*Article* **Factors Affecting Physical and Mental Fatigue among Female Hospital Nurses: The Korea Nurses' Health Study**

**Hee Jung Jang <sup>1</sup> , Oksoo Kim 2,3 , Sue Kim <sup>4</sup> , Mi Sun Kim <sup>5</sup> , Jung Ah Choi <sup>3</sup> , Bohye Kim <sup>2</sup> , Hyunju Dan <sup>2</sup> and Heeja Jung 6,\***


**Abstract:** Nurses often experience work-related physical and mental fatigue. This study sought to identify the levels of physical and mental fatigue present among Korean female nurses and discern factors influencing their onset. This cross-sectional study analyzed data from the Korea Nurses' Health Study (KNHS). A total of 14,839 hospital nurses were assessed by hierarchical regression analysis. The mean scores of physical and mental fatigue were 12.57 and 5.79 points, respectively. After adjusting for confounding variables, the work department had a significant influence on both physical and mental fatigue, that is, nurses working in special care units experienced greater degrees of both physical and mental fatigue than those working in general units. Nurse fatigue is an important consideration to monitor to ensure nurses' continued wellbeing as well as good patient safety levels. Therefore, it is necessary to establish a strategy to mitigate nursing fatigue while considering the characteristics of specific departments. In nursing practice, the introduction of a counseling program and guarantee of rest time that can alleviate the mental and physical fatigue of nurses working in special care units should be considered.

**Keywords:** physical fatigue; mental fatigue; female; nurses

#### **1. Introduction**

Nurses are required to perform appropriate nursing care and treatment at the frontlines of patient care and, as a result, often experience work-related physical and mental fatigue. According to a prior study, 50.2% of hospital nurses reported work-related chronic and acute fatigue and 84.9% of female nurses have experienced physical and mental fatigue [1,2]. These high proportions of physical and mental fatigue among nurses must be given attention as an important issue. Physical fatigue is caused by physical labor, such as long hours of standing, lifting, or changing the positions of patients, and appears to be a symptom of full-body discomfort and difficulty in tasks requiring strength [3]. In particular, healthcare professionals such as nurses encounter the risks of high exposure to posture-related harm, which can lead to musculoskeletal disorders and become a major factor of physical fatigue [4]. Mental fatigue is caused by work-related emotional stress such as patients' demands and expectations, which results in lethargy, decreased levels of concentration, or lack of motivation for work [3]. Both physical and mental fatigue negatively affect an individual's biological, psychological, and cognitive processes [5,6]. High levels of physical and mental fatigue affect nurses' personal health and health-promoting behavior [7]. Fatigue can also decrease nurses' work performance and may impair their

**Citation:** Jang, H.J.; Kim, O.; Kim, S.; Kim, M.S.; Choi, J.A.; Kim, B.; Dan, H.; Jung, H. Factors Affecting Physical and Mental Fatigue among Female Hospital Nurses: The Korea Nurses' Health Study. *Healthcare* **2021**, *9*, 201. https://doi.org/10.3390/ healthcare9020201

Academic Editors: Alberto Modenese and Fabriziomaria Gobba

Received: 6 January 2021 Accepted: 10 February 2021 Published: 13 February 2021

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

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

ability to perpetuate safe behaviors in the workplace [8,9]. Therefore, it can be argued that persistent fatigue is not just a nurse's personal problem but also an issue that directly affects patient safety and the quality of care.

According to previous studies, married nurses, those who work in shifts, and those who receive less income display a higher incidence of fatigue [5,10,11]. Nurses with a lower quality of sleep also reported more fatigue, whereas those without depressive symptoms and who had better-perceived health felt less fatigue [2,12]. Level of fatigue varies by work department as well. Emergency room (ER) nurses are more likely to experience high levels of fatigue relative to nurses working in other nursing departments [1,13]. This discrepancy is attributed to the characteristics of these respective special care departments, including the need to manage complicated and life-threatening medical problems, the existence of physically and mentally demanding job requirements, the persistence of increased pressure to perform well, and the use of complex technologies [1,5]. In addition, to the best of our knowledge, none of the previous studies on fatigue by workplace evaluated differences between physical and mental fatigue among nurses according to their work department. As such, there are limitations in identifying differences and in suggesting the relevance between physical and mental fatigue among departments. This study aimed to (1) identify the levels of physical and mental fatigue among Korean nurses and (2) investigate the factors affecting physical and mental fatigue.

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

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

This study was cross-sectional in design. We analyzed data from the Korea Nurses' Health Study (KNHS), which is described in detail elsewhere [14]. The KNHS is a largescale, prospective cohort study of female nurses that investigates the effects of occupational and lifestyle risk factors on Korean women's health. This study only included female nurses as participants due to the increasing demand to understand women's health issues, considering the dramatic changes in the fertility rate and social changes arising from the rapid economic growth in Korea. The study population of the KNHS included female hospital nurses of child-bearing aged between 20 and 45 years old. Data were collected online, through the KNHS website, from July 2013 to November 2014. Voluntary participation was encouraged by advertising the study on the KNHS website and social networking sites; furthermore, the research team visited the job training sessions held at the hospitals to encourage personnel to participate in the study.

A total of 20,613 nurses participated in module 1, the initial baseline survey, and 14,839 nurses who worked in general care departments and special care departments were included in the analysis for this study. Based on the specialty of the job and the traits of the working environment, the emergency room (ER), intensive care unit (ICU), operating room (OR), and post-anesthesia care unit (PACU) were classified as special care departments, while general wards (i.e., internal medicine, general surgery, and obstetrics and gynecology) and outpatient units were classified as general care departments.

#### *2.2. Measurements*

The questionnaires of KNHS were similar to those of the Nurses' Health Study 3 (NHS 3) performed in the United States (US) [14]. A multidisciplinary advisory committee translated the NHS 3 questionnaires, then modified or eliminated some questions to ensure relevance to Korean health research needs and accurately reflect cultural issues.

Fatigue, the primary variable of interest, was measured using the Chalder Fatigue Scale (CFS). This questionnaire consists of 11 items, where items 1 through 7 measure physical fatigue, while items 8 through 11 measure mental fatigue. Answers are assessed using a four-point Likert scale with zero points indicating less than usual, one point indicating no more than usual, two points indicating more than usual, and three points indicating much more than usual. Total physical fatigue scores can therefore range from zero to 21 points, while total mental fatigue scores range from zero to 12 points, with higher scores indicating more severe levels of fatigue in both cases. The Cronbach's alpha values of the original study were 0.85 (physical fatigue) and 0.82 (mental fatigue; 14) [15]. The Cronbach's alpha value of this study was 0.91.

Stress was measured using the four-item Perceived Stress Scale-4 (PSS-4). Responses were rated on a five-point Likert scale ranging from zero points indicating not at all to four points indicating very often. Possible scores ranged from zero to 16 points and higher total scores suggested the existence of higher levels of perceived stress in respondents. The Cronbach's alpha of the original study was 0.72 [16], while the Cronbach's alpha of the current study was 0.52.

Depressive symptoms were measured via the Patient Health Questionnaire-9 (PHQ-9), a self-reported instrument used to identify the severity of depressive symptoms [17]. The PHQ-9 consists of nine items with a total score that ranges from zero to 27 points. Scores of zero to four points, five to nine points, 10 to 14 points, 15 to 19 points, and 20 points or higher indicate minimal, mild, moderate, moderately severe, and severe depressive symptoms, respectively [17]. The Cronbach's alpha of the current study was 0.90, which is similar to that of the original paper (0.87).

We used the Jenkins Sleep Questionnaire to measure sleep disturbance [18]. This questionnaire consists of four items assessed using a six-Likert scale ranging from one point indicating not at all to six points indicating every night. Total scores here can range from four to 24 points and higher scores suggest the respondent is experiencing more severe sleep problems. The Cronbach's alpha of the original study was 0.79, while that of this study is 0.86 [18].

Participants were also asked to rate their health as good, fair, or poor. Separately, to investigate factors affecting nurses' fatigue, we adjusted for the following general covariates based on a literature review: general characteristics (i.e., age, level of education, marital status, annual income, shiftwork), body mass index (BMI), and sleep and psychological health (sleep problems, perceived health, level of depression, stress). All of these factors were included as potential confounding variables.

#### *2.3. Data Analysis*

The data were analyzed using the SPSS Statistics version 26.0 software program (IBM Corp., Armonk, NY, USA). Descriptive statistics were adopted to examine the frequencies and percentages, while Pearson's correlation coefficient was used to examine the associations among variables. Factors affecting the levels of physical and mental fatigue among female hospital nurses were analyzed via hierarchical regression analysis. The threshold for statistical significance was *p* < 0.05.

#### *2.4. Ethical Considerations*

This study received ethical approval from the Korea Centers for Disease Control and Prevention (2013-03CON-03-P). Anonymity was assured, and informed consent was obtained from the participants online.

#### **3. Results**

Table 1 presents the participants' general characteristics and the differences observed for each variable according to both physical and mental fatigue. Out of the 14,839 nurses, 62.9% were aged 29 years or younger, almost half had completed a four-year university or higher education program (50.9%), and most (70.2%) were single. Approximately half of the participants (42.1%) earned an annual income of lower than \$30,000 (USD), while 38.9% earned an annual income of between \$30,000 and \$39,999 USD. Most of the nurses worked shiftwork (78.3%) and 65.7% had normal BMI values. The majority of the nurses rated their health as either good (41.1%) or fair (48.7%); most of the nurses had no sleep problems (64.9%) and, with regard to depressive symptoms, most said they were experiencing minimal (32.9%) or mild (37.4%) levels of depressive symptoms, while only 3.7% reported severe depressive symptoms. The mean score of stress was 6.61 of 16 points. Considering

the work department, 62.0% worked in general units. The mean score of overall fatigue was 18.36 of 33 points, suggesting a moderate level of fatigue was prevalent throughout the study population. More specifically, the mean score of physical fatigue was 12.57 of 21.0 points, whereas that of mental fatigue was 5.79 of 12.0 points.


**Table 1.** General characteristics of participants (*N* = 14,839).

Note: BMI = body mass index; M = mean; SD = standard deviation.

The following parameters were statistically different according to mental fatigue (Table 1): age, marital status, annual income, shiftwork, perceived health, sleep problems, level of depression, and stress level. Meanwhile, the following parameters were statistically different according to physical fatigue: age, marital status, level of education, marital status, annual income, shiftwork, BMI, perceived health, sleep problems, level of depression, and stress level.

Table 2 displays the hierarchical multiple regression results, which were employed to discern which factors affect physical fatigue. As shown in model 3, the work department also significantly influenced physical fatigue (β = 0.027; *p* < 0.001). The adjusted R-square in the final model was 46.7% (F = 996.512; *p* < 0.001), revealing an increase of 42.7% in explanatory power compared with that of model 1. Age, marital status (being married), annual income (≤\$2999 USD), shiftwork, BMI, perceived health (both fair and poor), sleep problems, level of depression, and stress level were factors showing a significant association with physical fatigue.

**Table 2.** Hierarchical multiple regression analysis predicting physical fatigue (*N* = 14,839).


Note: BMI = body mass index; \*\*\* *p* < 0.001, \*\* *p* < 0.01.

Table 3 describes the results of hierarchical multiple regression, which were used to determine factors affecting mental fatigue (dependent variable). Socio-demographic factors and conduct of shiftwork were included in model 1. Health-related factors were then added in model 2. Finally, model 3 identified that differences between departments were statistically significant concerning mental fatigue. When the work department (independent variable) was added in the final model, the adjusted R-square was 36.5% (F = 651.819; *p* < 0.001), constituting an increase of 35.4% in explanatory power relative to model 1. The independent variable, work department, had significant influence on mental fatigue (β = 0.027; *p* < 0.001). Age, marital status (being married), shiftwork, BMI, perceived health (both fair and poor), sleep problems, level of depressive symptoms, and stress level were factors that had a significant relationship with mental fatigue.


**Table 3.** Hierarchical multiple regression analysis predicting mental fatigue (*N* = 14,839).

Note: BMI = body mass index; \*\*\* *p* < 0.001, \*\* *p* < 0.01.

#### **4. Discussion**

This study aimed to confirm the level of fatigue among Korean female nurses and to identify factors influencing the onset and worsening of physical and mental fatigue. The study results suggest that Korean female nurses show a moderate level of fatigue with average scores of physical fatigue and mental fatigue being 12.57 and 5.79 points, respectively. In particular, nurses working in special care departments showed higher levels of both physical and mental fatigue in comparison with nurses working in general care departments. In a previous study, nurses working in special care departments such as the ER, ICU, and OR reported higher levels of fatigue, which may be because these nurses must care for patients with greater severities of disease or injury and with higher levels of physical dependency [19]. Psychological and physical fatigue are predictors for medical administration errors. Psychological fatigue, above all, is related to incomplete or incorrect documentation of patients, negatively affecting the tasks performed by nurses [20]. Hence, the regular inspection of nurses' fatigue levels and the development of intervention plans could play important roles in maintaining patient care quality by enhancing nurses' work performance.

Hierarchical multiple regression was chosen to attempt to identify the factors influencing the physical and mental fatigue of nurses. Even after adjusting for confounding variables, work department had a significant influence on nurses' physical and mental fatigue. Some prior studies have reported that the work department is related to nurse fatigue although one study suggested the opposite—that the work department is not a factor influencing the fatigue of nurses [13,19,21]. Such a gap may have occurred due to different classifications of work departments in each study, making direct comparisons between them more difficult. One study suggested that nursing care models may have a bigger impact than the work unit itself on the chronic occupational fatigue of nurses. In particular, nurses in a total patient care model wherein primary nurses perform many functions express a higher level of fatigue than nurses working in a functional nursing care

model, where several nurses are given one or two assignments [21]. In this study, as we did not investigate the nursing care type or model of each department, the interpretation of data may be limited in some respects. However, when the results of prior studies are considered, we can infer that the work department is presented as a predictor of fatigue as each department boasts unique degrees of direct nursing care and physically or mentally demanding tasks. A study on physical activity and the level of fatigue of pediatric nurses working in the pediatric ICU and OR found that more than 50% of the physical activity performed by nurses in special care departments are related to their work [22]. Nurses working in the OR and anesthesia recovery room are also required to set up complicated equipment and often have to work while standing up [23]. Such results show that nurses in special care departments have a higher burden of physical activity, which can contribute to physical fatigue. Moreover, pediatric nurses in special care departments have also reported that they experience higher levels of fatigue from perpetuating interpersonal relationships [22]. Hence, the severity of physical work is considered to be significantly related to not only physical fatigue but also to mental fatigue.

According to a prior study on work-related factors that provoke physical and mental fatigue among nurses, inadequate time available and competing task demands are the factors that most frequently cause mental fatigue. Physical demand tasks, including lifting, pushing, and carrying, were factors that most frequently lead to physical fatigue [24]. Additionally, challenging working conditions, such as the hierarchies among healthcare workers, task orientation, and inflexible divisions of labor, result in occupational health problems among workers [25]. Hence, to reduce both the physical and mental fatigue of nurses, it is important to identify the nature and demands of the work performed within each department. However, in this study, the levels of physical and mental demands or burdens were not investigated and work units such as the ICU, ER, and OR were classified collectively as special departments. Future studies focusing on and further delineating the characteristics and influences of work according to individual work units would be beneficial.

One of the controlled variables, shiftwork, was also identified as a factor affecting mental and physical fatigue. While one study found that nurses working in shifts experience a high level of fatigue, there is also conflicting evidence concerning the nature of the relationship between nurses' shiftwork and fatigue. In a study of nurses working in four centers (pediatric, maternity, general, and emergency) at multispecialty hospitals, there was no meaningful relationship established between shiftwork and fatigue [20,26]. Elsewhere, in a literature review of work schedule characteristics and the fatigue profiles of nurses, an insufficient resting period had a more significant relationship with the level of fatigue relative to the number of night or evening shifts [27]. As such, the relationship between shiftwork and fatigue needs to be reviewed further with consideration of the role of rest periods among nurses working in shifts.

In addition, in this study, the level of depression was found to be an influencing factor for mental and physical fatigue. Few studies have investigated this relationship between the level of depression and fatigue among nurses, despite the higher levels of depressive symptoms exhibited by nurses, compared to other occupations, due to working in shifts [28]. Prior studies on college students revealed that those with moderate or severe fatigue showed significantly higher depressive symptoms and scored higher on a suicide risk measure than those with mild or no fatigue and those with mild fatigue, respectively [29]. Therefore, it is considered that proper management of depressive symptoms among nurses can have a relieving effect on mental and physical fatigue.

Finally, in this study, women in their 20s showed higher mental and physical fatigue than those in their 30s. These results indicate that mental and physical fatigue can be reduced through job adaptation as age and experience increase. However, in this study, nurses' work experience was not included as a covariate in the analysis; therefore, its relationship with fatigue could not be identified. Thus, it is necessary to understand the relationship between nurses' work experience and fatigue in future studies.

Identifying differences in factors that contribute to fatigue in nurses and providing appropriate interventions are important to effectively maintain nurses' health and reduce nurse turnover, especially in special care departments requiring longer staff training periods. Prior studies have suggested the effectiveness of relaxation exercise using deep inspiration and pursed-lip breathing techniques among ER nurses and the conduct of higher levels of physical activity contributing to a decrease in fatigue scores among pediatric nurses in a special care unit, respectively [22,30].

This study is limited due to its cross-sectional nature, which does not allow for an explanation of causal relationships among variables. However, as the KNHS is ongoing, future analyses of collected cohort data are anticipated to potentially shed more light on nurses' fatigue. Other limitations include not being able to control for direct variables (e.g., time spent standing up and the number of heavy objects lifted) when investigating differences in fatigue, according to work department, and selection issues. As this study only analyzed nurses working in internal medicine, general surgery, obstetrics and gynecology, outpatient units, ER, ICU, OR, and post-anesthesia care units, care is warranted when pursuing interpretations for nurses in other work areas. Finally, despite these limitations, the large sample size and wide participation of nurses nationwide provide significant data for understanding nurses' fatigue.

#### **5. Conclusions**

Fatigue in nurses can eventually lead to burn out and start a vicious cycle of deteriorating patient care due to turnover among nurses and increased workload for their colleagues. This potential lack of skilled nurses can be particularly detrimental to the services of special care departments. However, efforts to reduce the fatigue of these nurses have been insufficient. Practical policy measures to assess and mitigate fatigue in nurses are required, not only for nurses' wellbeing but also for patient safety. Moreover, it should be considered especially important in the ongoing Coronavirus 2019 (COVID-19) pandemic.

**Author Contributions:** Formal analysis, B.K. and H.D; Methodology, M.S.K., J.A.C. and H.J.; Project administration, O.K.; Supervision, H.J.J., O.K. and S.K.; Writing—original draft, H.J.J., O.K., S.K., M.S.K., J.A.C., B.K., H.D. and H.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Korea Centers for Disease Control and Prevention (KCDC) (nos. 2013E6300600 and 2013E6300601).

**Institutional Review Board Statement:** Approval of the research protocol: The study received ethical approval from the Korea Centers for Disease Control and Prevention (2013-03CON-03-P). All procedures performed in this study were done in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

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

**Data Availability Statement:** The datasets generated and analyzed during the current study are not publicly available because this government data needs time for data clearing and establishment of guidelines. The Korea Centers for Disease Control and Prevention is planning on opening this data to the public in the future.

**Conflicts of Interest:** The authors declare that they have no conflict of interest to report.

#### **References**


## *Article* **Palliative Care Professionals' Inner Lives: Cross-Cultural Application of the Awareness Model of Self-Care**

**Amparo Oliver <sup>1</sup> , Laura Galiana <sup>1</sup> , Gustavo de Simone <sup>2</sup> , José M. Tomás 1 , Fernanda Arena <sup>3</sup> , Juan Linzitto <sup>2</sup> , Gladys Grance <sup>2</sup> and Noemí Sansó 4,5,\***


**Abstract:** Compassionate professional qualities traditionally have not received the most attention in either critical or end of life care. Constant exposure to death, time pressure and workload, inadequate coping with personal emotions, grieving, and depression urge the development of an inner curricula of competences to promote professional quality of life and compassionate care. The COVID-19 pandemic highlights the universality of these problems and the need to equip ourselves with rigorously validated measurement and monitoring approaches that allow for unbiased comparisons. The main objective of this study was to offer evidence on the generalizability of the awareness model of self-care across three care systems under particular idiosyncrasy. Regarding the sample, 817 palliative care professionals from Spain, Argentina, and Brazil participated in this cross-sectional study using a multigroup structural equation modeling strategy. The measures showed good reliability in the three countries. When testing the multigroup model against the configural and constrained models, the assumptions were fulfilled, and only two relationships of the model revealed differences among contexts. The hypotheses posited by the awareness model of self-care were supported and a similar predictive power on the professional quality of life dimensions was found. Self-care, awareness, and coping with death were competences that remained outstanding no matter the country, resulting in optimism about the possibility of acting with more integrative approaches and campaigns by international policy-makers with the consensus of world healthcare organizations.

**Keywords:** compassionate care; compassion satisfaction; compassion fatigue; cross-cultural comparison

#### **1. Introduction**

Person-centered care, as a caring philosophy, holds that there is no appropriate healthcare unless it is compassionate [1]. Compassion or "suffering with"[2] has been defined as "a virtuous response that seeks to address the suffering and needs of a person through relational understanding and action"[3]. Moreover, kindness and equanimity are essential qualities in those who care for the dying. However, there is currently a great concern that these compassionate qualities are not always present in the care of the dying [3–5]. International studies have highlighted important levels of compassion fatigue in healthcare professionals in general [6–9], and in palliative care professionals in particular [10,11]. Specifically, in the Spanish context, 69% of nurses and 77% of physicians had, either firsthand or through close colleagues, experienced being the second victim within the following five years [12].

The latest literature focusing on person-centered care delivery considers the preferences, needs, and values of the receivers of these services [13–18]. The difficulty in

**Citation:** Oliver, A.; Galiana, L.; de Simone, G.; Tomás, J.M.; Arena, F.; Linzitto, J.; Grance, G.; Sansó, N. Palliative Care Professionals' Inner Lives: Cross-Cultural Application of the Awareness Model of Self-Care. *Healthcare* **2021**, *9*, 81. https://doi.org/10.3390/ healthcare9010081

Received: 27 November 2020 Accepted: 12 January 2021 Published: 15 January 2021

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

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

compassionate care is related to several stressors that affect palliative care professionals, such as exposure to death, inadequate time to deal with patients, growing workload and communication difficulties with patients and their families, or inadequate coping with their own emotional response to grieving, depression, and guilt [19,20]. Compassion is also linked to protective factors, such as training in emotional management and spirituality [21–25], self-care [26–29], empathy [30], awareness [31–36], or competency and attitudes toward death [37].

Attending the literature on compassion protectors, specifically based on Kearney and Weininger's awareness model of self-care [38], Sansó et al. [39] tested a mapping model with variables involved in palliative care professionals' quality of life: compassion satisfaction (CS), compassion fatigue (CF), and burnout (BO). The other variables included were self-care and awareness, which were positively related to coping with death. Coping together with awareness were posited as being promoting factors of CS due to their positive relationship, and both of them were shown to work as protectors, given their negative relationship with CF and BO [39].

This study, initially carried out in a national sample of Spanish palliative care professionals, is, to the best of our knowledge, the only attempt to study, in a multivariate framework, the inner life of palliative caregivers. This is of paramount importance, as keeping their equanimity, cultivating compassion, and also developing a strengthened sense of vocation and job satisfaction are recognized as key issues in the healing process [39]. However, a major constraint of this study was its rather specific European context, that is, the Spanish one. In order to strengthen the understanding of the stable effects, as well as those that are specific to different cultural contexts, this research was expanded to include other countries, namely, Argentina and Brazil. Finding the dimensions that are protective for the caregiver, the strength they possess, and the specific international circumstances under which they work may guide national policy-makers to make educated improvements to enhance the ability of professional caregivers to provide compassion.

There were two main reasons for the selection of the abovementioned countries. On the one hand, we wanted to test the generalizability of the awareness model of self-care approaching the relationship between specific aspects of professionals' inner lives through the evaluation of an adaptation of Kearney and colleagues' awareness model [19,38] in non-European countries. Argentina, while having a very different healthcare system, is one of the few South American countries where hospice and palliative care is widely provided throughout the country [40,41]. On the other hand, it was also of interest to test the model in other countries in which palliative care is still facing several obstacles (e.g., funding, establishment of inclusion criteria, and treatment discontinuity) before its complete integration, such as Brazil [40,42]. This country has a short history of palliative care and many of its services have been recently founded, mainly offering palliative care in hospitals [40,41,43]. Thus, both Argentina and Brazil differ from the Spanish palliative care system. Argentina has achieved "a measure of integration with other mainstream service providers together with wider policy recognition," as established by Clark and Wright [44].

In this context, the aim of this research was to test the awareness model of self-care, which integrates background and protective variables to explain professionals' inner lives in terms of quality of life in different countries with different idiosyncratic characteristics in their palliative care attention. For this aim, a multigroup structural equation modeling strategy was used. Compared with other analyses, such as regression analysis, path analysis generates a pictorial representation, which facilitates the interpretation of the model and the hypotheses for the reader, provides means to distinguish effects of one variable from another, and permits standardized errors of the observed variables [45]. In a multigroup context, that is, when studies involve more than one group or population, relationships can vary across these groups, and multigroup models can be used to examine such population heterogeneity [46]. These models study whether the observed variables remain unchanged in different populations—in our case, in the professionals of different countries. The test of equality or invariance of path coefficients across groups enables us

to examine similar behavior across groups [47], and therefore, to potentially generalize theories from one group to another. theories from one group to another. Our hypotheses are based on Kearney and Weininger's model [38], whose empirical evidence thus far only exists for Spain [39]:

from another, and permits standardized errors of the observed variables [45]. In a multigroup context, that is, when studies involve more than one group or population, relationships can vary across these groups, and multigroup models can be used to examine such population heterogeneity [46]. These models study whether the observed variables remain unchanged in different populations—in our case, in the professionals of different countries. The test of equality or invariance of path coefficients across groups enables us to examine similar behavior across groups [47], and therefore, to potentially generalize

Our hypotheses are based on Kearney and Weininger's model [38], whose empirical evidence thus far only exists for Spain [39]: 1. Competence in coping with death and awareness will be positive predictors of CS


*Healthcare* **2021**, *9*, 81 3 of 12

**Figure 1.** A priori structural model, based on an adaptation of Kearney and Weininger's model [38], validated by Sansó et al. [39] for the Spanish context. **Figure 1.** A priori structural model, based on an adaptation of Kearney and Weininger's model [38], validated by Sansó et al. [39] for the Spanish context.

#### **2. Materials and Methods**  *2.1. Design, Procedure, and Participants*  **2. Materials and Methods**

#### *2.1. Design, Procedure, and Participants*

The cross-sectional surveys of Spanish, Argentinian, and Brazilian palliative care professionals were conducted between 2013 and 2016. Prior to these surveys, the research protocols were approved by the ethics committees of the professional associations. Members from the Spanish Society for Palliative Care (Spain), the Brazilian National Academy of Palliative Care (Brazil), and the Pallium Latinoamérica Institute, Argentine Association of Medicine and Palliative Care and the National Institute of Cancer (Argentina) were encouraged to participate. The cross-sectional surveys of Spanish, Argentinian, and Brazilian palliative care professionals were conducted between 2013 and 2016. Prior to these surveys, the research protocols were approved by the ethics committees of the professional associations. Members from the Spanish Society for Palliative Care (Spain), the Brazilian National Academy of Palliative Care (Brazil), and the Pallium Latinoamérica Institute, Argentine Association of Medicine and Palliative Care and the National Institute of Cancer (Argentina) were encouraged to participate.

Data were collected using a secure and anonymous online platform, with participation being voluntary and requiring the responders' informed consent. Regarding the responses, 385 professionals completed the survey in Spain, 271 in Argentina, and 161 in Brazil. The participants' characteristics are described in Table 1. Data were collected using a secure and anonymous online platform, with participation being voluntary and requiring the responders' informed consent. Regarding the responses, 385 professionals completed the survey in Spain, 271 in Argentina, and 161 in Brazil. The participants' characteristics are described in Table 1.


**Table 1.** Sociodemographic characteristics of the participants.

Notes: M = mean; SD = standard deviation.

There were statistically significant differences among the countries in terms of the mean age (*F*(2.786) = 54.589, *p* < 0.001, *η*2 = 0.12), sex (*χ* 2 (2) = 8.674, *p* = 0.013, Cramer's *V* = 0.104), and profession (*χ* 2 (10) = 89.331, *p* < 0.001, Cramer's *V* = 0.233) distribution across samples.

#### *2.2. Outcomes*

Data were collected using the following measures (internal consistency can be consulted in Table 2):




Notes: S = Spain; A = Argentina; B = Brazil.

#### *2.3. Data Analysis*

The structural models were tested in MPLUS version 8 [56] with maximum likelihoodrobust estimation, given the lack of multivariate normality. Firstly, the a priori theoretical model [39] was estimated in the three samples (see Figure 1). Once an adequate fit was obtained for each individual sample, multigroup structural models were stablished in order to test for differences between countries. A multisample strategy was used to test the generalizability of the relationships. The multigroup sequence of models started with a configural or baseline model that had the same relationships but no constraints across groups. Then, a second multisample model was estimated, with all of the structural coefficients in the path model constrained to equality (constrained model). If this constrained model fit the data as well as the baseline model, this would indicate no differences between the samples or, in other words, no moderation effects due to the country. If potential interaction (moderation) effects were found, the modification indices of MPLUS were then used to test the adequacy of releasing each imposed constraint.

Model fit was assessed with chi-square, Comparative Fit Index (CFI), Standardized Root Mean Residual (SRMR), and Root Mean Square Error of Approximation (RMSEA). The following cut-off values were used to determine good fit: CFI above 0.90 and SRMR or RMSEA below 0.08. As a multisample context was used, the models were also comparatively assessed using the chi-square difference test (with no statistical differences meaning preference for the most constrained model) and CFI differences (with differences of 0.05 or less considered negligible) [57].

Missing data were dealt with using the full information maximum likelihood (FIML), which is adequate for both missing completely at random (MCAR) and missing at random (MAR) data and is the most recommended method for structural models [58].

We used the STROBE cross-sectional checklist when writing our report [59]. The research protocol received ethical approval from the Pallium Latinoamérica Institute (code 210316).

#### **3. Results**

Descriptive statistics for the variables included in the awareness model of self-care can be consulted in Table 2. In general, the means were medium–high for self-care, with higher means for the Spanish group. Moreover, high levels of awareness were found, with higher means for the Argentinian professionals. High levels of coping with death were also found, with higher scores for the Spanish professionals, as were high levels of compassion satisfaction, with higher levels for the Brazilian group. Lastly low–medium levels of compassion fatigue and burnout were found, with higher levels of compassion fatigue for the Spanish and Argentinian samples, and higher levels of burnout for the Brazilian professionals.

The model was independently tested in the samples and fit indices were adequate (see Table 3). Regarding the RMSEA, its performance has proved to be poor in small samples (as in the Brazilian case) and in models with small degrees of freedom, such as the tested


model (six degrees of freedom) [60]; however, our appreciation of the overall goodness of fit of the three samples should not change, despite this particular value.


Notes: CFI = Comparative Fit Index; RMSEA = Root Mean Square Error of Approximation; RMSEA CI = RMSEA 90% confidence interval; SRMR = Standardized Root Mean Residual.

> Once the adequacy of the model in each sample was established, the baseline model was tested. This model had no constraints across groups; all of the parameters were freely estimated and simultaneously tested in the three samples. This model showed a good fit (Table 3). Then, a model with all of the parameters constrained across the three samples was estimated, i.e., the fully constrained model. This model was the most parsimonious one, as only the Spanish sample was used for the estimation, whereas the estimates for the Argentinian and Brazilian samples were fixed to these first estimates. As Table 3 shows, the model fit was degraded: using a statistical criterion, the chi-square differences were statistically significant; using a subjective criterion, the CFI differences were within the limit put forward by Little [57]. These results provide evidence of some moderation effects of the country. In order to study these effects, modification indices were considered and the relationships penalizing the model's fit were released.

> The modification indices pointed to two constrains that, when released, improved the model's fit: the effect of specific training on coping with death in the Spanish sample and the relationship between CS and BO in the Brazilian sample. After releasing these constraints, the chi-square of this last model showed no statistically significant differences to the baseline model, as well as an irrelevant difference of 0.003 between CFIs (see Table 3). Consequently, the model was retained as the most parsimonious one.

> The parameter estimates offered evidence of a moderation effect on the relationship posed between specific training and coping with death for the Spanish palliative care professionals. As shown in Figure 1, when compared to the other countries, Spanish palliative care professionals' specific training had no effect on coping with death, whereas this training had a positive effect for both Argentinian and Brazilian professionals. As regards the second released parameter regarding the relationship between CS and BO in the Brazilian sample, the estimates pointed to a greater relationship in this sample when compared to the Spanish and Argentinian professionals. Both estimates were, however, negative and statistically significant, as hypothesized. All parameter estimates, either invariant or variant, are shown in Figure 2.

**Figure 2.** The most parsimonious model with the standardized parameter estimates. Notes: \* *p* < 0.050 and \*\* *p* < 0.001. The non-invariant parameter estimates are those that are duplicated. The first value in the relationship between specific training and coping with death refers to the Spanish sample; the second is that of the Argentinian and Brazilian samples. The first value in the relationship between compassion satisfaction and burnout is for the Spanish and Argentinian samples; **Figure 2.** The most parsimonious model with the standardized parameter estimates. Notes: \* *p* < 0.050 and \*\* *p* < 0.001. The non-invariant parameter estimates are those that are duplicated. The first value in the relationship between specific training and coping with death refers to the Spanish sample; the second is that of the Argentinian and Brazilian samples. The first value in the relationship between compassion satisfaction and burnout is for the Spanish and Argentinian samples; the second one is for the Brazilian sample. For the sake of clarity, standard errors are not shown.

the second one is for the Brazilian sample. For the sake of clarity, standard errors are not shown. Another important result was the considerable and homogeneous amount of variance explained by the most parsimonious model across the three sociocultural contexts (see Table 4). The variance for coping with death ranged from 11.5% (Spain) to 20.8% (Argentina). The protective variables allowed for almost 25% of the prediction of BO, no matter the country. When focusing on countries, a higher explicative power was reached for Another important result was the considerable and homogeneous amount of variance explained by the most parsimonious model across the three sociocultural contexts (see Table 4). The variance for coping with death ranged from 11.5% (Spain) to 20.8% (Argentina). The protective variables allowed for almost 25% of the prediction of BO, no matter the country. When focusing on countries, a higher explicative power was reached for Argentina.

> Argentina. **Table 4.** Variance accounted for by the coping with death and professional quality of life dimensions across countries based on the R<sup>2</sup> values.


Burnout 0.243 0.244 0.244 **Notes:** Results from the best fit, most parsimonious model.

#### Compassion fatigue 0.162 0.194 0.186 **4. Discussion**

**Notes:** Results from the best fit, most parsimonious model. **4. Discussion**  Person-centered palliative care is largely based on the attention of compassionate professionals. Despite its practical relevance, the recent literature claims compassionate qualities are not always present in professionals when working with patients at the end Person-centered palliative care is largely based on the attention of compassionate professionals. Despite its practical relevance, the recent literature claims compassionate qualities are not always present in professionals when working with patients at the end of their life and their families [3–5]. In the last decade, few theoretical approaches have tried to explain the reasons for this lack of professional competence [19,20,34], and empirical evidence based on these models, although robust, is yet limited to a particular European healthcare system [39].

of their life and their families [3–5]. In the last decade, few theoretical approaches have tried to explain the reasons for this lack of professional competence [19,20,34], and empirical evidence based on these models, although robust, is yet limited to a particular European healthcare system [39]. The aim of the present research was to investigate the generalizability of the model tested by Sansó et al. [39] in Spanish professionals of palliative care in two additional countries, namely, Argentina and Brazil. By testing a multigroup model, we evidenced the different effects of one variable on another and how these effects vary across our studied groups [46], and we pointed to generalizations in behavior patterns across populations The aim of the present research was to investigate the generalizability of the model tested by Sansó et al. [39] in Spanish professionals of palliative care in two additional countries, namely, Argentina and Brazil. By testing a multigroup model, we evidenced the different effects of one variable on another and how these effects vary across our studied groups [46], and we pointed to generalizations in behavior patterns across populations [47]. For such generalization purposes, we used evidence gathered in the previous literature. The model was mostly based on Sansó et al.'s work, although it included some improvements regarding professionals' inner life appraisal: self-care was assessed with all of the items of the Professionals' Self-Care Scale, and awareness was assessed with a shorter and more discriminant measure [50].

[47]. For such generalization purposes, we used evidence gathered in the previous literature. The model was mostly based on Sansó et al.'s work, although it included some im-

of the items of the Professionals' Self-Care Scale, and awareness was assessed with a

shorter and more discriminant measure [50].

The results supported Hypothesis 1: "Competence in coping with death and awareness will be positive predictors of compassion satisfaction and negative predictors of compassion fatigue and burnout." Both competence in coping with death and awareness promoted higher levels of compassion satisfaction and worked as protectors of compassion fatigue and burnout, with negative relationships with these two last constructs. These two relationships, competence in coping with death and quality of professional life and awareness with professional quality of life, have been well documented in the literature [34,37], although this is the first time they have been tested in several countries.

Regarding Hypothesis 2, "Having participated in training programs aimed at facing suffering and death, self-care and awareness will positively predict coping with death, and indirectly will predict professionals' quality of life (through a mediator effect of coping)," the results provided evidence on all of the assumed relationships, particularly between specific training and coping with death. The findings revealed that, while in Brazil and Argentina this relationship is significant, it is not in Spain. This lack of an effect of specific training on coping with death was already found in the Spanish sample studied by Sansó et al. [39]. Although the indicator used was the same in the three countries, "Have you done specific training to face suffering and death?," a possible explanation of the absence of an effect in the Spanish context could be the amount of courses healthcare professionals attend in this country. It is common for palliative care professionals to engage in a vast amount of training throughout their professional lives. This, together with the fact that we investigated an "older" sample, especially in terms of professional experience, could have made the question less discriminant in Spain. An additional result was the one offered by Hypothesis 2a, "These three variables will show positive relationships among one another," which was supported across the countries.

Finally, Hypothesis 3, "The dimensions of the professionals' quality of life, that is, compassion satisfaction, compassion fatigue, and burnout, will be interrelated. Burnout will be negatively related to compassion satisfaction and positively related to compassion fatigue, whereas compassion satisfaction and fatigue will be independent," was also sustained by the model, which also offered additional interesting context-dependent information. There was a stronger relationship between compassion satisfaction and burnout in Brazil compared to the other countries.

To summarize, our results highlighted the model's generalizability, showing that the key elements of professionals' inner lives, such as self-care, awareness, or coping with death, are competences that remain outstanding no matter the country, which suggests the convenience of being universally encouraged. On the contrary, two relationships could not be generalized: the lack of a predictive effect of specific training in the Spanish context of palliative care, and the negative relationship between compassion satisfaction and burnout, which was stronger in Brazil than in Spain and Argentina.

The Global Atlas for Palliative Care [61] indicates higher rates for adults in need of palliative care at the end of their life in the European and Western Pacific regions. Latin American countries show lower rates. Indeed, European and Western Pacific professionals of palliative care work with elderly patients, in comparison to Latin American professionals, where the end of life is a more natural path for younger professionals. This, however, did not affect the majority of the relationships specified in the current research.

The maturity of the palliative care system is another characteristic that could explain differences in the functioning of the model. The biggest variance accounted for by coping with death, satisfaction and fatigue compassion, and burnout, being explained by protectors in Argentina, could be partially understood by their major efforts in developing palliative professionals' inner curricula during the last decade. In addition, the Argentinean palliative system has encouraged specific training due to the role played by the *Pallium Latinoamérica* Institute [62]. If we focus on Latin America, clear differences arise in the palliative care contexts, as Argentinian palliative institutions emerged in the early 1980s, whereas in Brazil, they did not emerge until the late 1990s, with the main association (*Academia Nacional de Cuidados Palitivos,* ANCP) being created in 2005 [41]. Chile, Costa Rica, Argentina, and

Uruguay pioneered palliative care in this area; Brazil, and other countries such as Colombia, Mexico, and Paraguay, are in a medium state of development, while countries such as Honduras, Nicaragua, and Bolivia are the most delayed in this development. The Brazilian palliative care context is especially interesting for three main reasons: (a) professionals work with younger patients than in Europe; (b) they work in a context of great care discontinuity, as home care initiatives are not integrated in primary healthcare services [43], as it is the case in Spain; (c) caregivers' quality of life is strongly affected by the difficulties in home care and work overload because not only do professionals provide medical assistance in hospitals, but they also have to work together with the home-care team [63,64]. A more mature palliative care system would bring higher funding, more specific inclusion criteria, treatment continuity, better integration with other mainstream services, and wider policy recognition for those countries with a great tradition in this care. Moreover, the models did not significantly differ, and thus factors protecting professionals from burnout and compassion fatigue and promoting compassion satisfaction seem non-dependent on how well-established the provision of palliative care is.

Regarding the practical implications of this study, the findings evidence the fact that the practice of self-care, the development of awareness, and specific training enhance professionals' inner lives, directly influencing their quality of life and likely the quality of their caregiving. Working on the variables that increase professionals' quality of life, a double objective can be achieved: professional wellbeing can be improved (understood as the presence of high compassion satisfaction and low burnout and compassion fatigue), and professionals' efficacy as healing agents in the palliative care encounter can be optimized through an enhanced ability to use themselves as healing agents in clinical encounters [32].

This study presents some limitations to bear in mind. The first limitation is the low response rate of this kind of study, with a non-incentivized self-report questionnaire. Despite such difficulties, the sample size obtained provided a robust dataset to explore the validity of the awareness model of self-care in different countries with different idiosyncratic characteristics in their palliative care attention. Secondly, it is worth noting that the possibility of response bias is present. To reduce the likelihood of such a bias, the respondents were informed that the research was anonymous.

#### **5. Conclusions**

This study highlights both the similarities and differences across palliative care professionals of different populations. Such similarities in behavior patterns have been assumed many times but were tested in this study for the first time. Therefore, this study offers evidence of the ability to generalize scientific evidence, including the importance of selfcare, awareness, and coping with death for palliative care professionals in different parts of the world.

In conclusion, the contribution of this work is its provision of the first cross-cultural evidence (including two languages and three countries) on the suitability of a comprehensive model to address the relationship between protectors and quality of work life, as well as its quantification of the relationships in the model so that policy-makers can prioritize actions. The benefits from recent interventions in contexts, such as palliative care, with high emotional demands to promote professional quality of life are very encouraging [65,66] and are well structured [67].

In light of our results, even when healthcare systems are not mimetic and show great differences, the protectors of professionals' quality of life are the same and have the same quantitative effect. That is, the model is generalizable across countries and health systems. This is of special importance, taking into account that preventing burnout and compassion fatigue and enhancing compassion satisfaction are a requisite for both the quality of patients' care and occupational safety. Compassion is key to meeting patients' needs, including those on the surface as well as those kept more hidden, and is also crucial for institutional benefits. Compassionate professionals are able to work more and work to a better standard, and, most importantly, can provide more and better-quality care. Thus, interventions attending to the predictors of professionals' quality of life, such as mindfulness-based stress reduction interventions or compassion-based training, must be on the agenda of world health agencies and policy-makers from now on.

**Author Contributions:** A.O., L.G., N.S., J.L., G.G. and G.d.S. designed the research, to which the rest of the authors contributed. A.O., L.G., J.M.T. and F.A. contributed to the data analysis. A.O., L.G. and N.S. wrote the first version of the manuscript, which was critically reviewed by all signatories, who approved the final version. All authors contributed to the interpretation of the results and a critical review of the manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** F.A. was awarded by Brazilian CNP and J.P.L. by Argentinian National Institute of Cancer during the course of this research. This research was supported by Project RTI2018-094089-I00: Longitudinal study of compassion and other Professional Quality of Life determinants: A national level research on palliative care professionals (CompPal) [Estudio longitudinal de la compasion y otros determinantes de la calidad de vida profesional: Una investigacion en profesionales de cuidados paliativos a nivel nacional (Comp-Pal)] (Ministerio de Ciencia e Innovacion—Agencia Estatal de Investigacion/FEDER).

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board of Pallium Latinoamérica Institute (code 210316).

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

**Data Availability Statement:** The data that support the findings of this study are available from the corresponding author, Noemí Sansó, upon reasonable request.

**Acknowledgments:** The authors thank all of the participants and palliative care professional associations' representatives in Argentina, Brazil, and Spain for their valuable support. They also give thanks to Enric Benito for encouraging this research, and Stamm and The Center for Victims of Torture (www.cvt.org) for the use of the ProQOL measure.

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

#### **References**


## *Article* **Predictors of Health-Related Quality of Life among Healthcare Workers in the Context of Health System Strengthening in Kenya**

**Rose Nabi Deborah Karimi Muthuri 1,\* , Flavia Senkubuge <sup>1</sup> and Charles Hongoro 1,2,3,4**

	- Gauteng Province, South Africa

**Abstract:** Kenya is among the countries with an acute shortage of skilled health workers. There have been recurrent health worker strikes in Kenya due to several issues, some of which directly or indirectly affect their health. The purpose of this study was to investigate the predictors of health-related quality of life (HRQOL) among healthcare workers in public and mission hospitals in Meru County, Kenya. A cross-sectional study design was undertaken among 553 healthcare workers across 24 hospitals in Meru County. The participants completed the EuroQol-five dimension-five level (EQ-5D-5L) instrument, which measures health status across five dimensions and the overall self-assessment of health status on a visual analogue scale (EQ-VAS). Approximately 66.55% of the healthcare workers reported no problems (i.e., 11,111) across the five dimensions. The six predictors of HRQOL among the healthcare workers were hospital ownership (*p* < 0.05), age (*p* < 0.05), income (*p* < 0.01), availability of water for handwashing (*p* < 0.05), presence of risk in using a toilet facility (*p* < 0.05), and overall safety of hospital work environment (*p* < 0.05). Personal, job-related attributes and work environment characteristics are significant predictors of healthcare workers HRQOL. Thus, these factors ought to be considered by health policymakers and managers when developing and implementing policies and programs aimed at promoting HRQOL among healthcare workers.

**Keywords:** health-related quality of life; health measurement; work environment; healthcare workers; health systems

#### **1. Introduction**

In 2014, the United Nations (UN) General Assembly consisting of 193 Heads of State, universally adopted resolution A/RES/70/1 on 'Transforming our world: the 2030 Agenda for Sustainable Development' that envisions "A world with equitable and universal access . . . to health care and social protection, where physical, mental and social well-being are assured" (p. 3) [1]. The resolution contains 17 Sustainable Development Goals (SDGs) and 169 targets. SDG 3 emphasizes healthy lives for all persons of all ages [1]. The sixtyninth World Health Assembly (WHA) stated that SDG 3, and other health development agendas, could not be achieved without investing in and improving the health workforce [2]. According to a World Health Organization (WHO) report titled, "A universal truth: no health without a workforce" [3], it is paramount to put the health workforce at the center of health policy discourse aimed at strengthening health systems, improving public health outcomes, and achieving health development agendas [3]. The health workforce plays a pivotal role in the achievement of global health agendas such as Universal Health Coverage and SDG 3 by 2030 [3]. However, one of the most significant challenges in the Kenyan health system is a critical shortage of skilled healthcare workers [4].

**Citation:** Muthuri, R.N.D.K.; Senkubuge, F.; Hongoro, C. Predictors of Health-Related Quality of Life among Healthcare Workers in the Context of Health System Strengthening in Kenya. *Healthcare* **2021**, *9*, 18. https://dx.doi.org/ 10.3390/healthcare9010018

Received: 11 December 2020 Accepted: 22 December 2020 Published: 25 December 2020

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

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

The WHO recommended minimum threshold of skilled health workers is 2.3 per 1000 population [4]. By 2006, the skilled health workers' density in Kenya was 1.8 per 1000 [4]. A decline was reported in 2018, when the density of skilled health workers was 1.74 per 1000 population (an equivalent of 17.4 per 10,000 population) [5]. In addition to the acute shortage of skilled health workers [6], the health workforce is also facing neglect in health systems development [7]. In Kenya, the "brain drain" [8], migration [9], poor working conditions [10], poor human resources for health management [11,12], low salary, and delayed payment (or non-payment) of salaries [13], are some of the many challenges which the health workforce encounter, which leads to recurrent health worker strikes [14,15]. In 2017, Kenya experienced two countrywide 100-day doctors strikes, and 150-day nurses strikes [13], which also contributed to low health worker retention. Such strikes adversely impact the quality of healthcare [13] and threaten national and devolved health systems development [16].

In 2016, the WHO report titled 'Global Strategy on human resources for health: Workforce 20300 was published [17], following the adoption of related resolution WHA67.24 in 2014 [18]. One of the principles of this global strategy is to "Uphold the personal, employment and professional rights of all health workers, including a safe and decent working environment . . . " (p. 8) [17]. According to the Constitution of Kenya Article 43 (1) (a): "Every person has a right to the highest attainable standard of health, which includes the right to healthcare services, including reproductive health care." (p. 31) [19]. This signifies that health is a vital component for every individual (including a health worker), as recognized at the organizational, national, and global levels. Researchers and specialists across various disciplines have a unifying belief that health is vital in healthcare service delivery [20], health systems strengthening [21], and overall human development.

The Constitution of the World Health Organization defines health as, "the complete state of physical, mental and social well-being, not merely the absence of disease or infirmity" (p. 1) [22]. In this study, we assess the health-related quality of life (HRQOL). Currently, there is no universal definition of HRQOL [23]. In this study, HRQOL is defined as a multifaceted concept that delves into the assessment of ones' self-perceived health status using a multidimensional classification system [24]. Therefore, HRQOL is a value attributed to life, specifically focusing on health-related functional ability (or inability) and perceptions at an individualistic contextual realm [25]. HRQOL has been assessed among healthcare workers in various countries such as China [26], Pakistan [27], South Africa [28], and Greece [29], among others.

In Kenya, studies on HRQOL have been conducted among various populations such as children in Schistosoma haematobium-endemic areas [30], people living with irrigation schemes [31], women living in informal settlements [32], patients undergoing antiretroviral treatment [33–35], patients on maintenance hemodialysis [36] and patients who have undergone cataract surgery [37]. However, a study on HRQOL among healthcare workers in Kenya is yet to be conducted. Therefore, the present study contributes to bridging the existing knowledge gap in the public and private-not-for-profit (mission) hospitals. This study aims to investigate the predictors of HRQOL among healthcare workers in public and mission hospitals, Meru County, Kenya. The three research questions that this study aims to answer are:


This study will contribute to the existing literature that policymakers can use to inform the development of an evidence-based health workforce policy. It will also contribute to raising the awareness among policymakers and health development partners on the pivotal role of healthcare workers' HRQOL in health workforce strengthening, development of resilient health systems, and achievement of the SDG3 target 3.8 on Universal Health Coverage (and indeed attainment of all the remaining 12 targets).

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

#### *2.1. Study Design*

A cross-sectional study design was used to investigate HRQOL among healthcare workers in public and mission hospitals in the Meru County of Kenya.

The study was conducted between 15 June and 30 July 2020, which was during the Coronavirus Disease 2019 (COVID-19) pandemic. However, at that time, the COVID-19 cases were relatively few in Meru County. By the end of July, a total of 32 cases had been recorded [38]. Therefore, the COVID-19 pandemic did not adversely impact the data collection process. The healthcare workers were highly cooperative during the data collection phase of this study. However, despite there being low numbers of cases in Meru County at the time of data collection, it is important to note that globally, the COVID-19 pandemic shocked health systems and resulted in healthcare workers experiencing psychological distress and psychosomatic symptoms [39]. For example, in China and Singapore, a narrative review revealed that the COVID-19 global pandemic resulted in healthcare workers experiencing enormous stress especially during spikes of cases which were experienced at different periods across different countries worldwide [39].

#### *2.2. Study Setting*

Meru County is one of the forty-seven counties in Kenya. The total population by July 2020, was approximately 1,545,714 people [38]. It is primarily a rural area located on the eastern slopes of Mount Kenya and is known for livestock rearing and agriculture, specifically, cash crop and food crop farming [40]. By 2019, there were 183 health facilities across the entire healthcare referral system in Meru County. This study focused on the sub-county and county level, public and mission hospitals (*n* = 24). In Kenya, the majority of hospitals are categorized according to hospital ownership. In the rural areas, hospital ownership is primarily public and mission. This means that public hospitals are owned and operated by the government. In comparison, mission hospitals are owned and operated by private not-for-profit religious organizations.

#### *2.3. Study Population and Sample*

In Meru County, the total number of human resources for health (HRH) in public and mission sub-county and county hospitals was 1872 by 2019. The present study focused on healthcare workers, also known as healthcare professionals. Healthcare workers, in this study, are individuals who have been trained in the medical field to apply evidencebased medical procedures and principles, geared towards achieving quality healthcare delivery [41]. Our focus was on medically trained healthcare workers, excluding auxiliary staff. Thus, the total number of healthcare workers eligible to participate was 954. The sample size (ss) was calculated using the following formula [42]:

$$\text{ss } = \left(\frac{\mathbb{Z}^{2\*}\left(\mathbf{p}\right)\*(1-\mathbf{p})}{\mathbb{C}^2}\right) \tag{1}$$

Based on the formula, the total sample size of 566 was determined by the following parameters: the population was 954, Z = Z value at 99% confidence level, C = confidence interval of 3.46, and p = response distribution percentage of 50%.

Using simple random sampling a total of *n* = 566 healthcare workers were selected across the all the public and mission (*n* = 24) hospitals to participate in this study. The health professional cadres presented in this study were doctors, clinical officers, nursing personnel, dentistry personnel, pharmaceutical personnel, medical laboratory scientists, nutritionists, public health specialists, mental health specialists, physiotherapists, radiologists, and health records officers.

#### *2.4. Data Collection*

#### 2.4.1. Sample Characteristics

Sample characteristics were collected in the socio-demographic section of the instrument. The personal and job-related attributes, and work environment characteristics, constituted the independent variables in this study. The personal and job-related attribute data obtained were in the following categories: hospital ownership, health professional cadre, age, marital status, gender, household size, education attained, years of professional experience, hours worked in a week, in-service training, staff housing, and type of employment. Work environment characteristics data on the healthcare workers' perception of their working environment related to hygiene, water, sanitation, and occupational hazards in the hospital, were also obtained.

#### 2.4.2. The EQ-5D-5L

The EuroQol-five dimension-five level instrument (EQ-5D-5L) developed by the EuroQol Research Foundation [43] was used to measure HRQOL among healthcare workers. By 2019, the EQ-5D-5L had been translated into more than 180 languages and applied globally [43]. The EQ-5D-5L assessed the respondents' self-perceived health across five dimensions, namely: mobility, self-care, usual activities, pain/discomfort and, anxiety/depression [43]. Each respondent indicated what they felt across the five dimensions, depending on the boxes ticked, and a five-digit number, e.g., 11,111 (denoting full health), was generated for analytical purposes as per the EuroQol User Guide [43]. The last question in the instrument is the EQ-VAS (EuroQol Visual Analogue Scale) which required the respondents to assess their overall health status on a scale 0–100, where 0 signifies the worst health one can imagine, and 100 signifies the best health the respondent can envisage [43]. The EQ-5D-5L instrument was used after obtaining permission to use the Kenyan version for this study as instructed by the EuroQol Research Foundation. The research team, consisting of the principle investigator and two research assistants, explained the study both in written form (informed consent form) and verbally. After signing the informed consent form, respondents were given the self-complete paper version of the questionnaire. The respondents were informed that they could ask the research team any questions regarding the study, and they completed the questionnaire anonymously. Upon completing the questionnaire, the respondents would return it to the research team. On average, each respondent completed the questionnaire within 10 min.

Pretesting of the data collection instrument was performed among healthcare workers to evaluate its contextual validity and lucidity. The section of the personal, job-related and work environment characteristics was modified to enhance the contextual applicability in our setting.

#### 2.4.3. Statistical Analysis

Data entry was performed in Excel (Microsoft, Washington, DC, USA) and exported to STATA 15.1® (StataCorp., College Station, TX, USA). Analysis of the EQ-5D-5L selfcomplete paper version was conducted according to the EuroQol User Guide [43]. From the respondents scores across the five dimensions, the EQ-5D-5L health profiles were obtained. From this, the EQ-5D index values were calculated using the EQ-5D-5L Crosswalk Index Value calculator for Windows [43]. After obtaining the index values, measures of central tendency (including the median and interquartile range) were estimated using STATA 15.1® [43]. Using the EQ-VAS as the dependent variable (i.e., the self-reported overall health status score), analysis of variance (ANOVA) and linear multivariate regression analysis was performed using STATA 15.1®. The linear multivariate regression model estimated was [44]:

$$Y\_k = \beta\_0 + \beta\_1 X\_{1k} + \beta\_2 X\_{2k} + \dots + \beta\_{25} X\_{25k} + \in\_k \tag{2}$$

where *β*<sup>0</sup> indicates the constant or intercept term capturing the unexplained variations in the dependent variable Y (i.e., EQ-VAS), *β*<sup>1</sup> indicates the slope coefficient measuring the amount by which *Y* will change when *X* changes by a single unit, *k* ranges from 1 to *n*, in this case the 25 independent variables, *X*1*<sup>k</sup>* = stands for the *k*th observation value for the independent variable *X*1, and ∈*<sup>k</sup>* is the error (disturbance) term that captures errors in model specification and other factors that influence healthcare workers' EQ-VAS (overall health status score) but are not explicitly considered in the model.

The predictors of the healthcare workers' overall health status were assessed using this model. A *t*-test was performed to determine whether each individual variable regression slope coefficient was statistically significant at 90% or 95% level of confidence.

#### *2.5. Ethical Considerations*

Following permission from the Meru County Government Department of health [CGM/COH/1/17(50)], permission was sought from all the hospitals that participated in this study. Subsequently, written informed consent was obtained from each respondent, before they anonymously and voluntarily completed the self-administered questionnaire. Before this, the research protocol underwent a sequential three-step approval process. In South Africa, the University of Pretoria, Faculty of Health Sciences Research Ethics Committee approved the protocol of this study [718/2019]. In Kenya, the United States International University Africa, Institutional Review Board, also granted Kenyan ethical approval [USIU-A/IRB/130-2020]. Subsequently, the National Commission for Science, Technology and Innovation, Kenya, granted a national research license number [901924] to perform this study in Kenya.

#### **3. Results**

The total number of respondents in this study was 553 healthcare workers out of 566. It yielded a response rate of 97.7% because thirteen questionnaires were excluded from data analysis, due to 50% or more questions not being answered. The response rate could be attributed to various factors, including the fact that the questionnaire was asking about the healthcare workers themselves, thus they were inclined to participate. As mentioned earlier, data collection was conducted during the country's early onset of the COVID-19 pandemic, during which the healthcare workers' workload was less because people generally avoided visiting hospitals, due to fear of contracting the contagious COVID-19 virus. No incentives were offered or given to respondents, they all voluntarily participated in this study.

#### *3.1. Sample Characteristics*

Table 1 presents the percentage frequency distributions of the personal and job-related characteristics of the healthcare workers, overall (*n* = 553), and by hospital ownership (sub-sample). From a total of 553 respondents, 74.48% worked in public hospitals and 21.52% in mission hospitals.

**Table 1.** Overall and sub-sample percentage frequency distributions of personal and job-related characteristics.



**Table 1.** *Cont.*


**Table 1.** *Cont.*

Note: The exchange rate as of 22 December 2020, was, USD 1 = KES 110.38.

#### *3.2. Work Environment Characteristics*

Table 2 presents the frequency distributions and percentages of the work environment characteristics measured among the healthcare workers, in three categories overall (*n* = 553), public (*n* = 434) and, mission (*n* = 119) hospital ownerships.

**Table 2.** Overall and sub-sample percentage frequency distribution of work environment characteristics (*n* = 553).



#### **Table 2.** *Cont.*

#### *3.3. EQ-5D-5L Health Profile, Index Value and EQ-VAS*

The EQ-5D-5L health profile showed that approximately 66.55% of all the respondents reported no problems across all the five dimensions. Nevertheless, 33.45% of the healthcare workers in this study reported problems within the dimensions assessed. In public hospitals (*n* = 434), about 64.75% of the healthcare workers had no problems across the dimensions,

but 35.25% experienced health problems across the dimensions. In mission hospitals (*n* = 119), 73.11% of the respondents did not experience any problems across the five dimensions, leaving 26.89% who confirmed experiencing health-related problems; thus implying that not all healthcare workers are at their best health state, with approximately more than 30% experiencing problems across all the dimensions assessed.

The median of the EQ-5D-5L index (IQR) value was 0.900 (0.595–0.900, on a scale of 0 to 1) overall and in both public and mission hospitals. This implies that the healthcare workers' health profiles were relatively high, with a median score of 0.900, which was 0.1 below 1, where 1 signifies full health. However, there is room for improvement, because the EQ-5D-5L index value scores fell short of full health by a value of 0.1.

The EQ-VAS presented the results of the healthcare workers' self-assessed overall health status, on a scale of 0–100 [43]. About 68.72% of the healthcare workers rated their overall health greater than 90 (where 100 indicates the best health you can imagine). The 553 respondents had a median of 90, first quartile = 80, third quartile = 100, minimum = 20 and maximum = 100, and four outliers = 20, 36, 40 and, 49. Figure 1 presents the box-andwhisker plots of EQ-VAS by hospital ownership. *Healthcare* **2021**, *9*, x 9 of 16 553 respondents had a median of 90, first quartile = 80, third quartile = 100, minimum = 20 and maximum = 100, and four outliers = 20, 36, 40 and, 49. Figure 1 presents the box-andwhisker plots of EQ-VAS by hospital ownership.

**Figure 1.** Box-and-whisker plot of EuroQol Visual Analogue Scale (EQ-VAS) by hospital owner-**Figure 1.** Box-and-whisker plot of EuroQol Visual Analogue Scale (EQ-VAS) by hospital ownership.

Among the public hospitals (*n* = 434), the median (IQR) was 90 (80–100). Approximately 25% of the healthcare workers' overall self-rated health status in public hospitals, was lower than 80. About 75% of the healthcare workers' overall self-rated health status was rated more than 80, with four outliers. The healthcare workers in the mission hospitals (*n* = 119), the median of their overall self-rated health status was 95 (90–100). About 25% of their overall self-rated health status was rated lower than 90. Approximately 75% Among the public hospitals (*n* = 434), the median (IQR) was 90 (80–100). Approximately 25% of the healthcare workers' overall self-rated health status in public hospitals, was lower than 80. About 75% of the healthcare workers' overall self-rated health status was rated more than 80, with four outliers. The healthcare workers in the mission hospitals (*n* = 119), the median of their overall self-rated health status was 95 (90–100). About 25% of their overall self-rated health status was rated lower than 90. Approximately 75% of their overall self-rated health status was rated more than 90, with two outliers.

#### of their overall self-rated health status was rated more than 90, with two outliers. *3.4. Overall Self-Reported Health Status by Hospital Ownership*

ship.

*3.4. Overall Self-Reported Health Status by Hospital Ownership*  The ANOVA results revealed a statistically significant difference between the public and mission hospital healthcare workers' overall self-reported health status (EQ-VAS) (*p* = 0.0057). Hospital ownership explained 1.38% of the variance in healthcare workers' over-The ANOVA results revealed a statistically significant difference between the public and mission hospital healthcare workers' overall self-reported health status (EQ-VAS) (*p* = 0.0057). Hospital ownership explained 1.38% of the variance in healthcare workers' overall health status (see Table 3).

EQ-VAS 553 0.0138 0.0120 7.69 0.0057 \*

sonal, job-related and, work environment characteristics (*p* < 0.01) (see Table 4).

The linear multivariate regression model showed that approximately 13.73% of the variance in the overall health status among the respondents was explained by the per-

**Table 3.** ANOVA of EQ-VAS and hospital ownership.

all health status (see Table 3).

\* *p* < 0.05 indicates statistical significance.

*3.5. Predictors of Overall Health Status* 


**Table 3.** ANOVA of EQ-VAS and hospital ownership.

\* *p* < 0.05 indicates statistical significance.

#### *3.5. Predictors of Overall Health Status*

The linear multivariate regression model showed that approximately 13.73% of the variance in the overall health status among the respondents was explained by the personal, job-related and, work environment characteristics (*p* < 0.01) (see Table 4).

**Table 4.** Linear multivariate regression model results.


Table 5 presents the results of the linear multivariate regression model, including the twenty-five independent variables assessed association with the EQ-VAS (overall selfrated health status). The model showed six statistically significant predictors of overall health status among the healthcare workers (*n* = 553). Hospital ownership (*p* = 0.029), age (*p* < 0.001) and income (*p* = 0.069) were the three significant personal and job-related predictors associated with the healthcare workers' self-assessed health status. Moreover, the availability of water for handwashing (*p* = 0.018), presence of risk when using the toilet facilities (*p* = 0.015), and the overall safety of the hospital work environment (*p* = 0.001) were the three work environment-related predictors of healthcare workers' overall health status (see Table 5). Twelve of the twenty-five independent variables had negative coefficients, which implies that as values of those independent variables increase, the healthcare workers' overall health status decreases. On the other hand, thirteen variables had positive coefficients, meaning as the independent variables increase the overall health status of the healthcare workers increase.


**Table 5.** Results of linear multivariate regression of overall health status, and independent variables (*n* = 553).


**Table 5.** *Cont.*

\* *p* < 0.01 indicates statistical significance at 95% confidence level; \*\* *p* < 0.01 indicates statistical significance at 90% confidence level.

#### **4. Discussion**

HRQOL is based on an individuals' perception of their ability to execute functions associated with their health; related to the physical, psychological and occupational dimensions of life [24,43]. In this section, we discuss the HRQOL among healthcare workers in this study, compared to prior studies.

Overall, more than 30% of the healthcare workers studied reported experiencing problems across the dimensions. A study in South Africa found that up to 45% of the healthcare workers under study experienced problems across the five dimensions [28], thus the issue of HRQOL is in force. Both the South African study [28] and this study dispel the misconception that healthcare workers are automatically always in perfect health, due to their medical background. Thus, there is room for more action-oriented research to be done on healthcare workers' HRQOL. Health managers should consider implementing programs on health promotion behavior, and self-efficacy, which have been reported to have a positive impact on HRQOL [45], and thus, could enhance the healthcare workers' HRQOL.

In this study, a statistically significant difference between overall health status among healthcare workers in public and mission hospitals was revealed. As the hospital ownership changed from public to mission, the overall health status of healthcare workers increased by 3.079%. Healthcare workers in mission hospitals reported experiencing higher overall health status (73.11%) compared to their counterparts in public hospitals (64.75%). A study in Brazil revealed that healthcare workers in public hospitals had the lowest HRQOL scores compared to the private and philanthropic hospitals [25]. It appears that there is a need for interventions to increase HRQOL, especially in the public health sector. Most of the respondents in this study were from the public health sector, therefore policymakers and hospital managers should consider developing and implementing policy based on these research findings. Details regarding the predictors and possible solutions are discussed below.

Age was found to be a significant predictor of healthcare workers' HRQOL in this study. As the age of the healthcare workers increased, their overall HRQOL decreased by 0.431%. However, a study in Brazil revealed the older health workers were, the better their HRQOL compared to their younger counterparts [29]. The differences between the results in Kenya and Brazil may be attributed to contextual or cultural differences which influence the perception of age. Based on these findings, age-friendly employment policies need to be developed and implemented within the hospitals and the health system at a large scale. Age-friendly employment policies, such as creating an ergonomic work environment supporting older healthcare workers, will enhance their health [46]. Guaranteed financial incentives, and relatively flexible work schedules that allow work–life balance are some age-friendly strategies that could promote older healthcare workers' health and encourage them to work in the health system longer [46].

Income was positively and significantly associated with HRQOL among healthcare workers in this study. The higher the income of the healthcare workers under study, the higher their overall HRQOL. At an individual level, an association has been reported between income and health, particularly in situations where there are scarce goods and services available to the public [47]. In this context, this may partly explain the recurrent health worker strikes due to delayed or missed payments [13], which elicit feelings of scarcity and uncertainty of payment of the income for which they have worked and on which they greatly depend. Kenya's age dependency ratio of 71.3% in 2019 indicates that children (0–14-year-olds) and the elderly (65 years and above) are dependent on those working for survival [48]. Therefore, delayed pay and missed pay among healthcare workers jeopardizes survival of health workers and their dependents, and aggravates the income inequalities which adversely affect population health [49] especially, in a lowermiddle income such as Kenya with approximately 33.4% of the population living below the international poverty (Int\$) line of Int\$1.9 per day [50]. Thus, national and county policymakers should develop and implement strategies that facilitate timely payment and provide equal opportunities for promotions and incentives. This kind of action could potentially increase the HRQOL and eventually the retention of healthcare workers, especially in rural and remote areas in Kenya.

Previous studies revealed contrary results regarding the personal and job-related characteristics among healthcare workers. For instance, sex was a significant factor among health professionals in Turkey, where males had a higher HRQOL compared to their female colleagues [51]. This study revealed that sex was inversely related to HRQOL but was a non-significant characteristic among the respondents. In this study, the health professional cadre was also non-significant. A Turkish study on the other hand, reported higher HRQOL scores among physicians and health technicians compared to nurses and midwives [51]. Similarly, in Italy, the professional role significantly impacted the HRQOL, where nurses reported lower HRQOL scores compared to doctors and occupational safety and health technologists [52]. Although the professional cadre was a non-significant predictor in our study, further research needs to be done country-wide, to ascertain if this is similar or different in other locations. In addition, more studies on HRQOL across health professional cadres will inform future directions of health development, specific to the professional cadre needs [51].

In this study, length of work experience was a non-significant characteristic of healthcare workers' HRQOL; this was similar to a Turkish study [51]. On the contrary, in Italy, the longer a healthcare workers' career, the lower their general health score [52]. The differences in results could be attributed to contextual factors such as culture, location, and the period of study. As much as length of work experience was a non-significant predictor of HRQOL, age was a significant predictor in this study. The healthier healthcare workers are, the longer they are likely to work [46]. Hence, diversification of hospital organizing services, policies, and strategies such as age-friendly benefit packages promoting their health for example assistive devices were necessary: for example, comprehensive health insurance covers that facilitate restorative surgery, and acquisition of nutritional supplements [53], are some ways that could promote their health and, enable them work longer in the health system. However, the authors recognize that more research needs to be done in multiple settings to inform evidence-based policy and strategies towards promoting healthcare workers' HRQOL for longer job retention.

A healthy and safe work environment is valued by health providers and is paramount to the health worker performance and retention [54]. Improved performance among healthcare workers has been attributed to safety and hygiene; this subsequently has increased client satisfaction [54]. In this study, as the overall safety of the hospital work environment increased, the overall HRQOL of the healthcare workers increased. The presence of risk when using the toilet facility decreased the overall HRQOL among the respondents. This finding implies that the higher the perceived risk of the hospital work environment, the lower the healthcare workers' perception of their HRQOL. The healthcare workers' availability of water for handwashing increased the overall health status by 4.478%. Weinberg and colleagues [55] reported that the high-performance work environment in hospitals significantly correlated with better performance, better retention, and better-quality healthcare among the healthcare providers. Thus, policymakers and hospital managers need to consider the benefits and importance of designing a high-performance work environment because of its potential benefits related to the quality of healthcare delivery and patient outcomes [55].

According to Herzberg's Two Factor Theory on job attitudes [56], the predictors of healthcare workers' HRQOL are income or salary and work environment. Following this theory, hygiene factors are also known as job dissatisfiers. Hygiene factors are extrinsic to the job [57]. In this study, low salary and poor work environment were major dissatisfiers. Thus, hospital managers and health authorities should be explicit in the implementation policies of salary increment, financial incentives and payment of healthcare workers [57]. In relation to the environment, hospital managers and health policymakers should eliminate the dissatisfaction contributing to a poor working environment. Based on the findings, this could be achieved through improving the hospital safety, hygiene and work environment, in order to make the work environment in hospitals satisfying for healthcare workers to have a better HRQOL and to perform optimally.

The healthier healthcare workers are, the better the relationship they will have with colleagues, and they will deliver better healthcare services to patients they encounter daily [52]. The results in this study should be viewed with some limitations in mind; hence, opportunities for future research.

Firstly, this was a cross-sectional study; therefore, only correlations could be reported. Future studies using a longitudinal approach to monitor and evaluate the HRQOL of healthcare workers are essential to capture the trends accurately and modify health policy accordingly. The second limitation is that the sample reflects the healthcare workers in one of the forty-seven counties in Kenya, thus limiting the generalizability of these results to the entire country. Future studies need to be done in the other 46 counties in order to assess the similarities and variations in HRQOL among the healthcare workers in the different localities countrywide. Thirdly, due to the self-reported nature of the questionnaire used, the possibility of response bias is present. To reduce the likelihood for such a bias, respondents were informed that the research was anonymous, and their honesty would be valued. In future research, the HRQOL could be assessed alongside government and mission programs aimed at improving the health and wellbeing of the health workforce.

#### **5. Conclusions**

This study highlighted personal, job-related, and work environment predictors of HRQOL among healthcare workers in public and mission hospitals Meru County, Kenya. It is evident that some personal, job-related, and work environment characteristics are significant predictors of HRQOL among healthcare workers. The majority of the respondents reported perfect health, thus through evidence-based policy development and implementation of HRQOL programs, other health workers with problems stand a chance of attaining a higher HRQOL. This study emphasizes the importance of involving the healthcare workers in the decision-making process of promoting their HRQOL, because some of our results differed with prior studies also among healthcare workers. It is evident that not every healthcare worker is in perfect health, as is the misconception based on their medical

background. This finding implies that health policymakers and managers should aim at empowering and enhancing the changeable HRQOL among healthcare workers at the individual, organizational, and health system levels. Designing evidence-based mediumand long-term policies and programs would ensure effective implementation, and health workforce strengthening. In order to ensure sustainability within the national and county health systems, an inter-sectoral collaboration between the public and private sectors is recommended during the development (and revision) of health workforce policy aimed at HRQOL and wellbeing among healthcare workers in Kenya.

**Author Contributions:** Conceptualization, R.N.D.K.M., C.H. and F.S.; methodology, R.N.D.K.M.; software, R.N.D.K.M.; validation, R.N.D.K.M., C.H. and F.S.; formal analysis, R.N.D.K.M.; investigation, R.N.D.K.M.; resources, R.N.D.K.M.; data curation, R.N.D.K.M.; writing—original draft preparation, R.N.D.K.M.; writing—review and editing, R.N.D.K.M., C.H. and F.S.; visualization, R.N.D.K.M., C.H. and F.S.; and supervision, C.H. and F.S. All authors have read and agreed to the published version of the manuscript.

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

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Faculty of Health Sciences Research Ethics Committee of the University of Pretoria, South Africa (protocol code 718/2019 on 30th January 2020). In Kenya, approval the study was approved by the Institutional Review Board of the United States International University-Africa, Kenya (protocol code USIU-A/IRB/130-2020 on 21 February 2020).

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

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available until this article is published.

**Acknowledgments:** We appreciate the Department of Health, Meru County, Kenya. We thank the hospital management teams who permitted the present study to done in their health facilities. We are grateful to all the participants of this study. We are grateful to the University of Pretoria management for supporting this study. We thank God, for providing the resources that enabled us to do this research.

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

#### **References**


**Adriana Burlea-Schiopoiu 1,\* and Koudoua Ferhati <sup>2</sup>**


**Abstract:** The aim of the paper is to identify a set of the key performance indicators (KPIs) in order to provide managers and employees from the healthcare system with recommendations for evaluating, monitoring, and controlling the critical factors that influence the performance of the healthcare sector in Algeria during a pandemic crisis. During February–August 2020, a cross-sectional survey design was administrated to medical employees from hospitals situated in the northeastern part of Algeria. Our findings proved that the four groups of KPIs correlate to each other, and during this period, the triple relationship among human factor-technology-medication plays a decisive role in reducing the pressure on the medical system and overcoming the crisis. In order to increase the efficiency of the decision-making process, a hierarchy of KPIs is recommended in terms of their impact on the performance of medical staff. The practical importance of our research consists in ranking KPIs on four clusters that support managers to focus on both the human factor (clinical errors, infection rate, and medication errors) and the technical elements of maximum importance (laboratory test time, location of the facility, and sufficient air).

**Keywords:** key performance indicators; healthcare system; pandemic crisis; COVID-19; Algeria

#### **1. Introduction**

The recent crisis that the entire world is facing caused by COVID-19, a pandemic that has forced all organizations, whether public or private, to rethink their mission and vision. Thus, the efficacy of the healthcare sector depends greatly on the rapidity to adapt to the new dramatic situation. Before the World Health Organisation (WHO) had declared the COVID-19, a pandemic crisis, on 11 March 2020, Algeria had the situation under control, although the first two cases were registered on 25 February 2020. The rapid evolution of the pandemic coronavirus crisis requires that the common strategies should be oriented toward ensuring the health of the population and a continuous assessment of the events to give priority to future needs. The situation in Algeria is not far from the rest of the world, with a total of 55,081 cases and 1880 deaths recorded until 22 October 2020, with a mortality rate of 11.70% [1]. This high rate represents a threat to the national health situation of the country that leads to making a study to understand how the healthcare sector is dealing with this crisis and what are the available and efficient managerial tools to help managers and healthcare staff to better control facilities [2].

The international statistics of WHO [3] place Algeria in the fourth position in Africa (with 50,914 cases) after South Africa (with 669,498 cases), Ethiopia (with 72,700 cases), and Uganda (with 7364 cases) with a total of 1,172,342 confirmed cases and 25,481 deaths in the continent as the last update.

Watkins et al. [4] analysed the impact of the pandemic situation on SMEs from Australia and they observed that only six per cent of Australian SMEs had a plan for avoiding the pandemic crisis. Thus, the other 39 per cent consider that the pandemic has no impact

**Citation:** Burlea-Schiopoiu, A.; Ferhati, K. The Managerial Implications of the Key Performance Indicators in Healthcare Sector: A Cluster Analysis. *Healthcare* **2021**, *9*, 19. https://doi.org/10.3390/ healthcare9010019

Received: 30 November 2020 Accepted: 23 December 2020 Published: 25 December 2020

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

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

on their business and over 60 percent called upon the need to prepare a viable strategy to tackle a pandemic situation. As a result, in 2007, the Australian SMEs were not prepared for a pandemic situation and 13 years later SMEs worldwide are in the same situation.

Watkins et al. [5] considered that a strategy for the pandemic situation among SMEs depends on the perception of the risk level and the resources that are available to prepare and frame the strategy.

Ivo Hristov and Antonio Chirico [6] identified the impact of key performance indicators (KPIs) on company performance in the framework of sustainable strategies and they arrived at the conclusion that existing literature does not provide ample evidence about how to address the crises from managerial perspective.

At this moment, healthcare managers, officials, and policymakers are seeking to answer the following questions:

How will we manage to streamline the healthcare sector in real-time to overcome this pandemic generated by the novel coronavirus?

Can KPIs be used as a tool enabling a quick reaction of the healthcare sector's facilities to the adverse effects of COVID-19?

What is the possible set of KPIs that leaders in the healthcare sector are required to check in order to deal with the consequences of the virus and re-establish "normality" in the social and business environment?

The answers to these questions were found in the last months from many scholars around the world from economic point of view, but not from the healthcare sector perspective [7–10]

These analyses motivated us to approach this pandemic crisis from another perspective based on KPIs as instruments of human competencies in the fight against this enemy.

The range of factors affecting sustainability performance is varied, from the small details to the big issues in healthcare facilities. Thus, most of them are not obvious to the decision makers (e.g., maintenance and building design) because of lack of technical skills (e.g., attention towards the aspects of bioclimatic passive strategies and incorrect architectural and flexible layouts, which often limit the clinical processes and causes the decrease of performance and sustainability criteria) [11].

The use of KPIs as tools to improve the effect of public health measures and as indicators of the measures that need to change in function of true morbidity and mortality rates.

#### **2. Literature Review and Premises of Research**

Mohamed Khalifa and Parwaiz Khalid [12] classified KPIs on three levels of performance (operational, tactical, and strategic indicators), six levels of performance dimensions (safety, effectiveness, efficiency, timeliness, patient-centeredness, and equity) and three levels of system components (structure, processes, and outcomes). They have identified fifty-eight KPIs, classified into ten categories.

According to the literature review and based on our preliminary research, considering the particularities of the pandemic COVID-19, 41 indicators were selected for the KPI of the healthcare system in Algeria. The structure of the KPI is the following: ten indicators for social sustainability, eight indicators for economic sustainability, nine indicators for the internal process and fourteen indicators for the technical domain.

#### *2.1. Social Sustainability Indicators*

Social sustainability indicators (SSI) for healthcare facilities facing a crisis can be ambiguous to define and apply. Vallance et al. [13] (p. 342) affirm that SSI is "a concept in chaos". SSI have been organized under the broad categorical concerns of well-being, values, agency, and inequality [14].

Some researchers consider that SSI can be used to measure the performance in a state of overall well-being [15], and that KPIs should mainly contribute to improvement of the people's life [16].

In consequence, we need to be very specific while setting the indicators for social sustainability that helps track the critical metrics within a facility during a global pandemic crisis. As a result, there have been many attempts to categorize the various approaches to social sustainability indicators (Appendix A).

#### *2.2. Economic Sustainability Indicators*

The assessment of sustainability performance is assumed to be appropriate to the healthcare industry. Many studies were carried out with the scope of identifying an initial set of potential KPIs from an economic perspective [17] that could be used for sustainability performance evaluation to keep facilities operating with the minimum regular conditions of sustainability while facing the global pandemic (Appendix B).

A detailed explanation of ESI from a company point of view (Novo Nordisk A/S) was realized by Morsing et al. [18] and they arrived at the conclusion that KPIs are related to social and economic objectives, but are reflected on internal environment and also external environment.

#### *2.3. Indicators of Internal Process*

The indicators of internal process (IIP) in healthcare facilities are a well-defined performance measure used to monitor, analyse, and optimize all relevant processes and practices of the facility's staff to increase patient satisfaction and diminish any possible errors or damages (Appendix C).

#### *2.4. Technical Indicators*

Because of the complex technical and architectural nature of healthcare facilities, it requires a special set of indicators that suit the specific strategic objective's action plan. When facing a global pandemic such as COVID-19 every detail becomes important and overlooking, even the thinnest element such as insulation or air quality may lead to losing control over the crisis (Appendix D).

The KPI's constitute a management control tool that is used for planning and prioritizing actions, for making decisions and responding to problems in real time [19,20]. Elmar Hörner [21] arrived at the conclusion that one of the most important KPIs for measuring the success of the areas that need improvements in the pharmaceutical domain (e.g., Merck KGaA) is "Decision Making". The continuous monitoring of KPIs makes possible to answer the question of whether the objectives are achievable and, if yes, to what extent, which constitutes the basis for evaluation of the performance of contemporary facilities.

As the main role of a KPIs' system is tracking the performance and sustainability in the healthcare facility for the interest's spot; thus, it is fundamental to make an analysis of the facility's needs, strategy and goals. Therefore, in the case where the urgent need of crisis management is the main strategy, KPIs should serve this objective.

According to Schmidt et al. [22] (p. 760): "In order to successfully create new KPIs, it is crucial to analyse and understand the underlying cause and effect relationships as well as interdependencies between processes, equipment, and energy efficiency".

Recently, Daria Mikhailova [23] has chosen the KPIs to assess the performance of Pharmaceutical Project Management Quality System Effectiveness and she arrived at the conclusion that KPIs proved a practical support for both system monitoring and system interventions. Moreover, David Parmenter [20] underlines the importance of KPIs selection in the healthcare system and he advised us never underestimate the negative consequences of the choice of an inadequate KPIs.

Fernandes et al. [24] stated that all panellists finally accepted only eight KPIs. We allocated the eight KPIs of Fernandes et al. to our KPIs classes (e.g., SSI: providing in-person disease and medication education to patients; participating in interprofessional patient care rounds; providing discharge patient medication education; ESI: performing discharge medication reconciliation; and IIE: performing admission medication reconciliation (including best-possible medication history); completing pharmaceutical care plans; resolving drug therapy problems; and providing bundled, proactive direct patient care activities).

Matsuoka and Hirai [25] used the KPIs for explaining the core principles of Society 5.0 based on three factors (structural transformation, technological innovation, and quality of life). In addition, Re Cecconi et al. [26] used the KPIs to underline the importance of the technical indicators as strategic tools for improving the decision-making process. Jiang et al. [27] starting from the premise that performance measurement is of vital importance for the healthcare systems, especially during crisis periods, they proposed a model based on KPIs to help managers to make good decisions under time pressure. In the same line, Niemi et al. [28] consider the mean lead-time (MLT) as one of the most important KPI.

Ramzi Shawahna [29], (p. 2) affirms that KPIs "are often developed for capturing the performance of healthcare providers and the provision of services. These KPIs are supposed to monitor if healthcare services were provided with consistency and efficiency". Based on both theoretical and practical arguments, we propose the following hypotheses:

**Hypothesis 1.** *The economic sustainability indicators directly influence the indicators of internal process*.

**Hypothesis 2.** *The social sustainability indicators directly influence the economic sustainability indicators.*

**Hypothesis 3.** *The social sustainability indicators directly influence the indicators of internal process*.

**Hypothesis 4.** *The technical indicators directly influence the economic sustainability indicators*.

**Hypothesis 5.** *The technical indicators directly influence the indicators of internal process*.

**Hypothesis 6.** *The technical indicators directly influence the social sustainability indicators.*

#### **3. Methodology**

To select the KPIs, we have conducted in-depth interviews with four health management experts and one assistant professor from the Faculty of Medicine of Constantine 3 Salah Boubnider University to discuss and identify a list from the extracted indicators of literature review (185). The first set of indicators was developed from the literature review: 158 indicators were extracted from the discussions with the experts and were selected, then 62 indicators were divided into four main domains for sustainability assessment: social, economic, technical, and internal processes. The selected indicators were used to design the questionnaire for the first round of the Delphi method.

In the second stage, in consensus with Galanis [30], we employed the Delphi method in two rounds. In the first round, 20 public hospitals and managers for each hospital have been chosen to participate in the research, as they were involved in sustainability. The questionnaires were distributed to managers to investigate the implication of KPI's for primary healthcare facilities' sustainability performance.

In the second round, we distributed the questionnaires to the same hospitals' managers, and, in this stage, we eliminated the indicators with the lowest mean scores. After we received the feedback from the managers, we reduced the number of KPIs (we eliminated 22 indicators) and 41 KPIs were included in the final setup of our research.

This quantitative study has been conducted from the first week of recording the first three cases of COVID-19 disease in Algeria until the end of August 2020; investigation was made in two main wilayas: El Taref and Constantine, including different type of facilities: university hospital, public hospital centers, and neighbourhood's healthcare facilities.

During February–August 2020, a total of 300 questionnaires were distributed to Algerian hospitals that were identified as very important hospitals in the framework of the pandemic coronavirus. In total, 210 completed responses (response rate of 70 percent) were collected from medical staff who are directly involved in the battle with this invisible enemy. Respondents were required to evaluate the importance of every KPIs using a five-point Likert scale (1—not important at all to 5—very important).

We chose to use the SmartPLS [31] method to analyse the data and we started the research by assessing the measurement model to ensure that each construct's KPIs are reliable and valid.

#### *3.1. Sample*

The 88 respondents (41.9%) were from facilities situated in El Taref and 122 respondents (58.1%) from facilities situated in Constantine. The gender of the respondents was balanced with 96 females (45.7%) and 114 males (54.3%). The age of the respondents was distributed as follows: 21.4% between ages of 25 and 35 years, 37.6% between ages of 36 and 45 years, 24.3 between ages of 46 and 55 years, and 16.7% between ages of 56 years and over (M = 2.36; SD = 0.999).

A total of 109 respondents participated in the pilot study and at the final of this study all 41 KPIs were retained because they registered a loading factor above the threshold of 0.70 (Appendix E), in consensus with Sarstedt et al. [32].

#### *3.2. Measures*

We started the research by assessing the measurement model to ensure that each construct's KPIs are reliable and valid.

Internal consistency of the research model was assessed by partial least squares structural equation modelling (PLS-SEM), we started by examining the indicator loadings [32], ranging from 0.701 to 0.866 and indicating that the KPIs have a very good degree of reliability (Appendix E).

Next, we calculated the "reliability indicators" and higher values indicate increased levels of reliability. The main indicators exceed the minimum threshold of 0.7 [33] as follows: Cronbach's Alpha that measures internal consistency reliability ranged from 0.903 to 0.932 and represent good to very good reliability levels of KPIs, Dijkstra-Henseler's rho\_A ranged from 0.907 to 0.934 [34] and composite reliability (CR) ranged from 0.920 to 0.940. Thus, all the values exceed the minimum threshold value of 0.7 for all variables indicating that the measurement model has good reliability (Table 1).


**Table 1.** Descriptive Statistics, and Reliability and Validity of Measurement Model.

Source: Authors' own contribution based on SmartPLS.

The descriptive statistics indicate the values of the mean and we observe that a mean value of 4.01 out of 5 suggests that most of the respondents mainly agreed that IIP is very important KPIs for the healthcare system in a pandemic situation. Meanwhile, the SSI registered the lowest value (3.86) and an explanation consists of the particularity of this period and on pressure existing on medical personnel.

The pressure is reflected on value of loading factor of the IIP-C6 (Medical errors = 0.860), followed by IIP-C5 (Clinical errors = 0.847); IIP-C1 (Medication errors −0.822); IIP-C3 (Mortality rate = 0.789); IIP-C4 (Infection rate = 0.770) and finally IIP-C9 (Laboratory test time = 0.720).

The standard deviation shows that there are no relevant differences among the KPIs as the values are close to one another for ESI, IIP, and TI (ranged from 0.648 to 0.673); only SSI registered a value above 0.770.

We evaluated the extent to which any selected construct differs from the others and we tested the "convergent validity" and we used the average variance extracted values that are greater than 0.5 (from 0.530 to 0.622) and validate the latent variables for our model composition, in consensus with Hair et al. [35].

Average variance extracted (AVE) analysis was conducted for evaluating if we have a good convergent discriminant validity and if each construct exceeds the threshold value of 0.50. The result proves that all KPIs are retained.

The results prove that the indicators of internal process have the highest value of Cronbach alpha (0.918) which highlights the importance of this group of indicators in the context of the current pandemic coronavirus crisis.

We continued to assess the "discriminant validity" by calculating the Fornell and Larcker [36] criterion and for proving the relevance of the structural model. The highest correlation was registered between ESI-ESI (0.788) and the lowest correlation was registered between ESI-SSI (0.423).

In order to consolidate the assessment of the discriminant validity in variance-based structural equation modelling we used Heterotrait–Monotrait ratio (HTMT), which is considered superior to previous indicators [37] as Fornell–Larcker criterion and (partial) cross-loadings (Table 2).


**Table 2.** Heterotrait–Monotrait Ratio (HTMT) Test.

Source: Authors' own contribution based on SmartPLS.

The results show that the values of HTMT were smaller than 0.90 (ranged from 0.420 to 0.848), which means that this ratio meets the requirements of the Henseler et al. [38]. In order to sum, the model assessments prove a good evidence of validity and reliability.

#### **4. Results and Discussion**

The correlation between KPIs was used for verifying the relationship between all variables (Table 3).

The correlations between variables reveal to us that the age of respondents negatively influences three KPIs (IIP, ESI, and TI) and prove that under pressure the experience of the medical personnel is important for the decision-making process. The age of respondents is positively correlated with patient satisfaction, because no matter their age, the healthcare system employees are devoted to their job and to their patients.

The gender of respondents and the location directly and negatively influence two KPI's (ESI and TI) and directly and positively influence the other two KPIs (SSI and IIP). These correlations prove that the practical KPIS as ESI and TI are perceived as having a negative influence on decision-making process under pressure.


**Table 3.** Correlations between KPIs.

\*\*. Correlation is significant at the 0.01 level (2-tailed); Source: Authors' own contribution based on SmartPLS.

The location is an important variable, because the patients with COVID-19 are in some regions and they are treated in hospitals specially designated for this disease. Algeria has a public healthcare sector and it is accessible and free of charge for all citizens, financed by the government, given Algeria's young population. In close alignment with this long-term strategy, the government maintains an intensive immunization program.

The correlations between IIP and the other three KPIs are positive and prove that internal process is developed for the purpose of patient satisfaction (0.783) and considering the consequences of the infection with COVID-19, IIP is related to facilities of the hospital (0.685) in terms of qualitative and especially, quantitative KPIs.

We analyse the results and we first test the collinearity of the research model and we observe that the variance inflation factor VIF values ranged from 1.897 (Hospital readmission rate) to 3.852 (Indoor air quality) and is within the limits recommended by Hair et al. [35].

The results prove that there is no collinearity problem interfering with our KPIs and we continued to evaluate the research model by interpreting the coefficient of determination (R2), *f* 2 , and *P*. The coefficient of determination between 0.25 and 0.50 is considered good and above 0.50 are considered very well. Figure 1 shows values of R<sup>2</sup> , ranged from 0.254 to 0.673. In conclusion, the predictive power of the model and R<sup>2</sup> .

We arrived at the conclusion that all the KPIs are valid and reliable and we assess the research model and test the hypotheses (Table 4).

**Figure 1.** The Path Coefficients of the research model for KPIs. Source: Authors' contribution based on SmartPLS.


**Table 4.** Path Coefficients.

Source: Authors' own contribution based on SmartPLS.

The resulting effect size value of each KPI in the model ranges from 0.001 to 0.855, which are included in the category of very small to large [39]. The value of goodness of fit that is generated through the standardized root mean squared residual (SRMR) is equal to 0.08, which means that our model fits the empirical data [40]. We also tested our hypotheses with the coefficient parameter and the significant value generated from the 95% bias-corrected confidence intervals of each KPI.

The path coefficients provide significant value (at the *p* 0.05 level), only the relationship SSI -> ESI is not supported. Thus, the value of the coefficient (T) to the relationship SSI -> ESI is 0.491 with a *p*-value < 0.623. In conclusion, all hypotheses are supported except for the second hypothesis.

The particularities of our study due to the pandemic crisis do not allow us to affirm that our results support previous studies, because not many studies related to KPIs were developed during the pandemic crisis.

We can link these results to the pandemic crisis when the communication process and human factors are more important than economic and material factors. Hospital performance is a reference to key performance indicators (KPIs), especially to IIP and TI as promoters of ESI and SSI, because in this period quantitative assessments of hospitals became an indicator of the capacity of them to achieve the new goals by making efficient use of the limited resources available in the crisis period. This signifies that TI can be used to improve SSI and the ESI have a positive effect on IIP.

Healthcare facilities all over the world are dealing with major challenges to keep operating in a performing way during crisis time especially when it's facing a world pandemic such as Covid-19. This pandemic coronavirus took by surprise both the decisionmakers and the employees who faced an unpredictable enemy and impossible to be controlled. The fact that many organizations, public or private, were forced to discontinue their activity for an unspecified period of time has created a state of panic and uncertainty, upsetting society at all levels.

Our findings proved that KPIs play an important role in increasing the performance of healthcare systems, and, especially during the pandemic coronavirus crisis (Appendix F).

Figure 2 shows the four clusters of KPIs that can be a priority for hospital managers in a crisis period and every cluster includes different KPIs ranked by importance.

**Figure 2.** The Clusters of the key performance indicators (KPIs) as a managerial decision support during pandemic crisis. Source: Authors' own contribution

The first cluster includes 10 KPIs, two KPIs of SSI (related to the average hospital stay and on patient waiting time), five KPIs of IIP (related to clinical and medication errors, infection rate, average length of stay in the emergency room, and laboratory test time), and the other three KPIs are from TI as indoor air quality, sufficient air conditioning, and location of the facility.

This cluster proves the importance of this indicator in a crisis because under pressure and in very stressful conditions the decisions made by doctors concerning medical prescriptions or dozes might be mistaken [41]. Clinical errors tracking and assessment appears to be significant in crisis time and can affect the medical staff and equipment's effectiveness. The sufficient air conditioning allows checking the sufficiency of air out of the facility's HVAC system to ensure the well-being of staff and patients inside the healthcare facility. This indicator is crucial for the patients because one of the causes of death is insufficiency respiratory. Waiting time in the emergency room shows the value of assessing this metric in order to set less waiting time targets in crisis time.

These findings are in consensus with the findings of Mohamed Khalifa and Parwaiz Khalid [12] who considered that patient safety and infection control rates are very important indicators to gauge the quality of the healthcare system.

The second cluster comprises the other 10 KPIs as follow: two KPIs of SSI that measure the patient satisfaction and patient transfer rate to other facilities, three KPIs of ESI that indicate average care costs of insured patients, cost of work-staff and current cost per bed; two KPIs of IIP related to mortality rate and to medical errors, and three KPIs of TI concerning natural light penetration, vertical circulation and degree of thermal comfort. Patient satisfaction measures the degrees to which the medical service responds to patients' expectations is a high priority for the management strategy in a pandemic crisis. Mortality rate measures the rate of deaths while in a world pandemic and it is considered a structural tool for decision-making and setting the facility's strategy to deal with the crisis according to the given number. The degree of thermal comfort could affect or be affected by other technical and internal processes indicators related to the wellbeing inside the facility.

The third cluster includes 11 KPIs structured by categories as following: two KPIs of SSI (related to hospital readmission rate and to waiting time in the emergency room); three KPIs of ESI including average hospital expenses, cost of drugs, and equipment and also average care costs; one KPI of IIP (waiting time for admission to the operating room), and five KPIs of TI (waste management, energy and emissions control, quality of the building envelope, artificial lighting, and water consumption).

The cost of drugs and equipment needs to be taken into consideration because of the increasing of needed medicines, supplies, and special equipment as in a crisis, and to be correlated to the patient waiting time in a crisis that is also important because every patient needs medical care as soon as possible when he/she arrives at the facility to avoid any complications could expose his/her lives to danger. Quality of the building envelope plays a key role in the healthcare environment inside the facility as it controls directly other technical indicators ranked.

The fourth cluster includes 10 KPIs as follows: four KPIs of SSI (patient safety, rate of vacant patients in beds, patient complaint rate, and the number of new patients); two KPIs of ESI concerning average maintenance costs and costs per payer; one KPI of IIP related to bed occupancy rate and three KPIs of TI (acoustic insulation, distributions of medical devices, and hierarchy of functional spaces). The rate of vacant patients in beds allows the facility managers to set their priorities and to make decisions like patients transfer to other departments or facilities. The number of new patients indicates the particularity of this pandemic crisis because it is limited to receive more patients in the facility, because of the virus, which needs to be followed to make sure that the facility is ready for offering care to an expected number of patients. The hierarchy of functional spaces underlines every daily movement of the staff between rooms, controlling, services, and departments. As a result, such a metric gives more flexibility and performance in hard times where every second counts inside the healthcare facility.

The respondents considered as very important two indicators: infection rate and patient safety, because controlling infection rates and applying protocols in healthcare facilities is considered a key practice when facing pandemics. In this situation, patient complaint rate, like the patient's satisfaction, is not a high priority for healthcare facilities in a pandemic situation.

Waiting time for admission to the operating room gives the facility's managers a clear vision about the target time to set surgical operations to increase the internal processes performance, going to cost of work-staff that is considered as a lower priority in these circumstances.

Laboratory test time was not considered as a high priority indicator, and this might be justified by the healthcare facilities protocols in case of a pandemic crisis as test samples and results are oriented to the big test labs in the country such as Pasteur Institute.

Our results confirm the findings of McCance et al. [42] that analysed the eight KPIs clustered within the person-centered processes domain of the framework, and these KPIs were related to patient satisfaction, confidence, and implication in the decision-making process about his/her care.

To prepare action plans for in time strategies and to implement them successfully, the healthcare facilities continue to look for appropriate strategy implementation tools. Consequently, the measurement systems are used to evaluate the effects of the healthcare facilities actions. The role of such system is to support the management process as well as the process of implementation of the hospital's strategy, which should include not only solid technical and economic factors (particularly financial measures), but also the requirements of the corporate social responsibility and sustainable development standards, as well as the employee relations and value management requirements [43].

One solution is to use KPIs and keep controlling performance and sustainability through dashboards and scorecards to make frequent and continuous evaluation for the outputs to ensure a better crisis management for healthcare facilities and to make the strategic objectives to be achieved clearer and more convenient. In the case of Algeria, we did not find previous studies about KPIs for evaluation in healthcare facilities from a managerial perspective. For this reason, we consider this study important and it can add a real value to the research, because it helps the managers to evaluate sustainability performance of healthcare facilities in crises.

#### **5. Conclusions**

Our findings prove the opportunity for healthcare system employees and not only for hospital managers, to identify critical KPIs in a short period of time and with lower costs. It is very important to consider that the IIP are situated in the first place, which gives us an idea about the priorities of healthcare staff in a crisis. We recommend focalizing the improvements in the areas with high potential to propagate the factors of the pandemic crisis.

In the framework of a pandemic crisis, the performance of healthcare systems is related to its capacity to quickly respond to danger generated by COVID-19. In this period, it is difficult to adapt to an existing model, because the variables are completely different, and the weaknesses of hospitals are also different.

The practical implications are underlined by our model that provides hospitals' manager's solutions for the decision making process under pressure indicating the ways of improvements of quality of medical services by implementing suitable KPIs. Thus, these clusters of KPIs can be used as tools for developing sustainable healthcare systems not only in Algeria but also in developing countries that need financial material and human support to overcome the pandemic crises.

Our study fills the gap in the literature concerning the correlations between KPIs in the healthcare sector during a pandemic crisis. Moreover, the managers can establish realistic goals by using KPIs taking into account their level of importance as early-warning indicators that can point out forthcoming changes in the evolution of the crisis. Healthcare managers can use the clusters KPIs to evaluate executive performance and to develop strategies for saving lives.

In the last years, the KPIs were found as representative of overall healthcare systems around the world [24]. Paradoxically, the qualitative differences between healthcare systems around the world are reduced by the particularities of the crisis, because a crisis is a negative phenomenon at globally level and its repercussions are more or less evaluated by KPIs [27]. Moreover, during a pandemic crisis, the information plays an important role in reducing physiological and mental impact on the people [44] and for this reason, it is necessary to use the KPIs clusters in a flexible way and to adapt them to healthcare facilities [45].

The results could provide a guide to hospitals' decision-makers in order to have under control the situation of the Algerian healthcare system [46] and for the other countries healthcare systems, because our findings are in consensus with the results of the other researchers [47–49]

The limitations of our study are related to the sample size, because, considering the short period of analysis, we used a relatively small sample size, from Algerian hospitals and the findings should be critically analysed by considering the specificity of Algeria.

This study only analysed the relationship between four groups of KPIs without testing the direct impact of these KPIs on the performance of the Algerian hospitals.

Our Clusters of KPIs model can be adapted to healthcare systems from different countries, but it is important that every healthcare facility choose its own KPIs taking into account their human, financial and technical resources [50].

Future research should be oriented to testing the role of KPIs in the improvement of the hospitals' performance in relationship with the social responsibility and with the improvements of the commitment of the healthcare systems employees.

**Author Contributions:** Conceptualization, A.B.-S. and K.F.; methodology, A.B.-S. and K.F.; validation, A.B.-S. and K.F.; formal analysis, K.F.; investigation, K.F.; writing—original draft preparation, A.B.-S. and K.F.; writing—review and editing, A.B.-S. and K.F.; supervision, A.B.-S.; project administration, A.B.-S. and K.F. All authors have read and agreed to the published version of the manuscript.

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

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

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author. The data are not publicly available due to Koudoua Ferhati that is a PhD Student and she we will use the data for her final thesis.

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

#### **Appendix A**

**Table A1.** Social Sustainability Indicators (SSI).


#### **Appendix B**

**Table A2.** Economic Sustainability Indicators (ESI).


#### **Appendix C**

**Table A3.** Indicators of Internal Process (IIP).


#### **Appendix D**



#### **Appendix E**


**Table A5.** Code, Factor Loading and Variance Inflation Factor of KPIs.


#### **Table A5.** *Cont.*

#### **Appendix F**

#### **Table A6.** The importance of KPIs in pandemic crisis.



**Table A6.** *Cont.*

#### **References**

