Association Between Physical Activity and the Risk of Burnout in Health Care Workers: Systematic Review

Introduction

Background

Burnout is a multidimensional psychological syndrome resulting from chronic workplace stress. It is characterized by feelings of energy depletion or exhaustion, increased mental distance from one’s own job or cynicism, and reduced professional efficacy [1].

Burnout is gaining attention as a major public health concern for the mental health challenges of health care workers (HCWs) [2]. It poses a threat to the quality of care delivery, especially in terms of patient safety [3], and results in high resource consumption to face its consequences [4,5]. This became even more evident during the COVID-19 pandemic [6], which tightened physicians’ work conditions [7,8], significantly increasing the share of physicians and nurses expressing an intention to leave the profession [9,10], and highlighted the frailty of health care systems in emergency management [11]. Although COVID-19 no longer constitutes a public health emergency [12], HCWs continue to be particularly vulnerable to burnout [13,14], with weakened abilities to manage the typical difficulties of care work and increased exposure to emotionally challenging situations [15]. This is especially true for those directly engaged in patient care given the demanding nature of their roles [16]. They often work long hours, experience sleep deprivation, contend with irregular schedules, and are exposed to emotionally challenging situations. In addition, they face the pressure to master a vast body of clinical knowledge [17]. Furthermore, HCWs are at high risk of workplace injuries [18] and assaults [19,20]. These challenges are compounded by common life stressors such as work-home conflicts, educational debts, relationship status, the age of their children, and the employment status of their partners [21]. As reported in a literature analysis [6], burnout can result in various negative outcomes and criticalities for HCWs, including anxiety, depressive disorders, alcohol abuse or dependence, and suicidal ideation. Burned-out workers experience poor mental health even without a clinically diagnosable disorder, as also proved by the fact that, within this group, there are more mental health problems than in most other occupational groups [22].

Within the broad spectrum of possible public health options to address burnout in HCWs, prioritizing measures to promote mental well-being has become paramount [23,24]. This entails addressing cultural factors, particularly those related to stigma; ensuring protected access to mental health care services; and implementing active policies to encourage healthy lifestyles [25,26]. Among these, physical activity is recognized worldwide as a key strategy for promoting mental well-being [27,28] in addition to helping prevent and manage noncommunicable diseases. However, unlike physical health [29], determining the “optimal dose” of physical activity—considering the combination of intensity, duration, and frequency [30]—for mental well-being remains uncertain and dependent on the specific domain of investigation [27].

A preliminary search of previous systematic reviews on the effects of physical activity on burnout was conducted in Scopus, PubMed, and PROSPERO and resulted in 2 different studies. Naczenski et al [31] found that physical activity was an effective strategy to reduce burnout among workers from various sectors. Regarding health professionals specifically, Bischoff et al [32] identified a potential beneficial effect in health professionals of mind-body practices such as yoga or qigong on occupational stress, one of the conditions for burnout [1].

Objectives

Our study systematically investigated how physical activity was incorporated into studies on burnout among HCWs. Recognizing that strategies created without evidence lead to ineffective programs, wasted resources, and persistently poor health outcomes [33], our ambition was to contribute to the development of evidence-based public health policies.

First, we aimed to provide insights into the reported level of participation in physical activity and the extent of participants at high risk of burnout. We also intended to verify whether a correlation emerges between levels of physical activity and burnout and whether a dose-response association exists.

Our second purpose was to thoroughly assess the quality of the available evidence in terms of collecting, compiling, managing, analyzing, and using health data. This assessment could also offer indications to generate hypotheses for further research to strengthen the body of evidence.

These are essential steps along the road that leads to shaping public health strategies and resource allocation for HCWs’ well-being.

Methods

As recommended for systematic reviews of association studies [34,35], we adopted the population, exposure, and outcome (PEO) approach by considering physical activity habits as the exposure factor and burnout as the outcome. To ensure accuracy and transparency, we followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) [36] and the PRISMA-S (PRISMA literature search extension) [37] guidelines.

Eligibility Criteria

The eligibility criteria are detailed in Multimedia Appendix 1.

We included studies that examined the association between physical activity and burnout in health personnel directly involved in the provision of care services, such as physicians, nurses, and technicians, with no restrictions on demographics or workplace context. We considered both qualitative and quantitative measures of physical activity as eligible for inclusion. As we were interested in the association between physical activity and burnout, interventional studies were included regardless of the presence of a control group. A validated assessment tool for burnout (a general work-related stress outcome was not of interest) was mandatory for inclusion.

Our search was not limited by geographic context, funding source, or time horizon. We excluded studies in which physical activity was not distinct from other supportive strategies aimed at managing stress or building resilience or from other mindfulness practices. In addition, studies that focused on yoga or qigong were excluded as these are classified as meditation practices with only a light-intensity component of physical activity when performing respiration and poses [38-40].

Following the recommendation by Munn et al [35] to clearly report the exposure or risk factor and how it was measured or identified, we excluded studies that proposed only a simple question on physical activity without any reference to its frequency (eg, a generic “Yes/No” question on practicing “exercise” or even “regular exercise”). Furthermore, we excluded literature reviews, gray literature, conference proceedings, and unpublished material. Finally, we excluded studies that lacked a quantitative evaluation of the association.

Information Sources

We searched the following electronic databases up to July 2022: MEDLINE through the Ovid platform, APA PsycArticles, and Scopus. Using a “snowballing” approach, we manually screened the reference lists of included articles and conducted systematic citation tracking in Scopus, PubMed, and Google Scholar.

Search Strategy

In total, 2 authors, CGL and PM, developed search strings using the PEO framework to structure the research question. We selected search terms to identify HCWs, physical activity interventions, and burnout. Regarding burnout, to increase search sensitivity, we also considered terms related to the assessment instruments indicated by Rotenstein et al [41] and those identified in the more recent works by Edú-Valsania et al [42] and Shoman et al [43]. Moreover, to further increase the sensitivity of our search, we used the National Library of Medicine–controlled vocabulary thesaurus (Medical Subject Headings) with entry terms and synonyms. We limited the search to articles published in English. The complete search strategy can be found in Multimedia Appendix 2 [41-43].

Duplicates were removed by PM using an automatic procedure based on PubMed ID and digital object identifier, and this was performed in Microsoft Excel (Microsoft Corp).

Selection Process

Search results were retrieved from the databases and double screened independently by all the authors. The initial screening was based on the title and abstract using Rayyan (Rayyan Systems Inc) [44] for support. To refine and clarify the eligibility criteria, ensure consistency when applied by different reviewers, and train the team, a pilot phase was carried out on 500 works. The potentially relevant articles were retrieved for full-text screening, and their eligibility was determined as described in the previous step. Any disagreement was resolved through plenary discussion among all the authors until a consensus was reached.

Data Collection Process

Each selected study was randomly assigned to and independently evaluated by 2 authors to extract relevant data. Any disagreement was addressed as mentioned previously.

Data Items

We extracted 3 different types of information. First, we recorded the general characteristics, such as aim, context (country, workplace, and period), and population (type of HCW, number, and gender mix). Second, we summarized the methods used to measure burnout, the criteria used to quantitatively summarize the phenomenon in the study population, and a possible definition of severity. When possible, we included the frequency distributions of categorical outcomes and means of total scores with SDs. Third, we noted the methods adopted to measure physical activity, including the assessment of frequency, time, or intensity of activity and the percentage of participants practicing physical activity. Finally, we reported the adopted measures of the association between physical activity and burnout along with its strength (P values, odds ratios, and their 95% CIs), eventually derived by AB from published data, and a synthesis of the reported evidence. In addition, factors other than physical activity that are significantly associated with burnout were listed. A first list was drawn from bivariate analyses and reported only variables that were significantly associated (P<.05) with burnout. The consideration of even weak associations was hampered by the often incomplete presentation of these analyses. A second list came from multivariable analyses and included all the variables considered in the final (or presented) models regardless of whether physical activity was included in the model itself.

Synthesis of Evidence

The data are presented in a tabular form.

A narrative approach was adopted to provide an overall summary of the findings of the included studies and their biases, strengths, and limitations, with an in-depth discussion of the causes of heterogeneity.

Quality

The quality of the studies included in the analysis was assessed using the Joanna Briggs Institute Critical Appraisal Checklist for Cross-Sectional Studies [45]. This checklist consists of 8 items that are rated as “yes,” “no,” “unclear,” or “not applicable.” To further refine our judgment, we added “partial” as a fifth option.

Owing to the eligibility criteria adopted, 2 issues were always rated positively: item 1 (“Were the criteria for inclusion in the sample clearly defined?”), as we only considered studies that clearly reported associations in health personnel directly involved in the provision of care services, and item 4 (“Were objective, standard criteria used for measurement of the condition?”), as we only included studies that used validated questionnaires for burnout.

Item 2 allowed us to register whether a complete description of the participants and setting was provided. We assigned a score of “partial” or “no” if some or all the information was missing.

To assess whether the outcomes were measured in a valid and reliable way (item 7), we verified that the burnout measurement tools were used consistently with the dictates expressed by their developers.

We evaluated the validity and reliability of the exposure measurement (item 3) by positively appraising the use of structured questionnaires (eg, the International Physical Activity Questionnaire–Short Form [IPAQ-SF]) or automatic measurement devices such as pedometers, whereas the use of ad hoc questions to register the frequency or intensity of physical activity was considered a partial achievement.

Concerning the confounding factors, we did not assess how they were considered (item 5) and addressed in the statistical analyses (item 6) in studies exploring multiple wide-scope associations with burnout. On the other hand, we applied these items in studies that focused on a specific exposure factor (either physical activity or another variable of interest). In such cases, the use of statistical modeling, such as multivariable regression analysis, was considered a partial achievement in the absence of an advisable dedicated discussion on confounding factors.

For the appropriateness of the statistical analysis (item 8), a partial quality level was assigned if multivariable regression analysis, though possible, was not conducted or when it was conducted without an adequate method of model selection. Moreover, we assigned a partial level of quality if it was not possible to clearly understand all the details of the analyses because of omissions or results not being clearly reported.

Each study was evaluated for quality by 2 authors, and AB, a statistician, also reviewed items pertaining to statistics.

Discrepancies were resolved through plenary discussion among all authors until a consensus was reached. The quality assessment was not taken into account for eligibility purposes.

Results

Selected Studies

A total of 8937 records were identified, and after removing 2422 (27.1%) duplicates (Figure 1), 6515 (72.9%) publications remained following the initial screening. Of these 6515 publications, 6427 (98.65%) were excluded based on the title and abstract. These records were disregarded as they did not address physical activity, burnout, or health care personnel or because they were conference proceedings, reviews, or nonoriginal research (eg, letters or commentaries). Of the 88 studies selected for potential inclusion, 3 (3%) were not retrieved as they were published in journals not accessible through our organizations even after writing to the authors to request the accepted versions of their manuscripts [46-48]. Following full-text screening, 75% (64/85) of the records were excluded. Consequently, 21 independent studies involving 15,782 HCWs were included in this review (see Table 1 for details).

‎

Although 10% (2/21) of the studies explicitly focused on exploring the association between physical activity and burnout [62,67], the remaining studies investigated the association among several factors, including physical activity. Most of the studies were conducted in America (10/21, 48%) and Asia (9/21, 43%), with only 10% (2/21) of the studies [49,52] being conducted after the spread of COVID-19. In total, 43% (9/21) of the studies [51,54-56,59,60,62-64] focused on residents. All the studies included in this review (21/21, 100%) had a cross-sectional design.

Demography of Study Populations

The study populations consisted of diverse groups of HCWs with varying proportions of women individuals, ranging from 5.6% to 96.3%. The professional categories represented included physicians, nurses, and technicians from different specialties, which were present in 90% (19/21), 38% (8/21), and 14% (3/21) of the studies, respectively. The sample sizes ranged from 60 to 7197 participants. The mean age varied from 27.2 (SD 1.83) years to 49.9 (SD 12.0) years, with a few studies (5/21, 24%) reporting the median ages, ranging from 29 to 53 years.

Measurement of Burnout

In terms of burnout measurement, the 22-item Maslach Burnout Inventory (MBI) was the most commonly used questionnaire. Some studies (3/21, 14%) used MBI-related measures based on fewer items [63-65]. In total, 10% (2/21) of the studies used different questionnaires: the Professional Fulfillment Index [54], which is designed to measure physician well-being [70], and the Copenhagen Burnout Inventory [61], which was proposed as an alternative tool to the MBI for measuring burnout [71] (see Multimedia Appendix 3 [49,51-53,55,56,58-63,66-69,72,73] for details).

The severity of the condition in the investigated population was reported in all the studies (21/21, 100%), albeit with differences in terms of the dimensions considered and the criteria used for deriving a summary assessment of burnout across the dimensions and, in some cases, by providing measures of location and scale parameters. Further details are provided in Table 2.

In two studies, authors adopted tools not referable to BMI. Ng et al [61] adopted the Copenhagen Burnout Inventory [71], which comprises 3 categories: personal, physical, and psychological exhaustion, work-related physical and psychological exhaustion, and patient-related physical and psychological exhaustion. The reported frequency distribution among study participants is 63%, 56%, and 35% respectively. No specific criteria were defined for burnout.

Eckstein et al [54] used the Professional Fulfillment Index [70], which consists of three response categories: Professional Fulfillment (PF), Interpersonal Disengagement, and Work Exhaustion (WE), employing 5-point Likert scales (“not at all true” to “completely true” for PF items and “not at all” to “extremely” for WE and Interpersonal Disengagement items). All responses are scored from 0 to 4. However, in the considered study, the PF domain was not taken into account. A binary classification (presence vs absence of burnout) is adopted according to the following criterion: the average score of Interpersonal Disengagement and WE ≥1.33 [70]. Participants meeting this criterion for burnout comprise 38%. Regarding Location and scale parameters, the median score for WE, Interpersonal Disengagement, and burnout is respectively 1.50, 0.83, and 1.00.

With due caution regarding the aforementioned heterogeneity of definitions and cutoffs, the percentage of participants classified as burned out varied from 7% to 83%.

Measurement of Physical Activity

None of the included studies used objective measurement tools, such as pedometers, to assess physical activity (Table 3).

Structured questionnaires that distinguished among different typologies of physical activity or ad hoc items were considered to investigate the habits of HCWs. The IPAQ-SF, the Ricci-Gagnon scale, and the Arizona Lifestyle Inventory were used in the studies by Olson et al [62], Chokri et al [52], and McClafferty et al [60], respectively. The IPAQ-SF was the only validated tool used, although it tends to overestimate actual physical activity levels [84]. A total of 33% (7/21) of the studies [50,53,61,63,67-69] used an ad hoc question to assess the achievement of a given benchmark, with a threshold frequency ranging from 1 day every 2 weeks [63] to every day [53]. In the remaining studies, physical exercise was measured in terms of frequency and intensity [65], frequency and time [57], only frequency [49,51,54-56,58,59,66], or only time [64]. When it was possible to report a distribution of frequencies or times, we considered the lowest category that did not indicate “no physical activity at all.” Whenever possible, we adopted the most widely used benchmark of physical exercise of “at least 1 day per week” to define physically active participants. Using this criterion, we found that the percentage of active workers ranged from 44% to 82% for residents and from 46% to 87% for the other categories.

Methods of Association Assessment

Bivariate associations between physical activity and burnout were assessed using usual methods. Table 4 provides further details. All studies except those by Olson et al [62], Vinnikov et al [67], Ghoraishian et al [56], Feng et al [55], and Tiwari et al [66] also conducted a multivariable regression analysis. In all 5 cases, a multivariable analysis could have been conducted to obtain adjusted odds ratios. Logistic and linear regression were equally used in 43% (9/21) and 33% (7/21) of the studies, respectively.

Existence and Degree of Association

Table 4 also presents the evidence of the association between physical activity and burnout. Owing to the high heterogeneity of the studies, a direct comparison of the results was not feasible. As previously mentioned, the 2 sources of heterogeneity were the diverse and sometimes vague definitions of physical activity used in the studies and the variations in the definitions and cutoffs for measuring burnout.

A total of 14% (3/21) of the studies [49,57,64] reported a few results that were not consistent with the data, and in one case [58], the authors referred to supplementary material that was not available on the journal web page. In total, 80% (4/5) [55,56,62,66] of the studies that presented only bivariate analyses indicated that a lack of physical activity was associated with high values of at least one component of the MBI. Similar results were reported by half (8/16, 50%) of the other studies that presented bivariate analyses regardless of the measure of burnout adopted.

In studies in which bivariate analysis was preparatory to multivariable regression analysis [50-54,58,61,63,68], physical activity may not have emerged as a predictor in the multivariable models [51,52,54,63,68]. This is sometimes due to the lack of a significant (albeit weak) association already in the bivariate analyses [52,54]. However, this result is also influenced by the different approaches to variable inclusion and selection. When conducting a bivariate analysis as a preliminary step for variable selection in the multivariable model, a less stringent significance criterion than P<.05 should be considered. Variables that show weak individual associations can become important predictors when considered jointly (eg, the study by Hosmer and Lemeshow [85]). Unfortunately, the selected studies used very different criteria, ranging from P<.05 for each variable to no preselection at all. In addition, methods for model selection were not always applied or clearly stated.

In cases in which physical activity was found to be a significant predictor in multivariate analyses, a protective effect was observed, especially against emotional exhaustion [50,59,61,68,69] and depersonalization [50,51,53,59]. This held true across a range of activity frequencies, from as little as 1 session per week to daily engagement.

For the sake of completeness, factors other than physical activity that showed a statistically significant association with burnout have been listed in 2 separate tables: one for bivariate analyses and the other for multivariable analysis (Tables S1 and S2 in Multimedia Appendix 4). The predictors were grouped by topic (demography, health conditions, lifestyle, personal attitude, work-life balance, work organization and environment, work profile, and self-perception at work) and ordered within each class according to their frequency.

Quality Evaluation

The risk of bias was assessed using the Joanna Briggs Institute tool, and the details are presented in Multimedia Appendix 5 [45,49-69].

All the studies provided detailed descriptions of participants and settings, with the exception of the studies by Eckstein et al [54], in which gender specifications were missing, and Reed et al [64], in which the period of the survey, gender mix, and age were not reported.

Regarding the assessment of physical activity, only a minority of studies (3/21, 14%) [52,60,62] adopted structured questionnaires and, therefore, received a positive evaluation. Most studies (18/21, 86%) only partially fulfilled this criterion, either relying on simple ad hoc questions to assess the activity frequency, time, or intensity or referring to meeting a threshold of activity. This limitation was observed even in studies that claimed to have a specific focus on physical activity [55,67].

Confounding was never explicitly addressed. As mentioned in the Methods section, we applied the dedicated items to the 14% (3/21) of studies that focused on a specific exposure [62,67,69]. All these papers touched on this issue through their conducted analyses and, therefore, obtained a “partial” rating for item 5. Of these studies, 33% (1/3) conducted only a bivariate analysis and, thus, did not comply with item 6 [62].

As for the use of the MBI, only a few studies (9/21, 43%) received a positive evaluation for item 7. Contrary to the explicit indications of the MBI developers [86], the authors of most studies (11/21, 52%) [50,51,55-58,62,63,66-68] adopted a categorical classification of participants based on burnout dimensions for descriptive as well as inferential purposes. Moreover, Yang et al [69], although correctly adopting the score to assess each MBI dimension, defined and adopted an overall burnout score as the raw sum of the scores in each dimension (as reported in Table 4, even without considering that personal accomplishment should be interpreted in the opposite direction from emotional exhaustion and depersonalization). In all these cases except the studies by Ghoraishian et al [56] and Panse et al [63], we assigned a partial achievement of the item. Indeed, these 2 studies were ranked with a “no” as they did not even provide the rationale behind their choice of ranges of scores in each dimension. Except for 14% (3/21) of the studies—Ajab et al [49] and Vinnikov et al [67,68]—all the other studies referring to the MBI or related multi-item measures introduced a definition of overall burnout. Despite this, studies that also adopted the total score of the burnout dimensions in their inferential analyses [52,53,59,60] were ranked with a “yes.” Ng et al [61] and Eckstein et al [54] were classified as “yes” as they correctly adopted the measurement tools—Copenhagen Burnout Inventory [73] and Professional Fulfillment Index [54], respectively.

Regarding item 8, statistical analysis was not considered appropriate in 10% (2/21) of the studies [50,58], in which the significant differences between physicians and nurses in the dimensions of the MBI were not adequately handled in the multivariable analysis. The analyses conducted in another 14% (3/21) of the studies [49,56,64] were rated as unclear either because the presented results were not coherent with the declared statistical methods [64] or because of unexplained methods [49,56]. Goldberg et al [57], Bin Dahmash et al [51], Shanafelt et al [65], Vinnikov et al [68], Portero de la Cruz et al [53], and Chokri et al [52] clearly explained their statistical methods and conducted comprehensive analyses, receiving a positive quality assessment. In the other cases, a better statistical methodology could have been applied by conducting a multivariable analysis, considering more suitable criteria to include the explanatory factors in the multivariable analysis, or again by applying model selection methods to obtain more parsimonious and general models.

Discussion

Principal Findings

The COVID-19 pandemic has taken a heavy toll on HCWs in terms of physical and mental health, and the long-term effects of the pandemic will further increase the burden of HCWs’ mental health disorders and burnout [87,88]. It has been argued that addressing burnout could serve as a nonstigmatized and systemic approach to address a long-standing issue in medicine through mental health initiatives, whether prevention oriented or treatment focused [89].

In addition to the personal consequences for HCWs experiencing burnout, it is important to consider its impact on patient care and resource consumption. Emotionally and physically healthy HCWs are among the most relevant factors influencing health care service quality [90]. All these considerations emphasize the importance of implementing strategies to prevent and manage burnout at an individual, organizational, and cultural level. The literature on burnout management primarily presents fragmented solutions that are infrequently tested in practice. These solutions often align with 1 of 2 predominant lines of intervention: one emphasizes strengthening individual capabilities to navigate the inherent challenges of health care work, whereas the other acknowledges work organization as a contributing factor and attempts to intervene at that level. However, there is a growing realization that individual and organizational well-being are intricately interconnected, thus necessitating systemic solutions. These comprehensive approaches encompass organizational interventions, instilling a culture of well-being in the workplace, and integrating well-being into health care education for true efficacy [24,91]. A recent meta-analysis of 20 controlled trials found that the most effective existing interventions for reducing burnout were those targeting multiple facets of well-being [92]. Regrettably, these systemic solutions are complex and often come at a significant cost [93-95]; therefore, it is particularly relevant for research to precisely identify the characteristics that specific interventions must have to succeed. This review focused on physical activity as an effective factor in fostering a culture of well-being among HCWs, which is crucial for tackling the physical and mental consequences of work-related stress [24]. Physical activity has enormous potential to mitigate the physical and mental impacts of work-related stress [96,97]. Indeed, it is suggested that it facilitates psychological detachment from work and enhance self-efficacy [53], providing an opportunity to divert attention from stressful thoughts [50]. It has also been shown that moderate-intensity exercise training programs improve feelings of vigor, energy, and vitality [98]. In particular, greater effects occurred when combining resistance exercise with aerobic exercise compared with aerobic exercise alone [98]. This finding is consistent with the World Health Organization recommendations for physical activity, which underlie the importance for adults of regularly performing both aerobic and muscle-strengthening activities to support health, including mental health outcomes [96]. With this review, we aimed to systematically assess the strength of the evidence and, eventually, the dose-response association between physical activity and burnout in a physically and emotionally exhausting work environment. However, it is essential to note that all studies included in our review had a cross-sectional design. As is well known, randomized controlled trials with a longitudinal perspective are the gold standard to highlight any potential cause-effect relationship between an exposure and an outcome. Nevertheless, the cross-sectional approach is useful in highlighting potential relationships between burnout and related factors, aiding in the identification of a multiplicity of risk factors and mitigation strategies, as the complexity of the phenomenon requires. This is particularly true in our case as most of the included studies (19/21, 90%) were exploratory investigations that examined various aspects potentially related to burnout rather than focusing specifically on physical activity. As shown in Tables S1 and S2 in Multimedia Appendix 4, a wide range of variables were considered in bivariate analyses (74 issues) and included in multivariable models (58 issues), reflecting the interplay between burnout and demographic characteristics, health, lifestyle, personal attitude, self-perception at work, work organization and environment, job profile, and work-life balance.

As mentioned previously, physical activity is considered a potential tool to cope with distress, and the PEO approach adopted in this study was specifically structured in this direction. However, it is important to recognize the possibility that high levels of burnout negatively influence the level of physical activity [99]. According to the study by Olson et al [62], the “lack of energy” among burned-out residents suggests that high levels of burnout lead HCWs to reduce the time dedicated to leisure activities, including physical exercise. Stults-Kolehmainen and Sinha [99] found a similar result in the literature, stating that stress hinders individuals from being more physically active and has a negative influence on other health behaviors, including smoking, alcohol, and drug use. We suggest that future studies include sections dedicated to exploring the reasons for individuals’ inability to meet their desired level of physical activity. Factors such as time constraints, lack of interest, or underlying health conditions should also be accounted for. The Barriers to Being Active Quiz developed by the US Centers for Disease Control and Prevention [100] is an example of a tool that may support the exploration of the dynamic interactions among personal, socioenvironmental, and behavioral factors, serving as a knowledge base for promoting more active and healthy lifestyles.

This study confirms that the dose-response relationship between physical activity and psychological well-being and health-related quality of life is far from being established. Significant mental health benefits could be achieved even at physical activity levels below the public health recommendations [101,102]. This also emerged from our study and has direct implications on healthy lifestyle recommendations, especially for inactive HCWs for whom incorporating brief bouts of physical activity into daily activities may be a more realistically achievable goal than meeting the guideline-recommended physical activity levels. This can be crucial to promote physical activity and, therefore, trigger a virtuous circle with benefits for burnout. In addition, it may have broader implications for the general population as there are indications that physicians’ involvement in physical activity is linked to their propensity to advise patients on the advantages of exercise [103,104]. Recognizing that any physical activity is better than none and considering engagement in physical activity as a modifiable behavior, adopting a strategy of gradually increasing activity through small habit changes is deemed effective for establishing a consistent exercise routine. This approach can be facilitated by the use of activity trackers [105,106].

This systematic review highlighted another general challenge in quantifying the strength of the physical activity–burnout association and establishing the dose-response curve, primarily as various methods were used to detect burnout. Different self-reporting tools are indeed available grounded on different theoretical bases [42]. Most of the studies in our review (19/21, 90%) proposed the 22-item MBI or MBI-related questionnaires (Table 2), confirming the substantial monopoly of this tool in burnout research. All these studies fell into some form of misuse of the index. The 3-factor structure of the MBI implies, on the one hand, that each dimension must be treated separately and, on the other hand, that none of them should be ignored. In contrast, in some studies where it would have made sense (3/17, 18%) [59,60,62], personal accomplishment was excluded from the analysis. This practice is not uncommon as several authors consider personal accomplishment not as a reaction to stressful situations but rather as a personality trait or coping resource and, therefore, as not contributing to the comprehensive concept of burnout [107]. Another misuse of the MBI is its diagnostic application, which erroneously considers the MBI dimensions as symptoms of burnout [71]. The 7-point scale (from Never to Every day) used to report the feelings experienced by the respondents was intended by Maslach and the coauthors of the tool as an operational simplification of the measurement of a dynamic phenomenon evolving continuously over time rather than as the assessment of a dichotomous condition (absent or present) defined through an arbitrary cutoff. Considering MBI scores for diagnostic purposes would inevitably invoke the wrong concept of burnout as a disease or disability, ignoring decades of research and the recent statement from the World Health Organization [1]. The MBI was not designed as a diagnostic tool [86], and the cutoff scores established to classify people at low, moderate, and high levels of burnout were “intended primarily as feedback for individual respondents.” These scores were published up to the third edition of the MBI Manual [72] accompanied, however, by a strong recommendation to use the original total scores for any statistical analysis. In the fourth edition released in 2018 [77], the categorization was finally removed. Therefore, it is surprising that, despite the extraordinary diffusion of the MBI, there was a failure to implement the correct instructions for its use. In fact, all studies except the one by Goldberg et al [57] considered some classification of the severity (sometimes referred to as risk) of burnout in each of the 3 MBI dimensions also for inferential purposes. Furthermore, a variety of classification criteria, even when taken from the same reference [72], and algorithms for combining dimensions into a single overall burnout indicator, typically a high score in at least one dimension and sometimes in all dimensions (reversing personal accomplishment, if considered), further increased the heterogeneity of the analyses. These methodological considerations become even more crucial in future research, particularly now that apps providing burnout self-diagnosis are available and transparency is needed in the adopted assessing algorithms [108].

Moving on to the evidence on the association between physical activity and burnout that emerged from this review, the most compelling results came from multivariable analyses that considered emotional exhaustion and depersonalization as the outcome. In these analyses, physical activity appeared to be associated with a reduction in critical conditions. However, it is still unclear whether this association depends on the type, intensity, duration, or frequency of physical activity, as previously mentioned. Some of the included studies (2/21, 10%) suggested that exercising for at least one day per week is sufficient to see a positive effect [58,69], whereas others (2/21, 10%) suggested a frequency of 3 days per week [50,68]. Lebensohn et al [59] observed that, the more frequently HCWs engaged in physical activity, the greater the positive effect.

It is worth noting that, in a few of the studies that considered a multivariable analysis (2/15, 13%), physical activity was not included among the predictors. However, this can sometimes be related to strict variable selection methods.

One limitation of most of the included studies (18/21, 86%) was related to the modality of physical activity measurement. Future studies should consider using objective measurement tools such as pedometers or validated questionnaires such as the International Physical Activity Questionnaire [79] even in its short version [84]. These approaches enhance the accuracy and reliability of data collection. On the other hand, relying on simple questions about regular physical activity or adherence to recommendations for a healthy lifestyle, although suitable for preliminary exploratory investigations, limits the comparability of studies.

The use of precise and detailed definitions to collect measures of intensity and frequency can prove to be a valuable strategy to delve deeper into the dose-response relationship.

Limitations

Our systematic review has some limitations that warrant consideration. First, our search was confined to 3 databases and only considered English-language articles. Moreover, we did not perform a rerun of the search shortly before submission. However, to mitigate the risk of overlooking relevant papers, we used forward and backward citation tracking, including the use of Google Scholar.

Another potential limitation arises from our eligibility criteria as we considered only HCWs in direct contact with patients and excluded practices with only a light component of physical exercise.

Finally, the heterogeneity in measurement methods and statistical analyses, which we have extensively covered in the Results section, made a meta-analysis inappropriate and precluded the determination of any pooled effect size.

Conclusions

Our comprehensive overview of studies exploring the association between physical activity and burnout in HCWs revealed a significant level of heterogeneity in definitions, measurements, and analyses adopted in the literature. Our work aimed to advance effective public health practices by addressing this critical issue in the existing evidence. It is important to adopt a clear definition of burnout and physical activity and make thoughtful choices regarding measurement tools and methodologies for data analysis. This becomes particularly crucial when considering that burnout is not a diagnosable disease but rather a multifaceted psychological syndrome that emerges in response to chronic interpersonal stressors in the workplace.

Our findings strongly emphasize the beneficial connection between physical activity and burnout when a statistically significant association is present in the analyses. However, they also highlight the importance of a more in-depth investigation of the specific dependencies on exercise type, intensity, duration, and frequency, knowledge that currently represents a research gap in the field of burnout studies. Moreover, our considerations regarding the measurement of burnout and the comprehensive list of associated factors have the potential to enhance the quality of future studies. Our findings have significant implications for policy makers and health care professionals, underlining the importance of promoting physical activity as an easily accessible mitigation strategy for the well-being of the workforce and the overall effectiveness of the health care system.