based on Aeberhard et al. [1] and updated for 2019.

For ONS, the time expended by staff members was 10 min for a nutritional therapist, 10 min for nursing staff, and 2 min for a physician; for EN it was 10 min for a nutritional therapist, 40 min for nursing staff, and 2 min for a physician; and for PN it was 12 min for a nutritional therapist, 70 min for nursing staff, and 2.4 min for physicians. The costs were CHF 41.40 per hour for a nutritional therapist, CHF 47.11 per hour for nursing staff, and CHF 78.24 per hour for physicians (information provided by the Bern University Hospital). Data on materials costs, duration of NI and one-time costs were obtained from Aeberhard et al. [1]. Thus, the costs per patient for NI were calculated as CHF 187.57 for ONS, CHF 842.96 for EN, and CHF 1557.84 for PN.


**Table 2.** Costs of nutritional interventions (NI) per patient and intervention.

ONS: oral nutritional supplements; EN: enteral nutrition; PN: parenteral nutrition.

#### *4.3. Savings per Prevented Inpatient Day*

It was assumed that the reduction of LOS occurred at the end of the hospitalization period. The costs then were mainly related to accommodation, food, medical care, follow-up visits, hospital buildings and the like. Such LOS-dependent costs are also used for the remuneration of hospital services in DRG outliers. In such cases, in addition to the per-case rate, the hospital is reimbursed a daily rate for each day exceeding the upper trim point delimiting inlier LOS. These daily rates are based on cost-weights per day. The cost-weight per day was calculated at 0.126 as a weighted average across the SwissDRGs based on Aeberhard et al. [1]. This was multiplied by the base rate of the Bern University Hospital (CHF 10,900). Thus, the cost per prevented hospital day was estimated at CHF 1373.

#### *4.4. Additional Revenue due to SwissDRG Change*

The number of patients with SwissDRG changes were multiplied by the average increases in the cost-weights and the CHF 10,900 base rate of the Bern University Hospital.

#### **5. Results**

#### *5.1. E*ff*ect of a Systematic NS*

The patient flow and performance of systematic NS are summarized in Table 3.



DRM: disease-related malnutrition; p DRM: potential DRM; m DRM: manifest DRM; ONS: oral nutritional supplements; EN: enteral nutrition; PN: parenteral nutrition; LOS: length of hospital stay; DRG: diagnosis-related groups.

For a systematic NS, 20,000 patients per year would have to be screened on hospital admission. Assuming a detection rate of 20%, 4000 patients (including potential and manifest DRM) would be

detected. Of these, 45% (1800) will have potential DRM and 55% (2200) will have manifest DRM. Patients with potential DRM and a projected LOS of 12 days will experience a second instance of NS in the second week of their hospital stay. So, in total, 21,800 NS will be performed. The 1800 patients with potential DRM and an absolute 24.9% of the 2200 patients with manifest DRM (=997 patients) will receive ONS. Furthermore, 21.3% (=851 patients) will receive EN, and 8.8% (=352 patients) will receive PN. These NI will effect a reduction of LOS of 1.2 days per case in patients with manifest DRM. This will result in global savings of 2640 hospital days in 2200 patients.

In all 4000 DRM patients, DRM will be coded as a complication or comorbidity in the DRG system. This coding will cause a DRG change in 15% of all DRM coded cases (i.e., 600 patients). The resulting costs are summarized in Table 4.

A total of 21,800 NS will be performed per year at a cost of CHF 3.93 per unit of NS. Thus, the costs of NS will amount to CHF 85,583. The costs of NI are calculated in a similar way using the number of patients multiplied by the NI costs per patient. The costs for all 2797 patients with ONS (1800 with potential, 997 with manifest DRM) were calculated as CHF 524,694, the costs of the 851 patients with EN as CHF 717,076, and those of the 352 patients with PN as CHF 548,358. Overall, the costs of NI amount to CHF 1,790,128. Combined with the costs of NS, global costs amount to 1,875,711. These costs are balanced by savings of CHF 3,625,930 due to the reduction of LOS. Furthermore, additional revenue of CHF 2,877,600 results from DRG changes. So, after deduction of the costs, there is a net monetary gain of CHF 4,627,818 for the hospital.


**Table 4.** Costs resulting from a systematic NS.

Costs are indicated with positive (+), savings and additional revenue with negative (−) prefix. DRM: disease-related malnutrition; p DRM: potential DRM; m DRM: manifest DRM; ONS: oral nutritional supplements; EN: enteral nutrition; PN: parenteral nutrition.

#### *5.2. E*ff*ects of ND, Treatment and Coding of DRM*

Table A1 (Appendix A) shows the patient flow in the case of natural detection of DRM. In relation to the target population of 20,000 patients, the detection rate is 6.4%, which results in 1280 patients per year with DRM detected, treated and coded. Of these, 25% (320 patients) will be patients with potential DRM and 75% (960 patients) will be with manifest DRM. ONS will be provided to 755 patients (59%), EN to 371 patients (29%), and PN to 154 patients (12%).

It can be assumed that the average LOS in patients with naturally detected DRM would have been longer if they had not been treated for DRM and if NI had not reduced this LOS in the 960 patients with manifest DRM by 1.2 days. This amounts to a savings of 1152 hospital days. In 8.33% of the patients, i.e., in 107 cases, there is a DRG change.

Table A2 (Appendix A) shows the costs related to natural detection of DRM. The costs of NI amounts to CHF 141,652 for the treatment of 755 patients with ONS, CHF 312,906 for the 371 patients with EN, and CHF 239,284 for the 154 patients with PN. Total costs amount to CHF 693,842 per year. The savings due to a reduction of LOS amounted to CHF 1,582,224. The additional revenue resulting

from SwissDRG changes amounts to CHF 806,579. The costs are overcompensated for by the savings and the additional revenue, resulting in an overall net savings of CHF 1,694,961.

#### *5.3. Costs, Savings, and Additional Revenue Attributable to NS*

As compared with the actual state of a natural detection, a systematic NS would generate additional costs of CHF 85,583 for screening and CHF 1,096,286 for NI, totaling CHF 1,181,869. These would be compensated for by savings of CHF 2,043,706 due to a reduction in LOS and by additional revenue of CHF 2,071,021 arising from coding and DRG changes (Table 5).


**Table 5.** Net monetary effects of the introduction of a systematic NS.

#### *5.4. Sta*ff *Needed*

Table A3 shows (Appendix A) the calculation of the staff needed, broken down by professional group. The total number of working hours and days was based on the time needed per day and the duration of the therapy in days (see Table 1) and on the number of patients undergoing therapy. Details are given in Table A4 (Appendix A). The number of positions was calculated using an average working day of 8.4 h, 220 working days per year and a fulltime position, and a productivity of 80%.

For a systematic NS, 4.09 positions for nutritional therapists, 9.27 positions for nursing staff, and 0.82 positions for physicians is needed. The treatment of naturally detected patients with DRM presently requires 1.35 positions for nutritional therapists, 3.6 positions for nursing staff, and 0.27 positions for physicians. Thus, the additional staff needed amounts to 2.75 positions for nutritional therapists, 5.66 positions for nursing staff and 0.55 positions for physicians.

#### *5.5. Scenario Analyses*

The results show that the savings and the additional revenue are each separately greater than the costs. Scenario analyses were performed based on few important assumptions to determine the minimum savings or additional revenues needed in order to cover only the costs. To find a best estimate for the reduction of LOS, the value of a saved hospital day, the proportion of DRG changes, and the average increases in the cost-weights, the following scenario analyses were performed:

Analysis 1, cost consequences in case the savings = 0, shows that even if no savings could be realized, NS could still achieve a net benefit through the additional revenue of CHF 1,001,889.

Analysis 2, cost consequences if additional revenue = 0, reveals that, even if no additional revenue were realized, NS would still lead to net savings of CHF 1,750,218.

Analysis 3, minimal additional revenue needed if savings = 0, shows that, in order to compensate for the costs, additional revenue of CHF 1,875,711 is needed. This could be achieved (1) by reducing the proportion of cases with changes in DRG attribution to 9.8% while keeping the increase in the cost-weight constant, or (2) by leaving the percentage of cases with changes in DRG attribution unchanged while increasing the average cost-weight by only 0.29.

Analysis 4 shows that a 5% reduction in LOS with a constant value per day would be sufficient to achieve the minimum required savings of CHF 1,875,711. On the other hand, if the reduction of LOS were left unchanged, a valuation of CHF 710 per saved hospital day would be sufficient.

#### **6. Discussion**

The introduction of a systematic NS would be accompanied by yearly costs of CHF 1.875 million. These costs would be compensated for by savings of CHF 3.635 million from a reduction in LOS and by additional revenue of CHF 2.877 million due to changes in documentation and coding of DRM. These numbers include the costs, savings and additional revenue produced by natural detection of DRM. These were estimated at CHF 0.693 million for the costs, CHF 1.582 million for the savings, and CHF 0.806 million for the additional revenue. Thus, the introduction of a systematic NS would generate additional costs of CHF 1.181 million that would be compensated for by additional savings of CHF 2.043 million and additional revenue of CHF 2.071 million.

These figures are based on projections using data from literature analyses, use of secondary statistics and data collected from the Bern University Hospital. It is notable that the savings per se are almost twice the costs, and the additional revenue alone is about 1.5 times the costs. These results are based on a few central assumptions about the data collected. The most important of these are discussed here.

The detection rate of NS was estimated to be 20%. Compared with data from the literature showing prevalence rates of 20–50% and even greater, this is a cautious assumption. The assumption regarding the size of the target population seems well justified, as only patients with an expected LOS of >3 days were considered. In estimating the expected LOS of patients with DRM, neither the average LOS (barely 6 days) of all patients from the Bern University Hospital nor the LOS (20 days) of the naturally detected patients in the study by Aeberhard et al. [1] can be considered representative. The expected LOS in patients with DRM (12 days) is plausible given that the average LOS of the target population is 10 days.

The costs of NS are very low compared with those of NI. The costs as well as the savings and the additional revenue are determined by the detection rate and the number of patients with DRM. Varying the assumptions relating to the detection rate will therefore result in a corresponding almost proportional variation in the difference between the costs and the savings and the additional revenue. This difference depends strongly on the assumptions regarding the savings and the additional revenue.

The projected savings are based on assumptions about the reduction of LOS due to NI. These are based on the results of meta-analyses, reviews and single studies. The literature is not consistent, but it shows evidence of or tendencies toward reduction of LOS, complication rates or rates of non-elective rehospitalizations [5,6,10–13]. However, for an economic assessment, available data on types and frequencies of avoided complications are not specific enough. Reductions in rehospitalizations are also difficult to evaluate because of varying follow-up periods. Therefore, it seems justified to use the reduction in LOS as a proxy for the many and various effects, with evidence of an improvement in clinical and economic outcomes due to NI.

For each one-day reduction in LOS, CHF 1371 was saved. This reflects the high costs associated with medical care in Switzerland and at university hospitals. A Danish study found the costs of outliers to be \$224 per day in 2006. A Dutch study from the year 2005 estimated these costs at 476 €, and a US study reported costs of \$1770 in 2018. The global average costs per hospital day at the Bern University Hospital amounted to CHF 2848 (information provided by the Federal Office of Public Health). So, valuing the LOS-dependent cost with per diem additional rates for DRG outliers seems justified.

The amount of additional revenues received is determined by the proportion of cases with DRG changes and the resulting increase in the average cost-weight in these cases. In the study by Aeberhard et al. [1], the proportion with changes in DRG attribution amounted to 8.3%. Compared with other studies in patients without NS, this is quite low. The average increase in the cost-weight was 0.694, a figure confirmed by other studies as well. In patients with NS, these numbers are not valid. The proportion of cases with changes in DRG attributions was assumed to be 15%. This is a conservative estimate and corresponds with the lower of two values from studies with NS (15% and 27%). The average increase in the cost-weight in these patients was lower, amounting to circa 0.44 in both studies, a figure, which was therefore applied in the current study.

#### **7. Cost E**ff**ectiveness of Nutritional Therapy in the Post-Hospital or Community Setting: A Brief Statement**

Prevalence of malnutrition has been studied less often in the community setting due to the lack of systematic or standardized screening programs, for example at healthcare institutions or in the offices of general practitioners. It is well established, however, that between 20% and 40% of patients admitted from the community setting to a hospital are already malnourished. Also, nutritional status frequently worsens during a hospital stay, which means that a large proportion of patients, particularly those older than 60 years, are discharged to a community setting or to other institutions with malnutrition. It is estimated that only around 10% of malnourished patients are in fact hospital patients, with the rest dwelling in a community or nursing home setting [23].

Few studies have addressed the costs of malnutrition and the economic impact of nutritional support in these settings [24]. Due to feasibility, the majority of studies investigating cost effectiveness of nutritional therapy are in-hospital analyses, but these only reflect a short period in terms of the patients' needs for nutritional therapy. Malnourished patients, particularly those who are old, frequently need continued nutritional support. This is a challenge for the analysis of cost effectiveness, as the costs of management in one setting may be offset by greater cost savings in another setting, such as when patients are moving from one care setting to another and a more comprehensive perspective is needed. Not surprisingly, nutritional therapy is frequently discontinued after a hospital stay. Most studies that have assessed cost effectiveness of NI in the community-dwelling population are initiated at hospital discharge [25–28]. These frequently use oral nutritional supplements and are carried out for 3–6 months after a hospital stay. Studies of community-dwelling or nursing home residents more frequently address further types of nutritional support, such as dietary counseling, snacks between meals, or multi-component nutritional support.

Table A5 (Appendix A) gives an overview of the type and design of studies investigating the cost effectiveness of nutritional support following hospital discharge in the community-dwelling elderly or in nursing home residents. Despite the different methodologies used, nutritional therapy was generally found to be cost effective, with the higher costs incurred by nutritional support being ultimately associated with decreased use of health care resources and improved quality of life.

#### **8. Conclusions**

This is the first study to compare the costs of a systematic NS and subsequent adequate NI, with the combined effects of savings due to reduced LOS and additional revenue resulting from SwissDRG changes. The costs of the intervention could be separately compensated for by savings in the costs of basic treatment and by the additional revenue generated. NS with subsequent adequate NI is thus associated with high economic potential for the hospital. Moreover, the EFFORT trial was able to demonstrate an impressive effect of NI, with a significant reduction in mortality and severe complications as well as improvement of physical function and quality of life in medical inpatients [11]. These clinical benefits, in terms of savings in supplemental hospital costs, were not included in the current economic impact analysis. If they had been, this would have greatly increased the financial efficacy.

**Author Contributions:** Conceptualization, E.R. and Z.S.; data analysis and curation, O.E. and A.F.; writing—original draft preparation, E.R. and K.N.; writing—review and editing, A.F. and O.E. and P.S. and Z.S.; supervision, Z.S.

**Funding:** The APC was funded by the Research Fund of the Department of Diabetes, Endocrinology, Nutritional Medicine and Metabolism and in part by Nestlé Health Science (grant to the institution).

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

#### **Appendix A**


**Table A1.** Patient flow and resource use in connection with ND.

DRM: disease-related malnutrition; p DRM: potential DRM; m DRM: manifest DRM; ONS: oral nutritional supplements; EN: enteral nutrition; PN: parenteral nutrition; LOS: length of hospital stay; DRG: diagnosis-related groups.


**Table A2.** Consequences of the costs of ND.

Costs are indicated with positive (+), savings and additional revenue with negative (−) prefix. DRM: disease-related malnutrition; p DRM: potential DRM; m DRM: manifest DRM; ONS: oral nutritional supplements; EN: enteral nutrition; PN: parenteral nutrition.


**Table A3.** Hours, workdays and positions needed for NI in the case of systematic NS, ND and additional needs for systematic NS.

**Table A4.** Staff time per nutritional intervention and professional group.


DRM: disease-related malnutrition; p DRM: potential DRM; m DRM: manifest DRM; ONS: oral nutritional supplements; EN: enteral nutrition; PN: parenteral nutrition.


**Table A5.** Studies addressing the economic impact of nutritional support after hospital discharge in the community or nursing home setting.

#### **References**


© 2019 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* **Metabolic and Nutritional Characteristics of Long-Stay Critically Ill Patients**

**Marina V. Viana 1, Olivier Pantet 1, Geraldine Bagnoud 1,2, Arianne Martinez 1, Eva Favre 1, Mélanie Charrière 1,2, Doris Favre 1,2, Philippe Eckert 1,**† **and Mette M. Berger 1,\*,**†


Received: 30 May 2019; Accepted: 4 July 2019; Published: 7 July 2019

**Abstract:** Background: insufficient feeding is frequent in the intensive care unit (ICU), which results in poor outcomes. Little is known about the nutrition pattern of patients requiring prolonged ICU stays. The aims of our study are to describe the demographic, metabolic, and nutritional specificities of chronically critically ill (CCI) patients defined by an ICU stay >2 weeks, and to identify an early risk factor. Methods: analysis of consecutive patients prospectively admitted to the CCI program, with the following variables: demographic characteristics, Nutrition Risk Screening (NRS-2002) score, total daily energy from nutritional and non-nutritional sources, protein and glucose intakes, all arterial blood glucose values, length of ICU and hospital stay, and outcome (ICU and 90-day survival). Two phases were considered for the analysis: the first 10 days, and the next 20 days of the ICU stay. Statistics: parametric and non-parametric tests. Results: 150 patients, aged 60 ± 15 years were prospectively included. Median (Q1, Q3) length of ICU stay was 31 (26, 46) days. The mortality was 18% at ICU discharge and 35.3% at 90 days. Non-survivors were older (*p* = 0.024), tended to have a higher SAPSII score (*p* = 0.072), with a significantly higher NRS score (*p* = 0.033). Enteral nutrition predominated, while combined feeding was minimally used. All patients received energy and protein below the ICU's protocol recommendation. The proportion of days with fasting was 10.8%, being significantly higher in non-survivors (2 versus 3 days; *p* = 0.038). Higher protein delivery was associated with an increase in prealbumin over time (*r*<sup>2</sup> = 0.19, *p* = 0.027). Conclusions: High NRS scores may identify patients at highest risk of poor outcome when exposed to underfeeding. Further studies are required to evaluate a nutrition strategy for patients with high NRS, addressing combined parenteral nutrition and protein delivery.

**Keywords:** chronic critical illness; protein; Nutrition Risk Screening (NRS-2002); age; nutrition; vasopressors; shock; glucose; diabetes; underfeeding

#### **1. Introduction**

The intensive care unit (ICU) patient population has evolved over the last two decades with the appearance of an increasing number of patients requiring very long ICU stays, lasting up to several months after surviving the initial acute insult [1]. A long stay is usually defined as the requirement of more than one week of mechanical ventilation (MV) and of ICU therapy, but different definitions have been used [2]. Chronic critical illness (CCI) is the most frequent designation for these patients, which is characterized by lengthy hospital stays, intense suffering, high mortality rates, and substantial resource consumption [3]. Genetic influence has also been shown to be increasingly present, particularly in pediatric ICUs [4]. Preexisting chronic comorbidities are strong, independent predictors of this

condition [5]. In the ICU, mechanical ventilation, sepsis, Glasgow score < 15, inadequate calorie intake, and higher body mass index (BMI) have been identified as independent predictors of this condition [6].

Recently, a study including 185 chronic, critically ill patients showed that the preadmission nutritional status reflected by the Nutrition Risk Screening (NRS-2002) (hereafter NRS) [7] might be a good predictor of the outcome: a score ≥ 5 points seemed to be a cutoff predictor of mortality [8]. Entering the acute disease with deficits such as weight loss, being older, or malnourished, seems to constitute a metabolic handicap that threatens the patient's response capacity and survival. In 2013, the same group observed that inadequate nutrition (defined as the provision of less than 60% of needs) and organ failure (defined by the sequential organ failure assessment (SOFA) score) are mortality risk factors [6].

The optimal timing and amount of feeding has been much debated. Indeed, the intestine of the sickest patients is not always available to accommodate full feeding as shown by the French randomized controlled, multicenter, parallel group trial called NUTRIREA-2 including 2410 patients with septic shock [9]. Furthermore, many patients are at risk of the refeeding syndrome [10]. A recent Brazilian study showed in a cohort of 100 critically ill patients with a mean ICU stay of 19 days, that those 45 patients receiving a caloric intake ≥70% during the first 72 h of hospitalization did not present better outcomes in the short term (mechanical ventilation, length of ICU stay) or after 1 year (functional capacity, mortality) [11]. For both above reasons, i.e., risk of stressing the intestine and of relative overfeeding [12], the early full feeding strategy seems unable to promote a good outcome. In this context, the most recent European Society for Clinical Nutrition and Metabolism (ESPEN) guidelines propose increasing progressively (ramping up) the energy delivery whatever the feeding route over 3–4 days [13]. Furthermore, prescribed energy goals should be covered somewhere between day 4 and 7, and preferentially be determined by indirect calorimetry.

To better coordinate the treatment of the long stay patients admitted to our ICU, and to identify procedures that could be improved, the service initiated a dedicated program enrolling patients requiring more than 2 weeks of ICU treatment. The present study aims to describe the metabolic and nutritional specificities of these patients, to analyze those associated with poor outcome, and to identify a factor that might enable early detection of CCI risk.

#### **2. Methods**

With the approval of the Commission Cantonale d'Ethique sur la Recherche humaine (CER 2018-02018), consecutive patients admitted to the 35 bed multidisciplinary ICU of the Lausanne University Hospital were analyzed. The patients had been prospectively enrolled in the long-stay program of the ICU, a plan of action called PLS (Patients Long Séjour) created in January 2017 to improve the care of patients requiring more than 2 weeks of ICU therapy (clinical trial identifier: NCT03938961). The PLS program consisted of weekly interdisciplinary meetings addressing ventilation, nutrition, cognitive and functional issues, symptom management, psychological support, and complication prevention. The meetings ended with detailed evaluation and a plan of action.

#### *2.1. Patients*

The inclusion criteria were age >18 years, and inclusion in the PLS program with an ICU stay >2 weeks. The exclusion criteria were admission for major burns >20% body surface, traumatic brain injury, and patients who refused participation.

#### *2.2. Study Variables*

Data were extracted from the computerized information system (Metavision® iMDsoft, Tel Aviv, Israel, version 5.46.44) and included age, admission (dry weight) and discharge body weight, body mass index (BMI), NRS-2002 score [7], severity of disease (SAPSII), Sequential Organ Failure Assessment (SOFA) score on days 1, 2, 5, 10, 30, and discharge [14]), requirement of continuous renal replacement therapy (CRRT), admission for sepsis, presence of diabetes, daily intakes (energy, protein, glucose: raw data and per kg), all arterial blood glucose values, 24 h insulin dose, arterial blood lactate (maximal value), daily feeding route, pressure ulcers occurring during the ICU stay, discharge Medical Research Council force score (MRC), length of mechanical ventilation, ICU and hospital length of stay, ICU and 90-days mortality. Laboratory: blood prealbumin (determined weekly), C-reactive protein (CRP), and procalcitonin (mean of the ICU stay).

#### *2.3. Nutrition Protocol*

The ICU's protocol called NUTSIA (NUTrition aux Soins Intensifs Adultes) is based on the ESPEN guidelines [13]. Enteral nutrition is recommended as the first option when nutritional therapy is indicated (i.e., for patients for whom oral intake is not possible), and should be initiated within the first 3 days after hemodynamic stabilization. Parenteral nutrition (PN) may be used as combined feeding, or total PN from day 4 and earlier in selected malnutrition situations. Energy goals are 20 kcal/kg/day during the first week, to be increased or adapted thereafter—indirect calorimetry is recommended from day 7. The initial energy goal should be reached by day 4–5. Energy intakes include the energy from feeding products and oral intakes, and the energy resulting from glucose (drug dilution) and fat (sedative propofol), which are administered for non-nutritional purposes.

Feeding products were: Peptamen Intense (Nestlé, Switzerland), Promote Fiber plus (Abbott, Switzerland), Isosource Energy Fiber (Nestlé, Switzerland), Nutriflex Omega special (BBraun, Switzerland), and Pharmacy compounded PN in individual cases.

Energy balances were calculated as the daily difference between the total substrate intakes (nutritional and non-nutritional sources) and the prescribed value. The mean daily and the cumulated values were calculated for the first 10 days, and a cutoff of cumulated deficit −70 kcal/kg was considered critical [15,16]. For proteins, the daily delivery was compared to the recommended target of 1.2 g/kg/day. The cumulated deficit was calculated for the first 10 days and a cutoff of −300 g was considered critical.

Glucose control is handled by nurses, with a blood glucose (BG) goal of 6–8 mmol/L (6–10 mmol/L in diabetic patients) maintained using continuous insulin infusion. Blood glucose is presented as the daily mean value, and is measured on a point of care blood gas analyzer. Its variability is indicated as the daily standard error of the daily BG values. Daily total insulin was recorded.

The 2 dieticians (one full time position present on working days) attended to patients requiring >3 ICU days. These individuals are in charge of checking route of feeding, energy needs (indirect calorimetry), proteins delivery, and adapting the feeding solutions [17]. They attend the weekly PLS meetings.

#### *2.4. Other Variables*

Muscle strength upon discharge was assessed using the MRC score [18], where the patient's effort is graded on a *scale* of 0–5. Six muscle groups were bilaterally measured (abduction of the arm, flexion of the forearm, extension of the wrist, flexion of the hip, extension of the knee, and dorsal flexion of the foot). All muscle groups were scored between 0 and 5 (0, no visible/palpable contraction; 1, visible/palpable contraction without movement of the limb; 2, movements of the limb but not against gravity; 3, movements against gravity (almost full passive range of motion) but not against resistance; 4, movement against gravity and resistance, arbitrarily judged to be sub-maximal for gender and age; 5, normal). Therefore, the maximal score is 60 points.

#### *2.5. Statistical Analysis*

Data of the entire stay were extracted, but analysis was limited to the first 30 days. The data are presented as means ± standard deviation (SD), or median (interquartile Q1, Q3) depending on normal distribution, which was assessed through histograms and calculation of skewness and kurtosis. For non-parametric variables, the Kruskal-Wallis test was used. Two phases were considered for the analysis of the nutritional related variables: the first days (D1–D10), and the next 20 days of the ICU stay (D11–D30). The continuous variables in survivors versus non-survivors were compared using

the one-way and two-way ANOVA, while the categorical variables were compared using Chi2 tests. For non-parametric variables, Kruskal-Wallis rank sum test was used. Kaplan–Meier analysis and Cox regression were used to compare mortality. For the Kaplan–Meier analysis, the log rank test was used to compare the curves of each NRS group (3–4 and 5–6–7 points). The generalized linear model (GLM) was used to analyze response variables that were not normally distributed such as substrate administration or blood glucose over the 30 days. Statistical programs were JMP version 14.2 for Windows, (SAS Institute GMH, Böblingen, Germany), and R Version 3.5.3, 2019 (R Foundation for Statistical Computing, Vienna, Austria). Significance level was set at *p* < 0.05.

#### **3. Results**

#### *3.1. Patients*

The current study included prospectively 150 patients admitted between 1 February 2017 and 31 December 2018. During the same period, 589 patients were admitted who required ICU for >7 days, but the majority were discharged before entering the PLS program. Table 1 summarizes patient demographics and clinical outcomes, with the details of survivors/non-survivors at ICU discharge. Most patients (60.7%) were admitted for medical causes (17/91 with a surgical background), followed by emergency surgery, and elective surgery (52% were admitted for sepsis, and 48% needed continuous renal replacement during their stay).


**Table 1.** Demographics, severity of illness, laboratory and outcome variables according to ICU vital status.

Abbreviations: SOFA =Sequential Organ Failure Assessment, SAPS=Simplified Acute Physiology Score, BMI=Body Mass Index, NRS = Nutrition risk screening, CRP = C-reactive protein, Q1 and Q3 = q quartiles 25, 75.

Gender was equally distributed, but the non-survivors were older (*p* = 0.024) and tended to have a higher SAPSII score (*p* = 0.072). The admission SOFA score did not differ significantly but its evolution was different over time, decreasing significantly in survivors (*p* < 0.0001) (Figure A1). The cardiac component of the score remained elevated for a longer period of time (i.e., at 3–4 points) in the non-survivors. A significantly higher NRS score (*p* = 0.033) was observed in non-survivors. The Kaplan–Meier analysis showed that the scores NRS 5–6–7 are associated with a higher 90-days mortality of 43.8% versus 25.7% with NRS scores 3 and 4, respectively (Figure 1). Admission weight (mean 78.3 kg) and BMI were similar, but the weight change over time differed significantly (*p* < 0.0001) depending on the outcome. By the end of the stay, survivors had lost weight, while non-survivors had gained weight due to a persistent positive fluid balance. While 28 patients (18%) died in the ICU, further 26 patients died within 90 days, resulting in a total of 56 (35.3%) deaths—age and NRS differences became even more pronounced by day 90 (age *p* < 0.001; NRS *p* = 0.005; Table A1). A Cox regression model with 90-days mortality as an outcome showed an increased risk of death in patients with NRS ≥ 5 (HR 2.2 (1.18–4.2), *p* = 0.013), but not for the SAPS2 score (HR 1.0 (0.99–1.0), *p* = 0.235). Diabetes mellitus was present in 17.3% of patients, but was not associated with any mortality difference.

**Figure 1.** Kaplan–Meier analysis comparing elevated and low NRS scores. NRS: Nutrition Risk Screening and ICU: intensive care unit.

#### *3.2. Nutrition*

The proportion of days with intentional absence of feeding (fasting) was 10.8% of total days (Table 2 and Figure 2), being significantly higher in non-survivors (*p* = 0.038), with fasting contributing to the high day-to-day variability of intakes (Figure 3)—except for the number of fasting days (3 versus 2, *p* = 0.043). Table 2 shows that there was no difference between survivors and non-survivors. Fasting days were mostly observed during the first 5 days, but did occur throughout the 30 days. The feeding route was predominantly enteral, representing 55.7% of total days; combined enteral (EN) and parenteral (PN) feeding represented 8.1% and 14.1% of total days, respectively, and variable route combinations was used in 1% of total days.


Nutrition characteristics according to ICU outcome (D = day).

> **Table 2.**

**Figure 2.** Evolution of the route of feeding over time presented as percentage of all patients over the first 30 days—there is a variable time of fasting during the first week. Enteral feeding was predominant, with a stable proportion of combined enteral–parenteral feeding (Comb EN + PN), or total parenteral nutrition (PN), and a variable proportion of the combinations oral–enteral, or oral–parenteral. Abbreviations: EN = enteral nutrition, PN = parenteral nutrition, Comb = combined, PO = oral.

**Figure 3.** Individual delivery of protein and glucose by day during the first 10 days. Each line represents individual patient values. The erratic aspect aims at showing a phenomenon which is the extreme day to day variability and multiple interruptions that characterize the nutrition in the early phase. The thick dark lines show median (blue) and mean (black) values.

#### 3.2.1. Energy

The mean prescribed energy during the first 10 days was 21.7 kcal/kg/day, and 23.4 kcal/kg/day thereafter (1750 kcal/day). The individual patients were characterized by a high variability of feeding, and total protein and glucose intakes (Figure 3). The progression of feeding occurred over the first 10 days (and not over 5 days as per protocol), i.e., before inclusion in the dedicated program

(no difference between the two categories). Figure 4 shows the mean values for protein and energy, which were below the service's recommendations and below the prescribed values. Both groups were underfed, resulting in a median cumulated negative balance of −5266 kcal by day 10 (−70 kcal/kg). Nutrition delivery improved over time, but stagnated below the prescribed value.

Indirect calorimetry was available in 95 patients. The prescription was −236 kcal (range: −1391 to +483 kcal) below the measured energy expenditure. When this deficit is added to the daily energy deficit (difference between prescribed and delivered), the median deficit becomes −741 kcal/day (range: −2971 to −123 kcal per day) for the first 10 days.

#### 3.2.2. Proteins

The delivery was lowest during the first 10 days (Figure 4), with a median of 64.3 g/day (0.70 g/kg/day), which is below the 1.2 g/kg/day service NUTSIA recommendation. By day 10, 80 patients (53.3%) exceeded −300 g of cumulated deficit. Protein intake thereafter increased to 1 g/kg/day remaining below recommendation, with a tendency to be lower in non-survivors.

**Figure 4.** Mean protein and energy delivery with the resulting energy balance over the first 30 days according to ICU vital status (mean ± SD). The thick gray lines show protein target (1.2 g/kg/day), energy goal (prescribed value), and neutral energy balance. The differences in protein and energy delivery between survivors and non-survivors were significant after day 10 (*p* < 0.001). Energy balances were similarly negative.

Prealbumin, the lowest values were observed in patients with high NRS scores (*p* = 0.067). Prealbumin increased over time in the majority of patients in association with higher protein delivery (*r*<sup>2</sup> = 0.19; *p* = 0.027), and the increase (difference between first and last value of the ICU stay) was lower in non-survivors (*p* < 0.001).

#### 3.2.3. Blood glucose

Altogether, 30,769 arterial blood glucose values were available for analysis (4–10 blood samples per day). Blood glucose and related variables differed between survivors and non-survivors (Figure 5), and over time. While glucose intakes were identical and low, non-survivors had higher BG values during the first 10 days, but not thereafter. Variability was higher, but not significant. Insulin needs

were significantly higher and became again higher after day 20, while survivors' insulin needs declined, reaching below 30 u/day. Blood maximal lactate was also significantly higher during the first days in non-survivors.

The BG pattern differed in diabetic patients, being higher throughout the stay as per protocol, with higher variability and higher 24 h insulin requirements (Figure A2). This was particularly marked in surviving diabetic patients (Figure A3). Diabetic non-survivors were characterized by lower BG values and higher insulin needs during the first 5 days. Arterial lactate was elevated during the first 5 days in the majority of patients, being significantly higher in diabetic non-survivors.

**Figure 5.** Evolution of mean blood glucose (mmol/L), Blood Glucose (BG) variability (standard error = SD of the individual daily values), 24 h insulin (total dose/24 h), and glucose intake (total in g/day). The figure shows that the 24 h Insulin delivery is not related to the glucose intake, which was similar in both groups (data are represented as mean ± SD).

#### *3.3. MRC Discharge Score*

MRC discharge score was available in 91 patients. The median value on discharge was low with 34 points. In those patients who did not suffer significant persistent renal failure on discharge (which results in elevated creatinine values), the median serum creatinine was 52 μmol/L (a low value in adult patients, and a surrogate for muscle mass).

#### **4. Discussion**

The main finding in our study is that CCI patients are not equal upon admission. Many patients begin their ICU journey without metabolic reserves reflected by a high NRS score. However, the impact is not immediate. During the first 10 days, the patients were exposed to very low and highly variable energy and protein intakes, and many were exposed to prolonged and repeated fasting. The acute underfeeding generated important energy and protein deficits, resulting in an additional malnutrition diagnosis in these critically ill patients. Combined EN and PN feeding was minimally used, despite being included in the service's protocol based on randomized trials [19,20]. In patients at risk of malnutrition upon admission, the underfeeding further erodes a previously altered metabolic status, resulting in a poor outcome. These results confirm recent observational data [16], where patients at low risk are likely to have a reasonable outcome despite several days of inadequate nutrition, while

those at higher risk do not. The 2019 ESPEN recommendations [13] emphasize that no patient should be left longer than 72 h without initiating a balanced feeding. The majority of nutrition-related studies focus on the first days after admission, and have often included moderately sick or young patients that did not really benefit from artificial nutrition. By extracting the data of the entire stay, we observed that there was an early phase, and a later phase with a turning point around day 10. Thereafter, the patients were better fed, but the majority did not reach their prescribed goals.

The patient cohort was critically ill, as reflected by an elevated admission SAPSII score and an 18% ICU mortality, which was significantly higher than the general ICU patient's mortality (14.4%). The non-survivors were older, which belongs to the risk factors of CCI [1]. At the 90 days time point, a further 26 patients had died, increasing the total mortality to 35.3%. The initial SOFA score did not differ, but the score's evolution differed in survivors and non-survivors, remaining higher in the latter. The inclusion of the evolution of organ failure over time assessed by the SOFA score is an important strength of this study. Our data do not confirm those of a Brazilian cohort of 453 CCI patients, in which the initial SOFA score was a predictor of mortality [6]. One of the aims of the study was to find a variable facilitating the early detection of patients at critical illness risk. The total SOFA score did not predict the outcome, while the NRS score was able to detect them, confirming other Brazilian results [8]. A NRS score ≥ 5 should raise an alert, especially when associated with hemodynamic instability and high lactate values.

Timing and route of feeding in the critically ill has long been debated, and particularly the timing of PN initiation. The pragmatic CALORIES trial in 2400 mechanically ventilated patients randomized to EN or PN from day 1 showed no difference between the two routes [21]. The NUTRIREA-2 study included 2410 patients with septic shock, and showed that full early EN from day 1 was associated with more intestinal complications (*p* < 0.001) compared to the same dose of PN [9], with no benefit of EN on infectious complications. Providing early full feeding does not seem to prevent patients from becoming CCI, as shown by a Brazilian study including 100 critically ill patients. The study further shows that providing a caloric intake ≥70% in the first 72 h of admission did not improve short term or 1-year outcomes [11]. Among the possible explanations of these disappointing results, the early endogenous glucose production [20,22] in response to acute illness should be considered. This production covers approximately two thirds of energy expenditure during the first days of disease. In addition, enteral full feeding is not always tolerated, as shown by NUTRIREA-2 [9], and the risk of refeeding syndrome is a reality [10]. The dose of both feeding and vasopressors might be the explanation as shown by the secondary analysis of a large septic cohort which showed that EN providing close to recommended intakes of energy and proteins over the stay was associated with a better outcome [23]. A propensity-matched analysis, including 52,563 ventilated adult patients stratified by dose of noradrenalin [24], suggests that in patients on low- or medium-dose noradrenalin, early EN seems associated with a reduction in mortality but not in the patients requiring high-dose noradrenalin. The recommendation to provide early, but progressive feeding starting ideally during the first 48 h, while discouraging early full feeding whatever the route, formulated in the latest ESICM [25] and ESPEN [13] guidelines was built on such data.

The ICU's internal protocol recommends progressing to the goal by day 4–5 based on ESPEN recommendations, which was not achieved without any difference between survivors and non-survivors. The latter were characterized by a higher percentage of fasting days during the first days after admission (associated with hemodynamic instability), but also occurring at random thereafter. The reasons for the feeding interruptions are multiple in an ICU, and have not changed much over the last decades [26]. The elevated mean cardiovascular SOFA scores reflect a persistent hemodynamic instability, which is probably one of the reasons for withholding feeding, and for difficult EN progression. These poor results may need a change in our ICU feeding procedure. Volume based feeding needs consideration as recent studies show that it is safe, and effectively improves energy and protein delivery compared to traditional rate-based feeding [27].

An American study analyzed the discharge destinations of critically ill surgical patients according to the magnitude of their energy and protein deficit. The authors showed that nutrition was a major outcome determinant [16]. Yeh et al. used a cumulated deficit cutoff of −6000 kcal and of −300 g protein. The patients who remained below those cutoffs were three times more likely to be discharged home. The authors also observed a longer ICU stay and higher mortality in those patients exceeding these values. In the present cohort, the patients entered the PLS program only after 2 weeks. Large energy and protein deficits had built up, but did not differ between survivors and non-survivors, both being similarly underfed by day 10: 80 patients (53.3%) exceeded the −300 g protein deficit and 50 patients (33%) exceeded the −6000 kcal cutoff. Knowing that the indirect calorimetry values for energy expenditure were higher than prescription in the majority of patients, the real deficit is even greater. Malnutrition causes loss of lean body mass, which is a determinant of the outcome [28,29], with the early protein deficit being an important contributor to the loss. The positive response of prealbumin to higher protein delivery shows that increasing proteins delivery is a treatment option. Critical illness is characterized by a high degree of stress with an accelerated protein degradation that results in malnutrition, systemic inflammation, and organ dysfunction [30]. Supporting this hypothesis, Briassoulis et al. showed that in critically ill children, only 22.7% of patients without protein deficiencies versus 37% of those at risk or already deficient, developed multiple-organ system failure [31]. Further transferrin and prealbumin levels increased already after 5 days of early EN, and the patients with positive nitrogen balance had higher prealbumin levels [32]. In critically ill patients, this muscle loss occurs very rapidly as shown in the landmark study by Puthucheary et al. [33]. In patients with two or more organ failures, the mean loss of muscle measured by ultrasound of the thigh was 22% in 10 days. The present patients were in multiple organ failure as reflected by their high SOFA score. However, there is currently no standardized procedure to assess sarcopenia in long-stay catabolic patients [34], and only surrogates are available: low (33 points) muscle strength at discharge (MRC score) and the discharge creatinine values below those recorded at admission reflect the loss of muscle mass. These results highlight the importance of an early multidisciplinary awareness for metabolic and nutritional issues, i.e., waiting 2 weeks to address them is too long. The timing, dose, and route of feeding must be addressed more stringently and earlier, and the recommendations of the dieticians applied more diligently in the high risk patients already by day 3–4.

Among the factors associated with ICU weakness, hyperglycemia has been considered important, as insulin therapy might be an attenuating factor of muscle loss through its anabolic properties [35]. Hyperglycemia is associated with increased mortality in critically ill patients [36], an issue that appears to improve by tight glycemic control [37]. The debate regarding optimal glycemic control continues as diabetic patients are frequent and exhibit a different response to tight glucose control [38]. However, there is little data in patients with prolonged ICU stay. One study reported that tighter glycemic control was associated with improved outcomes in CCI patients with stress hyperglycemia, but not in CCI patients with diabetes [39]. In the present study, the initial BG values were elevated, independently of low glucose intake. BG normalized around day 10, and remained stable thereafter, being higher in the non-survivors despite similar glucose intakes. The observed changes in BG and insulin needs over time may be related to the decrease in lean body mass. The future non-survivors also required more insulin during the 2nd phase, possibly reflecting a metabolic derangement. Diabetic patients exhibited a different pattern compared to non-diabetics corresponding to the application of the internal protocol.

Limitations of the study: our current study is observational, including only 150 patients, but the patients were enrolled prospectively to the PLS program and the study was aimed at identifying early risk criteria. The strength of the data comes from the quality of data extraction from the computer system, resulting in longitudinal continuous daily data. Regarding fasting, the study was not designed to record the reasons for the feeding interruptions, so exact causes are missing but are unlikely to differ from other observations [26]. Furthermore, our study lacks a measure for the loss of lean body mass. Except for ultrasound, which was not available in our clinical setting, there is currently no standardized procedure to assess sarcopenia in CCI patients [34]. Whole body multi-frequency bioimpedance, phase

angle, and ultrasound assessment of the cross-sectional area of the thigh might become clinically useful tools, as they are non-invasive and little costly. We therefore consider integrating them into the nutrition protocol.

In randomized trials, our group showed that, in case of insufficient enteral feeding during the first 3 days (i.e., <60% of 25 kcal/kg/day which is <15 kcal/kg/day), supplemental PN (SPN) guided by indirect calorimetry to cover the measured value from day 4, reduced nosocomial infections [19] and related costs [40]. Furthermore, our group recently showed that this reduction was explained by modifications of the immune response and attenuation of the inflammatory response [20], with a trend to less loss of muscle mass (*p* = 0.07). This combined strategy should probably be applied more frequently, especially in patients with NRS ≥ 5 who obviously are not covering their needs, and who have persistent high cardiac SOFA scores. The two above Swiss SPN studies also show that a two-phase approach may be the most respectful of physiological mechanisms. Critically ill patients are generally unstable during the first days, and their endogenous energy production is elevated [22], providing 200–300 g glucose per day. This endogenous glucose production has an elevated protein cost, and 120 g/day were shown to be catabolized for neoglucogenesis. Full feeding during this period is therefore associated with a risk of overfeeding [12]. The present study shows that the very simple NRS score is probably able to identify patients who will not tolerate a prolonged underfeeding. This two-step, more physiological approach and the absence of difference between the enteral and parenteral feeding was recently summarized [41]. It is also important to integrate that "nutrition is more than the sum of its parts" as described by Briassoulis et al. [42], who emphasized that focusing only and separately on proteins or energy is not sufficient. An integrated, individualized approach, including all substrates and micronutrients is needed [43].

#### **5. Conclusions**

The present study shows that high NRS scores upon admission may identify patients at higher risk of a poor outcome, and those who require individualized nutrition therapy [43]. Exposing these high-risk patients to underfeeding further deteriorates their response capacity, including immune defenses [40], possibly favoring the development of chronic critical illness. The NRS score is not a malnutrition score, but identifies those patients who are at risk of high mortality at an early stage, this risk increasing further with underfeeding. Having no reserves, these patients will cope worse, and continue eroding their lean body mass, contributing to the poor prognosis [28,29]. Our study also shows that the most unstable patients, who become chronic, are exposed to discontinuous feeding, partly explained by persistent shock and high blood lactate values which question the clinicians about intestinal integrity, and the feasibility of EN [25]. In our previous studies [15], and in others [16], the cutoff of −70 kcal/kg of cumulated energy deficit was associated with increasing complications. Preventing such a deficit to build up requires identifying the patients at risk by their elevated NRS and hemodynamic instability. This would enable the introduction of combined feeding by day 4–5, while checking the real needs by indirect calorimetry. This strategy should be tested prospectively.

**Author Contributions:** Conceptualization P.E., E.F., O.P., and M.M.B.; methodology M.M.B., M.V.V., and P.E.; validation, M.V.V., O.P., G.B., A.M., E.F., M.C., D.F., P.E., and M.M.B.; formal analysis, M.M.B. and M.V.V.; methodology, M.V.V., E.V., A.M., and M.M.B.; supervision, M.M.B.; validation, M.V.V., G.B., M.C., and D.F.; investigation, M.V.V., O.P., A.M., E.F., M.C., D.F., and G.B.; resources, P.E. and M.M.B.; data curation, G.B., M.V.V., M.M.B., M.C., and D.F.; writing—original draft preparation, M.M.B. and M.V.V.; writing—review and editing, M.V.V., O.P., G.B., A.M., E.F., M.C., D.F., P.E., and M.M.B.; project administration, M.M.B., P.E., and O.P.

**Conflicts of Interest:** The authors declare no conflict of interest for the present research.

#### **Appendix A**

**Table A1.** Demographics, severity of illness, laboratory, and outcome variables according to the 90-day outcome.


**Figure A1.** Evolution of the total SOFA score (**A**), and of its cardiac (**B**) and respiratory (**C**) components over time. The total SOFA score is shown as box plots (median is the line within the box, whiskers are 10th and 90th percentiles, the points above and below indicate outliers; "Dout" on time axis is the day of discharge). The boxes B and C show smooth curve uniting the days of available SOFA. The gray bands represent the 95% confidence intervals. Total SOFA changes over time were significantly different (*p* < 0.0001) between survivors and non-survivors, driven by the cardiovascular component of the score. The respiratory score did not differ between groups, remaining around three points for a long period: respiratory insufficiency was a frequent reason for prolonged ICU stay.

**Figure A2.** Evolution of glucose variables in the patients with diabetes (*n* = 26) versus those without diabetes (*n* = 124). Blood glucose (BG) and insulin needs were significantly higher in diabetic patients (*p* < 0.0001), as was the BG variability (*p* < 0.001) through the stay, with similar glucose intakes.

**Figure A3.** Glucose variables according to ICU outcome in patients with diabetes (DM) versus those without diabetes. Blood glucose (BG) was highest (*p* < 0.001) in DM-survivors compared to all, While the insulin needs did not differ in non-diabetic patients between survivors and non-survivors, the DM-non-survivors were characterized by significantly higher insulin needs during the first 6 days (*p* < 0.001), and lower needs than DM-survivors thereafter. Arterial lactate was elevated in all patients groups, and significantly higher in DM-non-survivors during the first 7 days (*p* < 0.001).

#### **References**


© 2019 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* **Protein Intake, Nutritional Status and Outcomes in ICU Survivors: A Single Center Cohort Study**

#### **Peter J.M. Weijs 1,2, Kris M. Mogensen 3, James D. Rawn <sup>4</sup> and Kenneth B. Christopher 5,\***


Received: 16 December 2018; Accepted: 31 December 2018; Published: 4 January 2019

**Abstract:** Background: We hypothesized that protein delivery during hospitalization in patients who survived critical care would be associated with outcomes following hospital discharge. Methods: We studied 801 patients, age ≥ 18 years, who received critical care between 2004 and 2012 and survived hospitalization. All patients underwent a registered dietitian formal assessment within 48 h of ICU admission. The exposure of interest, grams of protein per kilogram body weight delivered per day, was determined from all oral, enteral and parenteral sources for up to 28 days. Adjusted odds ratios for all cause 90-day post-discharge mortality were estimated by mixed- effects logistic regression models. Results: The 90-day post-discharge mortality was 13.9%. The mean nutrition delivery days recorded was 15. In a mixed-effect logistic regression model adjusted for age, gender, race, Deyo-Charlson comorbidity index, acute organ failures, sepsis and percent energy needs met, the 90-day post-discharge mortality rate was 17% (95% CI: 6–26) lower for each 1 g/kg increase in daily protein delivery (OR = 0.83 (95% CI 0.74–0.94; *p* = 0.002)). Conclusions: Adult medical ICU patients with improvements in daily protein intake during hospitalization who survive hospitalization have decreased odds of mortality in the 3 months following hospital discharge.

**Keywords:** protein; malnutrition; critical care; mortality; outcomes; hospital readmission; ICU Survivors

#### **1. Introduction**

A hallmark of critical illness is muscle wasting related to a dramatic increase in muscle protein catabolism in the setting of inflammation [1]. Catabolic critical illness results in a protracted and dramatic loss of nitrogen and reduces exogenous amino acid deposition into endogenous proteins [2]. Nutritionally compromised hospitalized patients commonly present with sarcopenia [3]. Inflammation related endogenous skeletal muscle protein catabolism that can quickly progress to severe muscle atrophy [4]. In the critically ill, muscle mass loss and decreased strength are common complications [5].

Protein intake recommended for healthy adults is 0.8 g/kg per day. While the most common recommended protein intake in the critically ill is 1.5 g/kg per day [6,7], up to 1.5–2.5 g protein/kg per day may be most advantageous [7,8]. ICU patients commonly receive less than even the lowest guideline recommendation regarding adequate protein intake [7,9–11]. Recent trials [12–15] demonstrate that "calorie-supplemented, protein-deficient nutrition" [4] does not improve clinical outcomes. Early energy provision of 70–80% of measured energy expenditure appears to be associated with improved outcome [16,17]. Although the correct amount of protein to provide critically ill patients is unknown, higher than 1.2 g/kg is associated with reduced mortality in non-septic non-energy overfed ICU patients [16]. A recent large observational study suggests achieving ≥80% of prescribed protein intake in the ICU is associated with decreased mortality [18].

Although short term survival is studied in the critically ill regarding protein delivery [18], post-hospital discharge outcomes in such ICU survivors relative to protein intake and nutrition status is not known. In patients who survive critical illness, pre-existing malnutrition may be a risk factor for adverse events following hospital discharge [19]. As ICU patients with low protein delivery [18] and with malnutrition [20] have elevated in-hospital mortality, we studied the association between increases in daily protein delivery in malnourished critically ill patients would be associated with lower 90-day mortality following hospital discharge. We hypothesized that in hospital survivors who received critical care, those with higher daily protein delivery would have a lower risk of post discharge mortality and this effect would be magnified in patients with pre-existing malnutrition.

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

#### *2.1. Data Sources*

We obtained administrative and laboratory data from critically ill medical patients admitted to the Brigham and Women's Hospital (BWH), an academic medical center in Boston with 777 beds. Patient data was collected by the Research Patient Data Registry (RPDR) a clinical data warehouse for patient records at BWH that has been utilized and validated in other studies [20]. The Partners Human Research Committee Institutional Review Board granted approval for the study. Between 2004 and 2012 there were 801 unique patients ≥18 years, assigned the CPT code 99291 (critical care, first 30–74 min) [21] who had nutrition risk assessed, protein intake measured, and survived to hospital discharge.

#### *2.2. Exposure of Interest and Comorbidities*

All medical ICU patients at BWH are screened by a registered dietitian (RD). Patients at risk for malnutrition are then evaluated by an RD with a structured objective assessment. The exposure of interest, grams of protein per kilogram body weight delivered per day, was determined by an RD from all oral, enteral and parenteral sources for up to 28 days. We included all nutrition data collected by an RD in the ten days prior to ICU admission to two days following an ICU discharge. The determination of nutrition status is described in detail previously [20,22]. The RD determines the diagnosis of malnutrition based on literature [23,24], clinical judgment, and on data related to inadequate nutrient intake, muscle wasting, subcutaneous fat wasting and unintended weight loss [22]. As in our prior study, we categorized nutrition diagnoses a priori into four groups of increasing severity: malnutrition absent, at risk for malnutrition, non-specific malnutrition, or any protein-energy malnutrition [22]. Patients that meet criteria for non-specific malnutrition have risk factors for malnutrition (insufficient intake of energy, protein, and micronutrients) in addition to metabolic stress and/or obvious signs of malnutrition (muscle and/or subcutaneous fat wasting) without supporting biochemical or anthropometric data. Patients that meet criteria for protein-energy malnutrition have a combination of disease-related weight loss, obvious muscle wasting, peripheral edema, inadequate energy or protein intake and are considered underweight by percent ideal body weight [25]. Patients with absence of malnutrition are diagnosed by an RD as well-nourished and not at risk for malnutrition.Race was self-determined or determined by the patient's family or healthcare proxy. The Deyo-Charlson comorbidity index was determined using validated ICD-9 coding algorithms [26]. To define sepsis we used the presence of the ICD-9 codes 038, 785.52, 995.91, or 995.92 in the three days before ICU admission to the 7 days after ICU admission [27]. The vasopressors/inotropes

covariate is the prescription of any vasopressor or inotrope in the three days prior to ICU admission to the seven days after ICU admission [20].

The number of acute organ failures was a combination of ICD-9-CM and CPT codes relating to acute organ dysfunction assigned from three days prior to ICU admission day to the 30 days after [28,29]. Intubation was defined as the presence of ICD-9 codes mechanical ventilation (96.7×) following hospital admission [30]. Acute kidney injury is the RIFLE class Injury or Failure taking place in the three days prior to ICU admission and the seven days after ICU admission [31]. Chronic kidney disease is <60 mL/min glomerular filtration rate calculated from the Modification of Diet in Renal Disease (MDRD) equation using pre-hospital or hospital admission creatinine as baseline [32]. Changes from the expected hospital length of stay (LOS) are determined by subtracting the actual LOS from the geometric mean LOS for each Diagnostic Related Grouping (DRG) a classification system for hospital cases [33]. The geometric mean LOS is the United States national mean LOS for each DRG per the Centers for Medicare & Medicaid Services [34].

#### *2.3. End Points*

Our primary outcome was all cause 90-day post-discharge mortality. Mortality status for the study cohort was obtained from the United States Social Security Administration Death Master File. Mortality determination via the Death Master File is validated in our administrative database [21]. The entire cohort had mortality status present for the year following hospital discharge. The censoring date was 15 March 2012. All patients in the cohort had at least 90-day follow-up after hospital discharge or expired before the 90-days post-hospitalization.

#### *2.4. Power Calculations and Statistical Analysis*

90-day post-discharge mortality rate was 9.6% in our prior study nutrition in critical illness [35]. From these data, we assumed that 90-day post-discharge mortality would be 6.0% greater in patients with <1.0 g/kg/day protein delivery compared to those with >1.0 g/kg/day protein delivery. With a power of 80% and an alpha error level of 5%, the sample size required for the 90-day post-discharge mortality outcome was 393 patients with >1.0 g/kg/day protein delivery and 393 patients with <1.0 g/kg/day protein delivery.

We described categorical covariates via frequency distribution, and compared variables in outcome groups with chi-square testing. We compared continuous variables across outcome groups using one-way analysis of variance (ANOVA) or the Kruskal–Wallis test. We utilized mixed-effect logistic regression models [36,37] which contain both fixed and random-effects for analysis of the association between g/kg/day protein delivery and 90-day post-discharge mortality by use of the command xtmelogit [38] in STATA 14.1 MP (College Station, TX, USA). The dates of protein delivery of individual patients were used as the random effect. We assessed the following covariates for confounding: age, gender, race, Deyo-Charlson comorbidity index, energy delivery, nutrition status, acute organ failure, sepsis, metastatic malignancy, acute kidney injury and chronic kidney disease. We selected the final model confounders by analyzing the maximum model and then conducting backward elimination of covariates with a *p* > 0.10. The final model had an independent covariance structure of the random effects; a fixed effect for age, gender, race, Deyo-Charlson comorbidity index, energy delivery, nutrition status, acute organ failure and sepsis; and gaussian-distributed random intercepts and slopes. We then used mixed-effect linear regression to analyze the association between the change in daily protein delivery and hospital LOS where random effects accounted for the dates of protein delivery within individual patients. Data visualization was performed utilizing the DistillerSR Forest Plot Generator (Evidence Partners, Ottawa, ON, Canada). *p*-values presented are two-tailed with *p* < 0.05 considered to be significant. We utilized STATA 14.1MP (College Station, TX, USA) for all analyses.

#### **3. Results**

Patient characteristics were stratified according to 90-day post-discharge mortality (Table 1). In our cohort the mean (SD) age was 62.3 (16.6) years, 55.3% were male and 78.9% were white. The majority of patients were assessed by an RD within 24 hours of ICU admission (median [IQR] time between ICU admission and RD assessment was 0 [1, 1] days). The mean number of recorded nutrition delivery days was 15. A total of 8735 days with protein intake were determined. The mean peak protein daily delivery was 0.32 g protein/kg. Figure 1 shows protein delivery over time in the cohort.

The mean (SD) length of hospital stay was 22.1 (15.2) days. The 30-, 90- and 365-day post-discharge mortality rates were 7.1%, 13.9%, and 24.5%, respectively. Age, race, Deyo-Charlson comorbidity index, malnutrition, the acute organ failure score and chronic kidney disease are significantly associated with 90-day post-discharge mortality (Table 1).

**Figure 1.** Box plots of daily protein delivery up to 10 days in patients who did and did not survive to 90-day post-discharge mortality, showing the median (white line), the third quartile (Q3) and first quartile (Q1) range of the data and data outliers (observations outside the 9–91 percentile range).



Data presented as *n* (%) unless otherwise indicated. *p* values determined by chi-square unless designated by (\*) then *p* value determined by ANOVA. **†** Data available to determine Acute Kidney Injury in 638 patients. **††** Data available to determine Chronic Kidney Disease was present in 638 patients.

#### *Primary Outcome*

Higher protein delivery in ICU survivors was associated with lower 90-day post-discharge mortality. In our mixed-effect logistic regression model adjusted for age, gender, race, Deyo-Charlson comorbidity index, energy delivery, nutrition status, acute organ failure, sepsis and the random-effects structure, the 90-day post-discharge mortality rate decreased by 18% for each 1 g/kg/day elevation in protein delivery following ICU admission (OR 0.82, 95% CI 0.73–0.92). Following stratification for nutrition status, the same multivariable mixed-effect logistic regression model shows the 90-day post-discharge mortality rate significantly decreased for each 1 g/kg elevation in daily protein delivery. Specifically, in patients diagnosed with malnutrition (*n* = 473), 90-day post-discharge mortality rate was 30% (95% CI: 6–26) less for each 1 g/kg elevation in daily protein delivery following ICU admission (OR = 0.70 (95% CI 0.61–0.81; *p* < 0.001)).

We next determined if protein delivery was associated with longer term mortality outcomes. When patients were evaluated using a multivariable mixed-effect logistic regression model with the same covariates and structure as the primary outcome model, 180-day post-discharge mortality was significantly reduced for each 1 g/kg elevation in daily protein delivery (OR 0.74, 95% CI 0.67–0.83; *p* < 0.001). When patients were evaluated using the multivariable mixed-effect logistic regression model with the same covariates and structure as the primary outcome model, the 365- and 720-day post-discharge mortality rate was significantly reduced for each 1 g/kg elevation in daily protein delivery (OR 0.76, 95% CI 0.69–0.83; *p* < 0.001), (OR 0.77, 95% CI 0.71–0.84; *p* < 0.001) respectively (Table 2, Figure 2). All estimates are for each 1 g/kg/day increase in protein delivery.


**Table 2.** Associations between Protein Delivery and post-hospital mortality outcomes.

All estimates were produced via a mixed-effect logistic regression model adjusted for age, gender, race, Deyo-Charlson comorbidity index, energy delivery, nutrition status, acute organ failure, sepsis and the random-effects structure.

**Figure 2.** Forest plot of post-discharge mortality associated for each 1 g/kg/day increase in protein delivery in our study cohort. Odds ratios with corresponding 95% CIs shown of the individual outcomes under study. <sup>a</sup> Restricted to patients diagnosed with Malnutrition.

Finally we examined the association between protein delivery and hospital length of stay. In a mixed-effect linear regression model adjusted for age, gender, race, Deyo-Charlson comorbidity index, energy delivery, nutrition status, acute organ failure, sepsis and the random-effects structure, the actual hospital length of stay (LOS) was reduced by 0.87 days (95% CI −1.7 to −0.10) compared with the average LOS for the DRG for each 1 g/kg/day increase in protein delivery (*p* = 0.028).

#### **4. Discussion**

In our study, we examined the association between protein delivery in critical illness survivors and post-hospital discharge outcomes. We demonstrate that improvement in daily protein delivery is associated with a significant decrease in the odds of post-discharge hospital mortality. We also show that patients with improved daily protein delivery have decreased hospital length of stay.

Long-term morbidity and mortality are substantial problems in ICU survivors [39]. Adverse events following discharge from an ICU admission are known to include comorbidity, illness

severity and organ failures [40]. In critical illness survivors, the risk factors for outside hospital adverse events are not well known. Our results show an association between higher daily protein delivery during hospitalization and improved post-discharge mortality in critical illness survivors. While our observational study cannot be interpreted as causal, the protein delivery-post-discharge outcome association has biologic plausibility.

Our study may have limitations due to its observational design. Causality is limited and as such our observations should be considered hypothesis generating. As protein intake in our cohort is only collected in patients considered to be at least at risk for malnutrition, ascertainment bias may be present. Unmeasured variables may influence outcomes following hospitalization independent of protein delivery, which can result in bias. We are unable to adjust for gut failure which associated with lower nutrition tolerance, lower protein intake and contributes to adverse outcomes [41]. Though ICD-9 code assignment data is collected from individual provider encounters, determination of covariates via ICD-9 codes may not completely capture the true incidence or prevalence. Finally, residual confounding may remain despite multivariable adjustment which may contribute to observed estimates.

Our study has several strengths. We utilize mixed-effects models to incorporate multiple measures and sources of protein intake over a total of 8735 days which allow for more complete information to be captured in parameter estimations. Mixed-effects models are ideal for such longitudinal data because each patient may have an unequal number of repeated measures. Additionally, a highly trained RD utilized weight loss history, clinical assessment, anthropometric metrics and protein intake to make an in person assessment of nutritional risk and daily protein delivery. Finally, the Master Death File is shown to capture post discharge mortality in our cohort accurately [21].

#### **5. Conclusions**

Our data show that in critically ill patients who survive hospitalization, higher daily protein delivery during hospitalization is associated with decreased mortality following hospital discharge. Increasing protein delivery following the acute phase of illness may be a strategy to assist with critical illness recovery and longer term outcomes.

**Author Contributions:** Conceptualization, P.J.M.W. and K.B.C.; Methodology, P.J.M.W. and K.B.C.; Software, K.B.C.; Validation, K.B.C.; Formal Analysis, K.B.C.; Investigation, P.J.M.W., K.M.M., and K.B.C.; Resources, K.M.M., J.D.R., and K.B.C.; Data Curation, J.D.R., and K.B.C.; Writing—Original Draft Preparation, P.J.M.W., K.M.M., and K.B.C.; Writing—Review & Editing, P.J.M.W., K.M.M., J.D.R., and K.B.C.; Supervision, K.B.C.; Project Administration, K.B.C.; Funding Acquisition, K.B.C.

**Funding:** This manuscript was funded by NIH R01GM115774.

**Acknowledgments:** This manuscript is dedicated to the memory of our dear friend and colleague Nathan Edward Hellman. The authors thank Shawn Murphy and Henry Chueh and the Partners Health Care Research Patient Data Registry group for facilitating use of their database.

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

#### **References**


© 2019 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/).
