Relationship between Job Satisfaction, Burnout, and Depressive Symptoms in Physicians: A Cross-sectional Study based on the Employment Demand Control Model

We evaluated the relationship between job satisfaction, burnout syndrome, and depressive symptoms among Peruvian physicians using the job demand–control framework. We carried out a secondary data analysis of the National Survey of Satisfaction of Users in Health 2016 in Peru. The rst part was based on the association among the variables based on the job demand–control framework. The second part focused on estimating the acceptability of the proposed model based on the study variables through the structural equation modeling technique. A total of 2,100 physicians were included, and the prevalence of depressive symptoms was 3.3%. Physicians who had a work-related illness were more than twice as likely to have depressive symptoms (PR=2.23) compared with those who did not. The rst predictive model based on the variables, depressive symptoms, burnout syndrome, and job satisfaction had a low goodness-of-t index. Therefore, for a second evaluation, models with correlated errors were considered, and optimal goodness-of-t indices were found (CFI=0.974; RMSEA=0.060). Our study identied a stable model to explain the relationship between job satisfaction, burnout, and depressive symptoms among physicians. The results are consistent with the job demand–control framework and can be applied to decision making in occupational contexts in Peru.

Regarding the factors in uencing poor mental health outcomes among health care workers and physicians, existing evidence suggests that a number of them are directly work related 16 . These include working at emergency services, dealing with the patients' suffering and death, dealing with increased risk of infection 17 , or organizational factors, such as work overload, tensions between personal and institutional values 18 , inadequate quality of health care, and low professional autonomy 19 . More speci cally, job satisfaction is a complex and multidimensional concept that can be de ned as "the positive or negative attitude that an employee has towards working conditions, co-workers, the exercise of his or her profession in the work place" 20 . It is necessary to point out that although job satisfaction is a construct based on the worker's perception, it depends directly on the actual conditions of the workplace. Job satisfaction has been found to be moderately related to the presence of mental health problems, such as anxiety, depression, or burnout syndrome 21 , and conversely, the absence of burnout has been found to be a predictor of job satisfaction 22 . The existing evidence seems to point toward low job satisfaction being a factor in poor mental health outcomes and burnout among physicians. Job satisfaction among health care workers is crucial to public health, as it has been shown to reduce absenteeism-related costs and improve patient treatment care and satisfaction 23 . Hence, promoting high levels of job satisfaction among physicians is an essential aspect of ensuring that health care systems can deliver appropriate and safe care, thus improving overall public health 24 .
In recent years, several theoretical models have been put forward to de ne and explain the relationship between burnout syndrome, job satisfaction, and depression. However, empirical evidence remains scarce, and the few available studies are cross-sectional. Furthermore, the lack of a uni ed measurement instrument across countries limits existing interpretations 15 . Contrarily, different theoretical frameworks have attempted to explain the relationship between working conditions and work-related variables in mental health. One of the most critical of these frameworks is the job demand-control model, which predicts that "individuals with high-strain jobs characterized by high demands and limited control (decision authority and skills discretion) and possibly poor social support in the workplace (isolated strain: iso-strain) are at high risk of developing physical illness and mental health problems" 25 .
Some studies provided initial evidence on the structure of the relationship between the three variables, burnout, job satisfaction, and depression, in physicians and its consequences. Systematic reviews have shown that the perception of psychosocial stressors in health care facilities is related to the development of depressive symptoms in health care professionals 25,26 , and cross-sectional studies suggest that job satisfaction is associated with occupational factors and physicians' mental health [27][28][29] . The theoretical framework is used to create predictive models in cross-sectional studies and elaborate structural equation models (SEM) to explain the relationship between burnout, job satisfaction, and depressive symptoms in physicians. Using the SEM model, a previous study in health care workers in Brazil indicated a causal relationship between these three variables, where depressive symptoms were a predictor of burnout syndrome, and the absence of this syndrome predicted job satisfaction 22 .
Despite the existing evidence on the relationship between job satisfaction, burnout, and depression, there is no primary evidence of this relationship in middle-and low-income countries, including Peru. Since the existing evidence on burnout among physicians in Peru indicates high prevalence [10][11][12] , studying its relationship with job satisfaction and depression is imperative. Additionally, in the Peruvian health care system and socioeconomic context, inequities exist in health outcomes and access to health care 30,31 that may not be present in high-income countries where some of the evidence on the relationship between job satisfaction, burnout, and depression have been reported. It is important to note that the Peruvian health system is fragmented and segmented between public and private health care coverage and provision, with three levels of governance: national, regional, and municipal. More than 75% of the population has public health insurance in subsidized and contributory regimes 32 . According to critics, despite being a pioneer at the regional level in the establishment of a National Health Services System, the Peruvian health care system is characterized by structural ine ciencies, lack of continuity in care, lack of monitoring and evaluation 33 , insu cient public spending, and insu cient and geographically dispersed human resources and infrastructure 31 . Furthermore, physicians often work in more than one health care center, often sharing their working time between the public and private subsectors 34 .
Considering the systemic issues observed in the Peruvian health care system and the prevalence of burnout syndrome and mental health issues among physicians, which hinders their capacity to provide adequate services, failing to address their mental health may negatively affect population health in the long run while widening the existing gap in access to health care and adequate health insurance coverage 32,33 .
Within this context, our study evaluated the relationship between job satisfaction, burnout syndrome, and depressive symptoms among Peruvian physicians using the job demand-control framework model. It is the rst analysis of its kind in this country and has contributed to developing relevant knowledge for middle-and low-income countries in Latin America and other regions on the factors of common mental disorders among physicians. This study could also contribute to identifying solutions that would guarantee the health of health care professionals and consequently improve the overall population health.

Sociodemographic characteristics
The participants' sociodemographic variables were obtained from the database. A total of 34.5% were over 50 years old. Majority of the participants were males (68.8%), earned between 4 and 10 times the minimum wage per month (66.0%), and lived with a partner (64.5%). Around half (51.8%) had a medical specialty. Of all participants, 58% reported working in other institutions. Of these, 43% worked with the Ministry of Health, 34.4% were part of EsSalud, 8.4% worked in the armed forces and national police, and 14.2% worked in private clinics. Of all participants, 23% reported a work-related illness, 30.3% reported having a chronic illness, and 34.7% reported that they have suffered physical, psychological, or sexual violence ( Table 1). The proportion of physicians with depressive symptoms was 3.3%. Table 1 Characteristic of the participants included in the study (n=2,100). Note: Monthly income = Less than four minimum wages (≤$890), four to ten minimum wages ($890 to $2,225) or more than ten minimum wages (≥$2,225). The complex sampling weighting factor adjusted the percentages. * Model adjusted by sex, age, live with the couple, speciality, work in other institutions, monthly income, work-related illness, self-reporting having a chronic illness, type of organization, years working in the institution, and physical, psychological or sexual violence. Words in bold are signi cant values (p<0.05). RP = Prevalence Ratio.

Relationship between variables
We found that physicians who had had a work-related illness were more than twice as likely to present with depressive symptoms, compared with those who did not (PR: 2.23, 90% CI: 1.38-3.60) ( Table 1). The subdimensions of burnout were strongly correlated (|r| > 0.50), as well as the elements of depressive symptoms (r > 0.50). The job satisfaction scales were moderately correlated (r > 0.30). The general professional activity satisfaction scale and the dimensions of burnout were moderately correlated ( Table  2). Furthermore, anhedonia and the individual working conditions subdimension were moderately related to the dimensions of burnout (|r| > 0.30). There was a weak correlation between depressive symptoms and the three scales of job satisfaction, consistent with the job demand-control framework.

Structural regression model
The rst predictive model for depressive symptoms, burnout syndrome, and job satisfaction variables ( Figure 2) presented a low goodness-of-t index. This led to assessing the possible models with correlated errors, and we found that the errors of individual conditions and health services management were highly related. Hence, we decided to assess a second model taking into account these correlated errors (see Figure 2), and we found optimum goodness-of-t indices (CFI = 0.974; TLI = 0.961; SRMR = 0.045; RMSEA = 0.060 (Table 3). Note: X 2 = chi-squared; df = degree freedom; CFI = comparative t index; TLI = Tucker-Lewis Index; RMSEA = root-mean-square error of approximation; 90%CI = 90% con dence interval; SRMR = Standardized root mean square residual. The model values considered the weighting factor of the complex sample.
Our study established an explicative model in which job satisfaction negatively predicted burnout syndrome, and burnout predicted the presence of depressive symptoms ( Figure 2). This means that physicians who had low satisfaction with working conditions, health services management, or their general professional activity were at a higher risk of developing burnout syndrome. In other words, they would display higher levels of depersonalization and emotional exhaustion and lower levels of personal accomplishment. These elements of burnout in uenced the presence of depressive symptoms. This explicative model explained the relation between emotional issues, such as depressive symptoms, and organizational variables based on the job demand-control framework.

Main ndings and signi cance of the results
Our results support the hypothesis that job satisfaction in uences burnout and that burnout in uences the presence of depressive symptoms among physicians in Peru. These results are consistent with the job demand-control framework, which suggests that high levels of tension in the workplace and demanding working conditions for workers who may or may not have tools to control this demand put them at a high risk of work-related illness. When demands exceed control capacities, accidents may happen, and illnesses may be caused 35,36 . Speci cally, our study showed that with demands of the work environment and working conditions represented by the three scales of job satisfaction in uenced the physician's capacity for control and individual autoregulation, which was assessed by the burnout scale.
In turn, burnout in uences the presence of depressive symptoms among physicians in Peru. This provides evidence of a relationship between the three variables and the relevance of this model for mental health disorders, such as depressive symptoms.

Explanation of the plausibility of ndings
Several studies have found causal mechanisms to explain the relationship between the variables of job satisfaction, burnout syndrome, and depressive symptoms 25,37 . Among the main mechanisms are the factors linked to work conditions. First, the variables related to work conditions are tightly interrelated and linked to job satisfaction 38 . This explains why job satisfaction regarding health services management and the individual working conditions of the health care center, which are highly correlated, were identi ed in our study. Second, there is evidence that work conditions like work overload, mismanagement, and inadequate work environment 25,37 are essential factors in burnout prevalence. Third, there is plenty of evidence supporting the relationship between burnout and depressive symptoms, as the former predisposes one to negative affective feelings, such as sadness and anhedonia [15], potentially initiating clinical depression. Fourth, failure to treat problems in the work environment, in a timely manner, would lead to burnout syndrome and subsequently, the appearance of depressive symptoms 25,37 . Finally, there is evidence that the demand-control framework explains the relationship between job satisfaction, burnout, and depressive symptoms. Therefore, it is a valuable framework that could be used within the Peruvian context to explain depressive symptoms among physicians 25,33 .

Contrasting ndings within existing literature
Ensuring that health professionals feel satis ed with their work environment and the work itself is crucial as it contributes to ensuring the best possible quality of care for their patients. However, hostile work conditions, inadequate management, the perception of having limited opportunities for professional and personal growth, di culties in coping with tasks, among others, can lead to burnout syndrome. It is considered an illness, and health care providers must pay attention to it, as it affects both the health care workers and users of health services and leads to economic loss 22 . Thus, constant work-related stress and the inability of the employer to satisfy the needs and ful ll the rights of its workers can lead to depressive symptoms and other mental disorders. Furthermore, treating depression may span several years, resulting in nancial loss 39,40 . For this reason, it is crucial to address the root causes of depression and its symptoms, that is, the work conditions in the proposed predictive model. In this way, a range of negative consequences could be prevented at the national level, paving the way for improvement in public policies, laws, and regulations.

Relationship and comparisons between variables
Regarding sociodemographic factors associated with depressive symptoms among physicians, our study showed that the physicians who reported work-related illness were twice at risk of presenting with depressive symptoms. Other studies found that work-related illnesses, such as workplace accidents or back pain 41 , increased the patient's risk of developing depressive symptoms. Contrarily, age, gender, and monthly income did not show signi cant results. Previous studies in Peru showed that women over 75 years old and those belonging to the lowest wealth quintile are more likely to have depressive symptoms 42,43 . If wealth is to be considered alone as a factor in depressive symptoms, it could be said that health professionals would be at a lower risk of developing depressive symptoms. Physicians receive a higher income than the national average, on average four times the minimum monthly wages, and thus have better access to mental health care and higher educational levels, all of which are protective factors for depressive symptoms 42 . However, other studies among health professionals in Brazil, China, and Finland reported that sex, educational level, age, or the speci c profession (physician, nurse, or other health professionals) are not related to developing depressive symptoms 22,29,44 , which is consistent with our results.

Structural regression model
Our results support the job demand-control framework, which shows that workers with high levels of tension (burnout) as well as unfavorable organizational conditions (low job satisfaction), are at higher risk of developing health issues, such as depressive symptoms 25 . Research carried out in Brazil and Spain, which used models focused on job satisfaction, burnout, and depressive symptoms, found that the direct relationship between depressive symptoms and job satisfaction is minimal 22,45 . This supports our model, as their effect is mainly mediated by burnout. The abovementioned studies have put forward other types of models. The Brazilian study that focused on health professionals suggested that depressive symptoms predict burnout (n = 271), whereas the other study that focused on the members of the Spanish Scienti c Society of Periodontology found that depressive symptoms negatively in uenced personal accomplishment, which is a dimension of burnout (n = 170) 45 . However, these studies had two important limitations. First, neither of them used a framework that can explain their models, meaning that the directionality of the variables studied is not clear. Second, their respective sample size is minimal and not representative, and consequently, there may have been a selection bias. Thus, the model presented in our study represents a relevant contribution to the existing scienti c literature, considering that the job demand-control framework explains the results and can be generalized to Peruvian physicians.

Strengths and limitations
The strength of our study is the representativeness. We used a sample of primary care physicians in Peru, which means that our results can be generalized to this group. However, our study had some limitations.
First, the method used was cross-sectional, meaning that it is not possible to determine causality. Second, only physicians were included in the sample, excluding other health professionals, such as psychologists and nurses. Third, although validated instruments were used to avoid bias, this does not replace a clinical assessment of depression. As such, there may have been selection bias; however, the PHQ-2 is an instrument that has been shown to have high sensitivity and speci city scores 46 . Fourth, the data is from 2016, which indicates that these values may have changed in recent years. We however consider this very unlikely due to the nature of the Peruvian health care system 33 .

Implications in public health and recommendations
Peru is a lower middle-income country, and its health care system is overburdened and fragmented. Health care institutions are divided among different regulating entities and providers (Ministry of Health, EsSalud, the armed force and national police, and the private sector), with considerable disparities in terms of the resources available for health care provision. Therefore, our results are helpful for the Peruvian context but may also be relevant for other lower middle-income countries with a similar health care system. Our results showed that mismanagement, poor social support, and recognition of the needs and rights of health care professionals (low job satisfaction), and di culties to control high job demand may lead to burnout syndrome among health care professionals. If these issues remain unresolved, there may be adverse effects on the health of Peruvian physicians, such as a higher prevalence of depressive symptoms, and this may lead to years lost to disability as well as considerable costs for the state due to the limited human resources for health, absenteeism from work, and work saturation and overload 33,47 .
We thus recommend that policies and regulations be designed to improve the work conditions of health professionals.

Conclusion
Our study identi ed a stable model to explain the relationship between job satisfaction, burnout, and depressive symptoms among physicians. Policymakers could use this model to assess areas for improvement, such as working conditions. Improving management may prevent burnout syndrome and depressive symptoms among health professionals.

Study design
We

Depressive symptoms
The Patient Health Questionnaire (PHQ-2) was used to evaluate depressive symptoms in the last two weeks. This questionnaire had only two items, measured on a 4-point Likert scale (0 = not at all; 1 = several days; 2 = more than half of the time; 3 = nearly every day), and the scores ranged from 0 to 6. Using three or more points as a threshold, the PHQ-2 de ned depressive symptoms with good sensitivity (82%) and speci city (90%) 46 .

Sociodemographic variables
Data was collected on age, sex (male and female), whether the physicians had a specialty (no, in progress, and yes), whether they worked at another institution teaching, seeing patients, or performing administrative tasks (yes or no), whether they had a chronic illness (yes or no), whether they had a workrelated illness (yes or no), whether they had been victims of physical, psychological, or sexual violence in the workplace (yes or no), the type of organization in which they worked (Ministry of Health, EsSalud, armed forces and national police, or private clinics), and time spent working. Additionally, self-reported monthly income was evaluated and categorized according to the minimum wage (less than 4 times, 4-10 times, and more than 10 times the minimum wage). The minimum wage was 750 Peruvian soles (PEN), equivalent to about US$ 222.5 as on November 2020.

Analysis
Descriptive analysis Descriptive analysis of variables was performed, and a separate analysis was carried out for physicians with depressive symptoms (PHQ-2 scores ≥ 3 points) and for physicians without depressive symptoms (PHQ-2 scores ≤ 2 points). The analysis was adjusted by the weighting factor of the complex sampling.

Relationship between variables
A Pearson correlation was performed using a sample size weighting between the following variables: depressive symptoms (PHQ-2), burnout (MBI-HSS), and job satisfaction (three scales). These three variables were evaluated because they correspond to the theoretical framework proposed by Gray, Senabe, Naicker, Kgalamono, Yassi and Spiegel 47 , and Rothenberger 37 . Pearson's correlation coe cient was considered as effect size, considering weak (r = 0.10), moderate (r = 0.30), and high (r = 0.50) values 50 .
The chi-square tests were used to compare the relationship between sociodemographic characteristics and the variables of interest. Adjusted prevalence ratios (PR) were calculated using generalized linear models with robust variance estimations, assuming a Poisson distribution with log link functions 51 .
Potential confounders included in the adjusted model were sex, age, living with a partner, having a medical specialty, working in more than one institution, monthly income, work-related illness, self-reported chronic illness, working for the Ministry of Health, EsSalud, the armed forces and national police, or a private clinic, years of work in the institution, and experiences of physical, psychological, or sexual violence.

Structural regression model
For this analysis, we used the maximum likelihood estimation with robust standard errors 52 and Pearson matrices 53 . We adjusted the analysis by the weighting factor of the complex sampling. Two models were evaluated based on the hypothesis that job satisfaction in uences burnout syndrome, which in uences depressive symptoms 37,47,54 . The rst model (model 1) evaluated the relationship between job satisfaction, burnout, and depressive symptoms. The second model (model 2) was based on model 1 but considered correlated errors between the dimensions. Because the variable had four dimensions, it was very likely that these dimensions were strongly associated, and correlated errors were found between them 51 . The reason for using structural regression models over bivariate models was that they allowed for a single overall analysis of the different relationships among all the variables.
Evidence has been found that job satisfaction and burnout affect the mental health of health professionals, and a model was proposed to explain the dynamics between these three variables 55 , with labor dissatisfaction as the main predictor 22 . However, this study presents a model that explains how depressive symptoms occur in Peruvian doctors due to work dissatisfaction. This model explains a problem that persists in the health systems of several countries. Therefore, understanding this problem may help in the making of more e cient and effective interventional decisions 22,26 .
The models were evaluated based on different goodness-of-t indices. The comparative t index (CFI) and Tucker-Lewis Index (TLI) were used; values were considered optimal when they were higher than 0.95. The Standardized Root Mean Square Residual (SRMR) and Root Mean Square Error of Approximation (RMSEA) with a con dence interval of 90%, both with values adequate if <0.08 51,56 .

Ethical consideration
The survey was anonymous, and no information in the database could be used to identify the participants. Hence, conducting this analysis did not represent an ethical risk for participants since there was no access to con dential data. The study did not require the approval of an ethics committee because it came from a secondary database that is open access. Therefore, no primary data collection was performed and the study did not involve an ethical risk for participants. Ethics approval and consent to participate:

Abbreviations
The survey was anonymous, and no information in the database could be used to identify the participants. Hence, conducting this analysis did not represent an ethical risk for participants since there was no access to con dential data. The study did not require the approval of an ethics committee because it came from a secondary database that is open access. Therefore, no primary data collection was performed and the study did not involve an ethical risk for participants.

Consent for publication:
Not applicable.
Competing interests:   Figure 1 Flowchart of the participants included in the study.