Edited by: Renato Pisanti, Niccolò Cusano University Rome, Italy
Reviewed by: Debra Nelson, Oklahoma State University, USA; Margot Van Der Doef, Leiden University, Netherlands
*Correspondence: Charles C. Benight
Roman Cieslak
This article was submitted to Organizational Psychology, a section of the journal Frontiers in Psychology
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
This longitudinal research examined the relationship direction between burnout components (exhaustion and disengagement) within the context of personal resources measured by self-efficacy and social support. In line with the conservation of resources theory we hypothesized that exhaustion may trigger a spiral loss of personal resources where self-efficacy declines and subsequently, social support also declines and in turn predict disengagement. Participants in Study 1 were mental healthcare providers (
Job burnout is recognized as one of the key consequences of job stress (Maslach et al.,
Traditionally, burnout has been conceptualized as a prolonged response to chronic emotional and interpersonal stressors that occur in the work setting (Maslach et al.,
Across several models of burnout, exhaustion is one of its key facets. For example Melamed et al. (
Yet another prominent model of burnout (Demerouti et al.,
Dropping the personal accomplishment component is in line with theoretical developments (cf. Demerouti et al.,
Understanding the interplay between burnout components is also critically important to consider. Three different models have been proposed. First, Demerouti et al. (
Collectively these models proposed different directions for the relationships between burnout components. All three models suggested that exhaustion is a response to work stress (Leiter and Maslach,
There are a limited number of longitudinal studies investigating the direction of associations among job burnout components, yet the findings are relatively consistent. In support of the process model (Leiter and Maslach,
Self-efficacy and social support are among the most frequently examined resources that play important roles in understanding the development of work stress consequences such as burnout (Cordes and Dougherty,
Theoretical models explaining burnout consistently propose that control beliefs (including self-efficacy) and social support constitute critical resources that are important to consider (cf. job demands-control-support [DCS] model, Karasek and Theorell,
COR theory (Hobfoll,
We found one longitudinal study confirming that emotional exhaustion has an effect on self-efficacy and that self-efficacy may mediate the relationship between exhaustion and other burnout components (Brouwers and Tomic,
As suggested by Hobfoll (
Our studies investigated the associations between two components of burnout, exhaustion and disengagement within the context of personal resources. We investigated the importance of changes in two primary personal resources, burnout self-efficacy and work related social support. The associations were tested in two longitudinal studies conducted among human services workers working in the U.S. and Poland with military and civilian clients. Specifically, it was hypothesized that exhaustion at Time 1 would predict disengagement at Time 2. Second, we hypothesized that the exhaustion—disengagement association would be sequentially mediated by changes in self-efficacy and changes in social support. These mediating effects were tested after controlling for Time 1 disengagement. The hypotheses were tested controlling for years of work experience. This variable is one of the key determinants of burnout (Brewer and Shapard,
Study 1 was a part of a larger study investigating secondary traumatic stress and job burnout among behavioral healthcare providers for U.S. military personnel. Inclusion criteria for this study included (a) working as a behavioral healthcare provider at least one year, (b) providing services for U.S. military personnel, and (c) being indirectly exposed to trauma through their work. Two hundred and ninety four participants (mean age = 48.87 years old [
Mean age in years ( |
48.87 (12.76) | 50.62 (12.58) | 35.32 (8.48) | 34.97 (8.06) |
Female | 66.3% (195) | 71.1% (96) | 76.1% (233) | 79.3% (153) |
Male | 33.7% (99) | 28.9% (39) | 22.9% (70) | 19.2% (37) |
In a long-term relationship | 76.2% (224) | 72.6% (98) | 73.9% (226) | 77.7% (150) |
Not in a long term relationship | 21.4% (63) | 25.2% (34) | 25.5% (78) | 21.8% (37) |
High school | 0.3% (1) | 0 (0%) | 20.6% (63) | 18.1% (35) |
Associate's degree | 0.3% (1) | 0 (0%) | − | − |
Bachelor's degree | 2.0% (6) | 1.5% (2) | 21.6% (66) | 20.2% (39) |
Master's degree | 45.2% (133) | 51.1% (69) | 56.6% (172) | 60.1% (116) |
Doctorate degree | 52.0% (153) | 47.4% (64) | 1.0% (3) | 0.5% (1) |
116 CP (39.5%) | 50 CP (37.0%) | 148 HCP (48.4%) | 89 HCP (46.1%) | |
74 counselors (25.2%) | 39 counselors (28.9%) | 115 SW (37.6%) | 78 SW (40.4%) | |
57 SW (19.4%) | 28 SW (20.7%) | 38 others (12.4%) | 23 others (11.9%) | |
28 HCP (9.5%) | 9 HCP (6.7%) |
Participants completed questionnaires assessing job burnout, self-efficacy for job burnout, social support, and demographics.
Oldenburg Burnout Inventory (OLBI; Halbesleben and Demerouti,
An 11-item Burnout Self-Efficacy Scale was applied to measure self-efficacy for dealing with job burnout (Lua,
The Multidimensional Scale of Perceived Social Support (MSPSS; Zimet et al.,
We collected demographic information regarding the number of years of work experience, participants' age, gender, level of education, relationship status, occupation, and experiences with indirect exposure to traumatic events through their work.
The Institutional Review Board at the authors' institution in the U.S. approved this study. The details of the procedures were described elsewhere (Cieslak et al.,
We used the maximum likelihood estimation method to impute missing data for 135 completers using IBM SPSS Amos (version 22). Measurement items for burnout, change in social support and self-efficacy were included in the full information maximum likelihood imputation. The assumption of this approach to data imputation is that the missing data must be missing at random (MAR). To assess MAR, Little's missing completely at random (MCAR) tests, which is more restrictive than MAR, were conducted in IBM SPSS (version 22) using gender, profession, and intimate relationship status as references. Results of the Little's MCAR tests showed missing data were MCAR for items for MSPSS at T1,
We tested the cultivation hypothesis in the sequential mediation analysis using Mplus (see Figure
Table
1. Disengagement T1 | 0.74 |
0.67 |
0.57 |
−0.61 |
−0.58 |
−0.19 |
−0.18 |
0.01 | 2.35 (0.70) | 2.71 (0.64) | 4.92 | 0.53 | 0.22–0.51 | |
2. Disengagement T2 | 0.77 |
0.49 |
0.66 |
−0.45 |
−0.58 |
−0.12 | −0.19 |
−0.05 | 2.40 (0.75) | 2.77 (0.65) | 3.64 | 0.53 | 0.13–0.43 | |
3. Exhaustion T1 | 0.80 |
0.67 |
0.69 |
−0.69 |
−0.61 |
−0.22 |
−0.15 |
0.08 | 2.54 (0.70) | 2.82 (0.69) | 4.72 | 0.40 | 0.21–0.52 | |
4. Exhaustion T2 | 0.64 |
0.76 |
0.77 |
−0.57 |
−0.60 |
−0.21 |
−0.22 |
0.06 | 2.53 (0.76) | 2.81 (0.60) | 3.51 | 0.41 | 0.13–0.42 | |
5. Self-efficacy T1 | −0.58 |
−0.35 |
−0.52 |
−0.38 |
0.70 |
0.23 |
0.22 |
−0.04 | 5.89 (0.89) | 5.24 (0.98) | 6.26 | 0.69 | 0.44–0.86 | |
6. Self-efficacy T2 | −0.55 |
−0.62 |
−0.52 |
−0.61 |
0.56 |
0.23 |
0.24 |
−0.05 | 5.94 (0.81) | 5.22 (0.95) | 7.31 | 0.82 | 0.52–0.91 | |
7. Social support T1 | −0.32 |
−0.31 |
−0.29 |
−0.23 |
0.21 |
0.17 | 0.31 |
0.03 | 5.80 (1.08) | 4.93 (1.58) | 5.99 | 0.64 | 0.57–1.18 | |
8. Social support T2 | −0.29 |
−0.34 |
−0.29 |
−0.30 |
0.18 |
0.1 |
0.82 |
−0.06 | 5.73 (1.16) | 5.12 (1.36) | 4.37 | 0.48 | 0.33–0.89 | |
9. Work experience T1 | −0.30 |
−0.27 |
−0.19 |
−0.23 |
0.34 |
0.22 |
0.03 | −0.01 | 17.07 (11.36) | 12.92 (9.18) | 3.52 | 0.40 | 1.92–6.39 |
Repeated measures analysis of variance indicated that burnout self-efficacy did not change significantly from T1 to T2,
The examination of the hypothesized model assuming a sequential mediation effect of self-efficacy change and social support change in the relationship between exhaustion at T1 and disengagement at T2 showed that the model had adequate model-data fit, RMSEA = 0.032 (90% CI [0.000, 0.152]), CFI = 0.997, Tucker-Lewis Index (TLI; Tucker and Lewis,
Additionally, we tested a nested model with nonsignificant pathways (from self-efficacy change to social support change, from T1 exhaustion to social support change, and from social support change to T2 disengagement) constrained to zero. Results showed that the nested model was not significantly different from the hypothesized model,
The results of Study 1 did not support the cultivation hypothesis among behavioral healthcare providers working for U.S. military personnel. However, we found an indirect effect of a decline in self-efficacy in the relationship between exhaustion at T1 and disengagement at T2. In Study 2, to replicate these findings, the same model was tested among Polish professionals working with people suffering from an exposure to traumatic events.
Study 2 was a part of larger study examining work-related resources and demands among human services professionals who were indirectly exposed to traumatic events through their work. Inclusion criteria for this study were (a) working at least 1 year as a healthcare or social service provider, (b) providing services for civilians who were exposed to traumatic events, and (c) indirectly experiencing traumatic events through their work. Three hundred and six participants (mean age = 35.32 [
We used the Polish version of the same measurements as in Study 1 to assess burnout, self-efficacy changes, social support changes, and demographics. Back translation was used to establish accurate translation from English to Polish. Cronbach's alpha coefficients were 0.80 for disengagement at T1,0.81 for disengagement at T1,0.83 for exhaustion at T1,0.78 for exhaustion at T2,0.91 for self-efficacy at T1 and T2, and.96 for MSPSS at T1 and T2. As in Study 1, residuals between self-efficacy at T1 and T2 and residuals between social support at T1 and T2 were used as the change indices.
The Internal Review Board at the authors' institution in Poland approved this study. The details of the procedures were described elsewhere (Cieslak et al.,
The same analytical procedures and software were used as in Study 1 on 193 completers of the study. The Little's MCAR tests showed that missing data were MCAR for items for OLBI at T1,
Table
Repeated measures analysis of variance showed that burnout self-efficacy did not change from T1 to T2,
Results of the sequential mediation analysis of self-efficacy change and social support change in the relationship between exhaustion at T1 and disengagement at T2 showed that the model had adequate fit, RMSEA = 0.080 (90% CI [0.000, 0.162]), CFI = 0.980, TLI = 0.919, SRMR = 0.034. Significance levels of all coefficients were consistent with results of 95% bootstrap confidence intervals. Figure
As in Study 1, a nested model with constraints and the hypothesized model with no constraints were compared. In the nested model, three pathways were constrained to zero: from self-efficacy change to social support change, from social support change to T2 disengagement, and from T1 exhaustion to social support change. Results indicated that the nested model was not significantly different from the hypothesized model,
The invariance of the findings across the two studies was tested using a two-group model (see Table
Two-Model Group Model 1 | Hypothesized model | 10.03 | 1.67 | 0.984 | − | − |
Two-Model Group Model 2 | Significant pathways constrained to be equal | 16.75 | 1.86 | 0.974 | 6.71 | 0.011 |
Two-Model Group Model 3 | Covariances constrained to be equal | 33.54 | 2.80 | .947 | 23.51 |
0.037 |
Two-Model Group Model 4 | Residuals constrained to be equal | 10.14 | 1.27 | .984 | 0.11 | 0.000 |
Two-Model Group Model 5 | Significant pathways and residuals constrained to be equal | 16.88 | 1.53 | 0.974 | 6.85 | 0.011 |
Additional analyses aimed at testing invariance of the nested models across subsamples of men and women were conducted. The hypothesized two-group model without constraints was compared with the two-group nested models with constraints, assuming equal effects for both genders. The nested model developed for the test of invariance between Study 1 and Study 2 (cf. Two-Group Model 1) tested the invariance among men and women.
The two-group model with path coefficients constrained to be equal in men and women was not significantly different from the two-group model without constraints, Δχ2 = 6.17,
The comparisons conducted for data obtained in Studies 1 and 2 indicated that there were significant differences in the mean levels of the study variables (see Table
The results obtained in Study 2 were consistent with the Study 1 findings. Specifically, high levels of exhaustion at T1 led to a larger decline in self-efficacy, which in turn resulted in a higher level of disengagement at T2. Furthermore, the two-group model analyses indicated that the associations between the key investigated variables were similar across Study 1 and Study 2.
The findings obtained in two samples collected in different cultures provide novel evidence for the direction of the relationship between exhaustion and disengagement in the context of change in personal resources. Both samples demonstrate that exhaustion predicted disengagement approximately 6-months later. Additionally, the effects of exhaustion on disengagement were mediated by an index of change in self-efficacy beliefs where higher exhaustion led to a larger decline in self-efficacy across 6 months, which in turn resulted in higher disengagement levels.
The present study confirms the assumptions formulated in the process models advocated by Leiter and Maslach (
Our investigation attempted to test for the underlying mediating mechanisms which may explain why exhaustion predicts disengagement. Therefore, it goes beyond previous theoretical and empirical approaches that assumed and tested the direct effects of exhaustion on disengagement (Leiter and Maslach,
The effects of exhaustion on disengagement may be further explained by changes in other mediating mechanisms (e.g., personal growth) triggered by stressful events. So far, the mediating roles of evaluations of personal change (or self-evaluations other than self-efficacy) have been addressed in the context of indirect exposure to traumatic material at work (i.e., via traumatized client; cf. Shoji et al.,
In contrast to the “resource caravan” hypothesis (Hobfoll,
A lack of effect of social support on disengagement may result from the fact that this variable operates indirectly, via other resources. For example, social support may directly affect perceived personal growth (Shoji et al.,
Burnout and personal resources are relatively stable. Longitudinal research conducted over periods ranging from 4 months to 7 years indicated that approximately one-third of variance of burnout and about a half of variance of resources may be stable over years (for overview see Seppälä et al.,
The present study has its limitations. Our approach to burnout focuses on its two dimensions, which are included into some but not all burnout models (cf., Melamed et al.,
Although, both of our studies were longitudinal, there were only two measurement points. A four-wave investigation would be optimal to test a sequential multiple mediation model with two mediators and we plan to conduct this type of investigation next. Regarding a methodological limitation related to a longitudinal design, the research procedures did not allow us to explain reasons for dropouts at T2. Relatively high attrition rates limit the generalizability of the findings. Although for a majority of variables we found no systematic dropout patterns, we observed trends indicating a systematic character of dropout for two variables in Study 1 and one variable in Study 2. In Study 1, participants with high disengagement (T1) were lost at the follow-up. Therefore, the findings of Study 1 may better reflect the effects observed for those whose burnout was lower at T1. Importantly, the findings of Study 1 and Study 2 revealed similar patterns of associations, and there was no systematic dropout for burnout indicators in Study 2.
Another limitation refers to the choice of self-efficacy measure. Although our findings suggested that the scale had good reliability and shared less than 38% of variance with other constructs, confirming its discriminant validity, future research testing the validity of the burnout self-efficacy scale are needed.
The study is also limited in that we tested only one direction from exhaustion to disengagement, which is in line with previous findings (e.g., Toppinen-Tanner et al.,
Finally, in line with earlier findings (e.g., Shoji et al.,
In sum, this is the first longitudinal two-study cross-cultural investigation on how changes in personal resources mediate between exhaustion and disengagement, measured 6 months apart. Both studies consistently indicate that reductions in job burnout self-efficacy were determined by exhaustion and facilitated greater disengagement. Future research that includes the intersection of personal resources and environmental factors in untangling the negative components of burnout will help move this literature forward informing critical interventions. In particular, the findings may have some implications for prevention of the escalation of burnout. Interventions aiming at a reduction of negative consequences of work stress may target workers with higher levels of exhaustion and work to enhance their self-efficacy beliefs specifically related to the negative consequences of work stress.
CB, RC served as the P.I.'s on the project providing significant conceptual and design contribution and significantly contributing to drafting the manuscript. CB, RC, AR, CY were involved in data acquisition. KS, RC, AL, were involved in the statistical analyses for the projects. CB, RC, KS, AR, CY, AL AK were involved in manuscript preparation and final approval of the paper. CB, RC, KS, AR, CY, AL, AK agree to be accountable for all aspects of the work specifically to responding to questions related to the accuracy or integrity of any part of the work.
This research was supported in part by a research grant to Charles Benight awarded and administered by the U.S. Army Medical Research & Materiel Command (USAMRMC) and the Telemedicine & Advanced Technology Research Center (TATRC) at Fort Detrick, MD under Contract Number W81XWH-11-2-0153 and a research grant N N106 139537 from the Polish National Science Center awarded to Roman Cieslak. The contribution of Aleksandra Luszczynska is supported by the Foundation for Polish Science, Master program.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.