Work stress and turnover intentions among hospital physicians: The mediating role of burnout and work satisfaction

Table 1 shows correlations among the study variables. As can be seen, a strong positive relationship between work stress and burnout was found (.55, p < .0001), and a strong negative association emerged between burnout and work satisfaction (?.63, p < .0001) as well as between work satisfaction and turnover intentions (?.65, p < .0001).

Evidently, all three hypotheses were corroborated. A structural equation model (by AMOS 7) was computed to simultaneously test all the relationships as predicted by the three hypotheses ( Figure 1 ). However, a direct path from work stress to work satisfaction and another from burnout to turnover intentions were also incorporated into the model, because previous publications have borne out these links ( Moreno-Jiménez et al., 2009).

Figure 1

Research Model.

Note . The stantardized regression coefficients are depicted on the paths, while the correlations coefficients are shown in parenthesis.

N = 124.

*p < .05, **p < .01.

Since there is no single statistical test that best describes the strength of a model's predictions ( Hair, Anderson, Tatham, & Black, 1992 ), several measures of approximation were employed. In the Normed Fit Index (NFI), the Goodness-of-Fit Index (GFI), and the Comparative Fit Index (CFI) a degree of fit above .9 is considered sufficient ( Bentler, 1990 ). The approximation measures found here were above .95, and therefore meet the approximation criteria (NFI = .991, CFI = .995). In addition, lack-of-fit was measured by means of RMSEA (root mean square error of approximation), where the value of a suitable fit should be lower than .1 ( Jöreskog & Sörbom, 1989 ). In the current study, the results for lack-of-fit were sufficient (RMSEA = .088). In addition, ? 2(1) was 1.95, p > .05.

Specifically, Figure 1 indicates that work stress relates directly and positively to burnout (? = .55), burnout relates directly and negatively to work satisfaction (? = ?.37), and work satisfaction is directly and negatively connected to turnover intentions (? = ?.53), as hypothesized. Furthermore, however, we found a negative direct relationship between work stress and work satisfaction (? = ?.45) as well as a direct positive link between burnout and turnover intentions (? = .19), thereby lending support to the conclusion that the relationships between work stress and work satisfaction as well as between burnout and turnover intentions are not fully and solely directly mediated but rather partially mediated.

We used two statistical methods in order to test the validity and stability of our findings: (1) Spearman's (1904, 1910) disattenuated correlation coefficient in order to control for measurement errors; and (2) Browne's (1975) cross-validation coefficient to control for possible artifactual effects of the sample size. The disattenuated correlation coefficients have been presented in Table 1 (in parenthesis). We can conclude that there were not substantial measurement errors in our data, because the disattenuated correlations were in accordance with the predicted values, in regard to the reliabilities and correlations of our study (see Muchinsky, 1996; Zimmerman & Williams, 1997 ). Nevertheless, there are two correlations which reached .99 (burnout E and total burnout) and .98 (burnout D and total burnout) after disattenuation, and thus can be considered measuring the same construct.

In order to view more precise results in predicting the different variables (see Figure 1 ), we calculated Browne's (1975) cross-validation coefficient (Pˆc2). For the R 2 of burnout (.30), Browne's cross-validation coefficient was .29, meaning there is a 1% discrepancy between the original R 2 and Browne's cross-validation coefficient. For the R 2 of work satisfaction (.53), Browne's cross-validation coefficient was .51, meaning there is a 2% discrepancy between the original R 2 and Browne's cross-validation coefficient. For the R 2 of work satisfaction (.44), Browne's cross-validation coefficient was .41, meaning there is a 3% discrepancy between the original R 2 and Browne's cross-validation coefficient. In sum, these discrepancies do not prove to be problematic for the current study due to their low (although progressively increasing) values.

In the present study, we sought to extend the investigation of the relationships between work stress, burnout, work satisfaction, and turnover intentions among hospital physicians.

The hypothesis (H 1) – that a positive relationship would be found between work stress and burnout – was upheld. Medical personnel face enormous mental pressures that could result in burnout ( Pines, 2011 ). In their work, physicians are d to cope with numerous demands: clinical demands of the job (for instance, decisions about patients), academic demands (for instance, learning new technologies), and administrative demands (for instance, the need to consider financial considerations when treating patients). These demands could easily result in various pressures, from clinical responsibility to an overload of tasks under time constraints, emotionally fraught contact with patients and staff, and self-confidence in the ability to do their job properly ( Agius, Blenkin, Deary, Zealley, & Wood, 1996; Bruke & Richardson, 1990 ). Burnout is caused by ongoing mental stress or by other chronic work-related pressures ( Hobfoll & Shirom, 2000; Maslach, Schaufeli, & Leiter, 2001 ). A recent study of IT specialists ( Jung, 2013 ) found respectively that each work burnout factor (i.e., emotional exhaustion and cynicism) was predicted by job stressors (i.e., role overload, role insufficiency, and role boundary).

Our second hypothesis (H 2) – that burnout would negatively associate with work satisfaction – was also confirmed. This connection was reported in previous studies (i.e., Hombrados-Mendieta & Cosano-Rivas, 2013; Wolpin, Burke, & Greenglass, 1991 ). The intensity of the relationship depends on satisfaction measures (for instance pay, promotion, the work itself, supervisor, and the organization as a whole) and the sector (private or public) ( Tsigilis, Zachopoulou, & Grammatikopoulos, 2006 ). A study conducted in Israel on 890 doctors in six areas of specialization revealed similar findings: physicians employed in hospitals are more burned-out (the physical tiredness aspect) and express less satisfaction than independent HMO doctors ( Nirel, Shirom, & Ismail, 2003 ).

Finally, the third hypothesis (H 3) – that work satisfaction would negatively relate to turnover intentions – was corroborated. Dissatisfaction and turnover intentions are types of employee withdrawal. The withdrawal is intended to protect the individual from stress and its consequences ( Keaveney & Nelson, 1993 ). The results of the present study concur with other studies that see the worker's dissatisfaction or ‘psychological withdrawal’ as the first stage that motivates the withdrawal behavior that eventually results in turnover ( Hom & Griffeth, 1991; Krausz, Koslowsky, & Eiser, 1998 ).

Beyond the direct relationships, the results indicate that satisfaction mediates the relationship between stress and turnover intentions, whereas the direct link between turnover intentions and stress is not significant. Some studies have found weak correlation between work stress and turnover intentions (i.e., Bedeian & Armenakis, 1981 ), but most of the results a significant relationship between the variables ( Goodman & Boss, 2002; Moreno-Jiménez, Gálvez-Herrer, Rodríguez-Carvajal, & Sanz-Vergel, 2012 ).

The main explanation for not finding a significant link between stress and turnover intentions, in our opinion, could be the characteristics of Israeli physicians’ work environment. The number of hospitals in Israel is limited, and doctors realize that employment conditions in all hospitals are similar. Furthermore, a considerable number of hospital physicians in Israel work in a number of employment settings. Research has shown that most physicians (84%) in the examined specialties (including cardiologists and general surgeons who provide services in hospitals) work for more than one organization ( Nirel et al., 2003 ). That is to say, knowing medical organizations well, physicians are aware that moving from one setting to another is not lucrative. As a result, physicians do not engage in pipedreams that another hospital would be better and, consequently, their turnver intentions are lower despite their dissatisfaction.

Physicians who are fed up with the difficult work conditions in hospitals could abandon the profession entirely, moving on to biotechnology, pharmaceutical companies, medical research, teaching, etc. ( Dyke, Holtzman-Schweid, Bin-Nun, & Kushnir, 2011 ), but this obviously means intending to leave the profession rather than to leave the organization.

Also, the mediation of work satisfaction between perceived work stress and turnover intentions is known in the literature (for example, Fried, Shirom, Gilboa, & Cooper, 2008 – partially; Paillé, 2011 – fully). Our findings that work stress does not directly affect turnover intentions could be explained by stress creating negative feelings manifested in reduction of the worker's satisfaction. Turover intentions reflect the individual's attempt to cope with these negative feelings (that deplete his/her resources) by means of psychological withdrawal. Support of this explanation can be found in Lazarus’ (1999) process approach to coping with stress, namely that an individual that is exposed to stress performs: (a) primary appraisal – the individual perceives and interprets the significance of the encounter with the stressor as damaging, threatening, or challenging; (b) secondary appraisal – the individual evaluates possible response options to the damage, threat, or challenge. Subsequent emotions are rooted in the secondary appraisal (for instance, anger, anxiety, hope, etc.) ( Lazarus, 1999 , pp. 94-96). As stated, individuals that suffer from constant depletion of their resources perceive the chance for change as slim. Evaluation of response options becomes limited (the employee is burned-out and depleted) and raises negative feelings that are the ingredients of dissatisfaction. Thus, the individual concludes that action must be taken on oneself (rather than changing the stressor), while adjusting and accepting the stress situation or, alternatively, withdrawing from it, as is suggested in this model. Following Locke's (1976) widely-accepted definition of work satisfaction, and applying it to the present study's variables, it can be said that after the work is evaluated as stressful (due to demands from the doctors), an emotional reaction to work events is formed (probably negative, expressed as dissatisfaction) as are work-related behavior intentions (turnover).

We also found that burnout is a factor that partially mediates the relationship between work stress and dissatisfaction. The explanation is in the fact that work stress often, although not always, directly causes dissatisfaction. Hindrance stressors ( Cavanaugh, Boswell, Roehling, & Boudreau, 2000 ) are perceived as disruptive to the individual (prevent him/her from achieving desired goals) such as organizational politics, bureaucracy, or concerns over employment security ( Gilboa, Shirom, Fried, & Cooper, 2008 ). When faced with hindrance stressors, people believe that they do not have the d resources and coping mechanisms to face these demands no matter how much effort they invest ( LePine, Podsakoff, & LePine, 2005 ). Podsakoff, LePine, and LePine (2007) reported that while hindrance stressors decrease job satisfaction and increase turnover intentions, challenge stressors increase job satisfaction and decrease turnover intentions. In a study that examined stress and burnout of police officers, Pines and Keinan (2005) found that despite the heightened stress and potential burnout, the police officers reported satisfaction with their work. In their study, burnout was found to relate more to satisfaction and turnover intentions than to work stress. Despite the stress, burnout can be low if workers feel that their job is important and useful ( Pines & Keinan, 2005 ), or if their sense of self-efficay is intact ( Friedman, 2000 ). Challenge stress, which s a sense of meaningfulness, reduces burnout and allows people to feel satisfied with their stressful work (the interest, the challenge, etc.). As a result, there are fewer turnover intentions. Respectively, hindrance stress that disrupts the worker's ability to achieve significant goals enhances burnout, which in turn leads directly to turnover intentions in order to stem depletion of resources ( Moreno-Jiménez et al., 2009 ), but also is indirectly mediated by dissatisfaction due to the presence of negative feelings.

The present research has a number of limitations.

First, all of the study variables were examined concurrently. Long-term investigation (longitudinal study), which is important when examining dynamic variables, is missing. This is especially true for burnout studies, because burnout changes over one's employment period ( Dunford, Shipp, Boss, Angemeier, & Boss, 2012 ).

It is also possible that situation variables affected the study. The data were collected around the time of the big doctors’ strike. The strike was inundated with large media campaigns about doctors’ inferior wages and employment conditions. This could make generalization of the results to periods of stability and quiet difficult.

Another possible drawback is that many specialists and physicians work in more than one setting ( Nirel et al., 2003 ). This has implications for the doctors’ sense of overload, burnout – physical fatigue and cognitive exhaustion, low satisfaction, and less commitment to the primary place of work. Stress and burnout accumulate from numerous workplaces (more work hours per day). This study did not examine the respondents’ number of various work settings, which could affect their burnout and dissatisfaction at the hospital. This should be done in future studies.

In addition, the study did not control for individual differences that might affect the level of subjective stress. For example, individuals high in NA (negative affectivity) are more likely to experience discomfort and view their life as a series of stressors even in the absence of overt stress ( Watson, 1988; Watson & Clark, 1984 ). Therefore, it seems quite possible that negative affectivity may be driving the obtained relationships among the variables.

The research variables in this study were collected from single-source data, namely self-report questionnaires filled out by the respondents. We believed that, due to the nature of the variables (subjective stress, burnout, work satisfaction, and turnover intentions), subjective reports would be the most appropriate. However, there is a risk that the correlations do not reflect a relationship between the theoretic structures that we examined, but rather stem from the fact that a t rules sytem, or schematic structure, was used to evaluate items or measures that represent separate structures ( Avolio, Yammarino, & Bass, 1991 ), so that the observed relationships among variables were artificially ‘inflated’ ( Lim & Yuen, 1998 ). As regards our study, it is possible that the respondents ‘colored’ their answers in gloomier shades because of negative feelings rooted in stress. Hence, correlations between variables could suggest the emotional state in which the answers were given rather than a link between the variables themselves.

Finally, the study sampled only 124 interviewees, a relatively small sample. This is in accordance with Salgado's (1998) research, in which he reviews and compares sample sizes across three peer-reviewed journals. In his paper, he documents an increase in sample size over the years. This may possibly point to a weakness of the current research, that is the sample size. Clearly, similar studies on a larger scale would help to support both the reliability and validity of the current findings, and better serve the goals of reaching an all-encompassing model about the relationships between the investigated variables.

However, it is imperative to note that although the sample size might seem insufficiently large, we have used two methods in order to test the stability of the findings and control for the possibility of artifactual sample size effect on them. Based both on cross-validation coefficients ( Browne, 1975 ) and on disattenuated correlation coefficients ( Spearman, 1904, 1910 ), we conclude that the sample size of the current research does not prove to have significantly impacted on the results in terms of measurement errors and strength of the findings.

Medicine is a draining profession in itself ( Kushnir & Cohen, 2006; Pines, 2011 ). Additionally, doctors’ satisfaction with internal factors (meaningfulness, interest, challenge, self-fulfillment) is much higher than their satisfaction with external factors (unrewarding work environment, hard physical conditions, inability to lead a normal family life) ( Cooke & Chitty, 2004; Landon, Reschovsk, Pham, & Blumenthal, 2006; Van Dyke, Holtzman-Schweid, Bin-Nun, & Kushnir, 2011; Zuger, 2004 ). We suggest expanding the scope of the study and adding the variable ‘satisfaction with profession’. It could be interesting to investigate whether burnout and stress affect various types of satisfaction. It is possible that among doctors, satisfaction with their profession mediates the relationship between stress and dissatisfaction with the organization.

Additionally, turnover intentions are a form of psychological withdrawal. It should be examined whether, as a result of burnout and dissatisfaction, more serious withdrawal behaviors occur (such as tardiness and absenteeism), as was found in many models ( Hom & Griffeth, 1991; Krausz et al., 1998 ), or whether among doctors – due to their complex and important profession – a different withdrawal model occurs.

Doctors may experience challenge stress in major parts of their work. We therefore recommend an in-depth examination of the research model that differentiates between hindrance and challenge stressors. We believe that challenge stressors (as opposed to hindrance stressors) reduce burnout, and thus reduce turnover intentions, both directly ( Podsakoff et al., 2007) and mediated by satisfaction.

The present research was conducted among hospital physicians. Hospitals are bureaucratic organizations, which numerous stressors that negatively affect employees: overload, lack of independence and rewards, and stressors that stem from the service receivers ( Pines, 2011 ). We have seen that there are differences rooted in the doctors’ employment base (hospital or HMO) ( Nirel et al., 2003 ). We recommend examining the research model on physicians in less bureaucratic organizations, and comparing the results.

Observing the research model and its results empasizes the importance of treating physicians’ stressors and burnout in order to prevent their dissatisfaction and turnover intentions. Subjective perception of stress and the burnout closely related to it precede the rest of the unwanted organizational results, and should therefore be addressed first. We recommend that heads of health systems treat the physicians’ objective stressors, which affect subjective perceptions of stress, which is, in turn, related to burnout. We propose two levels of treatment to reduce stress: structural and organizational.

The structural aspect should involve reduction of objective stress on physicians by changing the actual structure of health services. Hospital doctors suffer from an overload of tasks, great responsibility, and they work long hours. We believe that a considerable number of tasks that are at present performed by hospital doctors should be transferred to other medical branches. Hospital physicians would deal with life-threatening emergencies, intensive care, and complex medical procedures. Other cases, not related to infective diseases (which could cause epidemics), can be supported by community clinics and home care. This would result in the development of new professions such as nurse's assistant, who would actively perform simple, routine medical procedures that were so far performed by hospital doctors and nurses, thus freeing them for the core roles of their profession. First examples of such steps already exist and are successful. We suggest expanding them to form a general rather than specific perception change, primarily due to developing medical technologies that enable it. For instance, in the past, treatment of diabetes was performed in hospitals, whereas today family members inject sick children with insulin and supervise sugar levels by simple technological means. Other medical procedures, such as changing bandages of bedsores, could be performed by family members or caretakers after short training. The benefits of the suggested structural change could have far-reaching consequences. It could reduce hospital costs, while decreasing overload and improving service to the patients. Furthermore, it could reduce physicians’ overload by eliminating routine tasks and grinding bureaucracy, which would enable a meaningful relationship with the patient, a sense of satisfaction and significance among doctors, spare time for updating, research, and rest, less mistakes, etc. Another benefit would be opening new employment avenues and creating jobs that would relieve doctors and nurses of routine procedures. Moreover, patients would be protected from hospital-related infections, their autonomy would be preserved (home care), and their connection to the social networks that provide them with emotional support would not be severed.

The organizational facet relates to the long work hours and chronic stress that deplete the physician's resources and increase burnout. We propose that organizations increase physicians’ resources wherever possible, so as to maintain a resonable level of resources, and enable recovery from stress. For instance, opening a gym for doctors in the hospital, setting time for yoga/meditation, providing effective tools for coping with stress, augmenting psychological resources (optimism, hope, resilience, etc.) that are the basis of quality coping. A recent meta-analysis has substantiated the claim that cognitive, behavioral, and mindfulness-based approaches are effective in reducing stress in medical students and practicing physicians, and may also contribute to lower levels of burnout in physicians ( Regehr, Glancy, Pitts, & LeBlanc, 2014 ). Indeed, a controlled experiment concerning primary healthcare professionals has significantly shown that the use of mindfulness-based programs as part of continuing professional education to reduce and prevent burnout promote positive attitudes among health professionals, strengthen patient-provider relationships, and enhance well-being ( Asuero et al., 2014 ). Optimal coping with stress at work to a certain degree reduces future stress, because it is a cycle that feeds itself. Proper coping with stress at work will enable the physician to be less impatient with the patient, which in turn improves the level of service and prevents possible problems from a patient who is dissatisfied with the service that he/she received.

The authors of this article declare no conflict of interest.