Improvements in Mindfulness Facets Mediate the Alleviation of Burnout Dimensions

The current study was conducted as a part of a randomized controlled trial (RCT) titled The Effectiveness of Mindfulness Practices in the Recovery of Burnout, funded by the Finnish Social Insurance Institution and registered under ClinicalTrials.​gov (NCT number: NCT01920230). The research protocol was approved by the Research Ethics Committee of the Central Finland Health Care District. As part of the ethical evaluation, a report on the data processing and storage was approved. Data were stored anonymously in the university data storage. The results of the RCT are presented in Puolakanaho et al. (2020). Participants were recruited using web page announcements and newspaper advertisements and via occupational health care units. Enrollment took place via a web questionnaire and was open to anyone interested in participating in the study. The selection was based on information provided by the applicants via the enrollment questionnaires and subsequent interviews. The inclusion criteria were the age of 25–60 years, ongoing work, daily available Internet connection, and high burnout according to the cutoff score of the Bergen Burnout Indicator (BBI). The BBI cutoff was the 75th percentile (39–47 points) of the age group, according to the manual by Näätänen et al. (2003). Candidates were excluded if they had serious psychological or somatic disorders, were in regular psychotherapy, or were susceptible to major pharmaceutically induced mood swings. The participants gave their informed consent and received the intervention free of charge. They were emailed web questionnaires before the intervention (pre), after the intervention (post), and 10 months after the post-measurement (fup10). The pre-measurement was completed within 2 weeks before the intervention started. Reminders were sent via email and telephone.

The participants were paired based on sex, age, and education. Then, either participant in each pair was assigned to an intervention group receiving an intervention in addition to treatment-as-usual (TAU; n = 133; 12 separate groups), and the other was assigned to a control group receiving only TAU (n = 133). Most of the participants (82%; n = 109 for both groups) were blindly randomized to the groups. The remaining participants (18%, n = 24 for both groups) were matched without randomization using the same criteria (sex, age, and education). This was done to allow for a sophisticated statistical analysis of the associations between the mindfulness facets and the burnout dimensions, given that a larger sample size was required to fit the criteria for structural equation modeling (Wolf et al. 2013).

The same inclusion criteria and matching procedure were applied for both the randomized and non-randomized participants to mitigate the risk of confounding variables affecting the results of the study. A pilot study with 24 intervention participants was conducted before the RCT, and a matching procedure was completed separately to obtain corresponding control participants for the pilots. These controls were collected from the participants who initially enrolled in the study and fulfilled the inclusion criteria but were not able to participate in the face-to-face meetings and were thus not randomized. The non-randomized intervention participants underwent the same intervention procedure as the randomized intervention participants, and the non-randomized control participants underwent the same procedure as the randomized control participants.

No significant differences were found between the randomized and non-randomized controls in terms of sex, age, education, and the main study variables (five mindfulness facets and three burnout dimensions) at the enrollment, pre-, post-, or fup10 measurements, as shown by an independent samples t test. The same was observed for differences between the randomized and non-randomized intervention participants, except that the latter experienced less exhaustion at the enrollment and had lower reduced professional efficacy at the pre-measurement than the former. The exact results of all the t tests in the methods, with the means and standard deviations for the compared groups, are reported in the supplementary material (Appendix A).

The sample that was used in the analyses consisted of 106 intervention group participants and 96 control group participants, yielding a total sample of 202 participants. All of them completed both the pre- and post-measurements. By the 10th month after the post-measurement, 23 participants had dropped out of the study. There were no significant differences between those who remained in the study and those who dropped out in any of the demographic or main study variables, according to an independent samples t test. A diagram of the participant flow is presented in the supplementary material (Appendix A).

The participants of the final sample were all Caucasian, and 80% of them were women. The mean age of the participants was 47.5 years (SD = 8.05, a range of 27–60 years), and 67% of them had a polytechnic or university degree. Of the participants, 30% had vocational education, and 3% had participated in other forms of education (e.g., short employment courses). They worked approximately 40 h per week (SD = 9.55). Of the participants, 75% were married or cohabiting, 13% were divorced or widowed, and 11% were single. Ten percent evaluated their economic situation as very good, 57% rated it as rather good, 29% described it as rather tight, and 4% considered it very tight.

Burnout was measured using the 15-item BBI (Näätänen et al. 2003). The BBI-15 has a subscale for each burnout dimension: exhaustion (five items, e.g., “I am snowed under with work”), cynicism (five items, e.g., “I feel dispirited at my work and I think of leaving my job”), and reduced professional efficacy (five items, e.g., “I frequently question the value of my work”). The 6-point response scale for this questionnaire ranges from 1 (completely disagree) to 6 (completely agree). This measure was chosen since it contains cutoff scores for Finnish samples and could thus be used to include only workers with the highest amount of burnout symptoms. The reliability and validity of this measure have been found to be relatively good by previous studies (Salmela-Aro et al. 2011; Näätänen et al. 2003).

The mindfulness facets were measured using the FFMQ (Baer et al. 2006). This questionnaire consists of 39 items measuring five mindfulness facets: observing (eight items, e.g., “When I’m walking, I deliberately notice the sensations of my body moving”), describing (eight items, e.g., “I’m good at finding words to describe my feelings”), acting with awareness (eight items, e.g., “When I do things, my mind wanders off and I’m easily distracted”, reverse-scored), non-judging (eight items, e.g., “I criticize myself for having irrational or inappropriate emotions”, reverse-scored), and non-reacting (seven items, e.g., “I perceive my feelings and emotions without having to react to them”). The 5-point response scale for this questionnaire ranges from 1 (never or very rarely true) to 5 (very often or always true). This measure was chosen since it has separate subscales for different facets. In previous studies, the reliability and validity of this measure have been relatively good (Baer et al. 2006, 2008). Cronbach’s alphas from pre- to fup10 measurements for burnout and mindfulness are presented in Table 1.

Descriptive statistics (means, standard deviations, correlations, and reliability calculations) were computed with SPSS Statistics 24, and other analyses were completed with Mplus 8 (Muthén and Muthén 2017). Latent change score modeling (McArdle and Hamagami 2001) with a measurement model was conducted for each combination of the mindfulness facets and burnout dimensions, yielding 15 independent models. The used model was the same for each combination and is presented in Fig. 1. Reliability statistics (Cronbach’s alphas) were calculated for the mindfulness and burnout scales, and these results showed good or excellent values (α = 0.81–0.95) for all mindfulness facets at all time points and for all burnout dimensions at the post- and fup10 measurements. At the pre-measurement, the alpha values were adequate for exhaustion (α = 0.75) and cynicism (α = 0.79) but questionable for reduced professional efficacy (α = 0.65). The measurement model was used to further eliminate the effect of error variance on the study constructs and to ensure the reliability of the constructs in the final model. To improve the variable-to-sample size ratio, construct-specific parcels were created according to recommendations from Little et al. (2002). Previously identified structures of the five mindfulness facets (Baer et al. 2006) and the three burnout dimensions (Näätänen et al. 2003) were utilized as bases for parcel creation. Confirmatory factor analyses (CFAs) were conducted at pre-, post-, and fup10 measurements for each construct to validate the use of these existing structural definitions for the mindfulness facets and burnout dimension. Since the CFA models fitted the data reasonably well and the factor loadings of the individual items were approximately the same size for the given factors, the parcels were created by combining the items in the order they are presented in the original questionnaire. This way, the individual items of each mindfulness facet were divided into three parcels, as were the items of each burnout dimension. In the parcels, an individual level that was stable over time and was unassociated with any other part of the model was detected. This was considered by adding a level correction to the model for the parcels. The scalar equivalence (e.g., equal factor loadings and equal intercepts of observed variables) was expected to hold across time.

Latent change score modeling combines features from cross-lagged regression modeling and latent growth curves (McArdle 2009; McArdle and Hamagami 2001). In latent change score modeling, variable Y is described at a time t, and ΔY_t is defined as the change in Y from t–1 to t (McArdle 2009). In the current study, the change scores were calculated for the factors instead of observed variables; hence, Y here referred to a factor. The coefficients relating to Y_t and Y_t-1 were constrained to 1, and there were no error terms in the equation for Y_t. Y_t was a direct sum of Y_t-1 and ΔY_t. This way, ΔY_t could be used as a latent variable that directly indicated the amount of change in the target variable between given time points. Latent change scores were created for the changes from pre- to post-measurement and from post- to fup10 measurement in the mindfulness facets and burnout dimensions (marked D in the figure). Modeling was conducted, adjusting the effect of non-normality with a robust full information maximum likelihood estimator. A few outliers represented real observations of the participants with the different intervention effects; thus, they were included in the models. The possible effects of these outlier observations on the study results were evaluated by comparing the results after the exclusion of problematic observations. These examinations showed no significant changes to the results. Standardized model results were reported, from which the magnitudes of the effects were directly observable without further effect size calculations.

The fit of the models was evaluated using the chi-square test (χ²), comparative fit index (CFI), Tucker-Lewis index (TLI), standardized root mean square residual (SRMR), and root mean square error of approximation (RMSEA). As recommended by Hu and Bentler (1999), a good model fit was assumed when the CFI and TLI were close to 0.95, the SRMR was close to 0.08, and the RMSEA was close to 0.06.

As shown in Fig. 1 (relevant pathways are bolded), β₂, β₉, and β₁₀ were the most relevant coefficients to the research questions about the mediation effects. β₂ showed whether the intervention affected the change from pre- to post-measurement in each mindfulness facet. β₉ and β₁₀ showed whether the change in a mindfulness facet from pre- to post-measurement predicted the change in a burnout dimension from pre- to post-measurement (β₉) and from post- to fup10 measurement (β₁₀). Furthermore, the pathways β₂ × β₉ and β₂ × β₁₀ showed whether the effects of the intervention on the burnout dimensions were mediated by the changes in the mindfulness facets (indirect effects). β₂ × β₉ showed how the change in the mindfulness facets during the intervention was connected to the change in the burnout dimensions during the intervention. Thus, mindfulness and burnout were measured at the same time points. In addition, β₂ × β₁₀ showed how the change in the mindfulness facets during the intervention was connected to the change in the burnout dimensions during the follow-up. Thus, mindfulness change was measured before burnout change. β₃ showed whether the intervention had an additional direct effect on the change from pre- to post-measurement in the burnout dimensions after the mediated indirect effect was considered. A 95% confidence interval for the coefficients (estimate ± 2 standard errors) was calculated to evaluate the differences between the burnout dimensions with regard to the significant predictors of change. Overlapping of the confidence intervals for the equivalent coefficients (β₂ × β₉, β₂ × β₁₀, β₂, β₉, and β₁₀) of the mindfulness facets in the case of each burnout dimension indicated that there was no difference in the significance of the predictors of change in the group-level analysis. A lack of overlap meant that the difference was significant.

Tables 3, 4, 5, 6, and 7 show the β values and their confidence intervals for the models. The tables were grouped based on the mindfulness facets such that each table presented values for all three burnout dimensions when one of the mindfulness facets was considered. The description of the results was based on the values and significances of the coefficients, namely, β₂, β₃, β₉, β₁₀, β₂ × β₉, and β₂ × β₁₀. The intervention was effective in alleviating all burnout dimensions. All mindfulness facets improved during the intervention compared with the skills acquisition in the control group (β₂ was significant with all combinations of mindfulness facets and burnout dimensions). There were both similarities and differences between the burnout dimensions in which mindfulness facets were mediators of change (based on the significances of the coefficients β₂, β₉, β₁₀, β₂ × β₉, and β₂ × β₁₀ and differences thereof).

The current study indicated that it could be essential to develop several mindfulness facets simultaneously to induce a significant change in burnout dimensions, although, in previous cross-sectional research, certain mindfulness facets have been more strongly associated with burnout (Kriakous et al. 2019; Taylor and Millear 2016). A large spread of the mindfulness facets mediated the change in burnout dimensions during the intervention, and the differences between the dimensions were minor. The importance of several mindfulness facets for short-term burnout change can be understood in the context of burnout literature. Burnout is a persistent, job-related state of ill-being that has wide-ranging effects for the functioning of the employee (Leiter et al. 2014; Morse et al. 2012). Besides, Warr (2012) characterized burnout as a cognitive-affective disorder that comprises thoughts, feelings, and memories and encompasses a wide range of experiences in daily living. Because of the all-encompassing effects of burnout, likely, a profound change in a way a person views their situation is required to alleviate burnout. Hence, simultaneous improvement in several mindfulness facets could be needed to achieve such a change. Improvement in observing and describing skills could help a person identify the factors associated with burnout, whereas improvement in acting with awareness could help the person take action to affect these factors in a way that the possibility for recovery increases. Furthermore, better non-judging and non-reacting skills could help one adopt an accepting and defused stance towards even difficult experiences and focus their resources on recovery.

Some mindfulness facets even mediated the additional reduction of burnout during the 10-month follow-up. For a long-term change in burnout dimensions, the skills of non-judging (for all dimensions), describing (for cynicism and reduced professional efficacy), and observing (for reduced professional efficacy) appeared to be more relevant than the others. If several mindfulness facets are needed at the beginning of burnout recovery to induce profound change, the unique relevance of some mindfulness facets over the others could be more visible after a basic level of all skills is achieved. The significance of non-judging, describing, and observing for further alleviation of burnout can be understood in the context of the cognitive-affective description of burnout (Warr 2012). Exhaustion could be considered a more affective component of burnout since it refers to the feelings of emotional and physical strain after work (Näätänen et al. 2003). When a non-judgmental way to evaluate experiences increases, a person could experience more self-compassion and be less demanding when working. These changes could help to reduce the experienced strain and thus alleviate exhaustion. In turn, cynicism and professional efficacy could be described as more cognitive components of burnout, given that they involve interpretations of oneself, others, and situations (Näätänen et al. 2003). A new way to observe and describe experiences could be needed for one to continuously develop a non-judging stance towards their conceptualizations of themselves and others. On the other hand, non-judging skills could help one benefit from better observing and describing skills and change the interpretations of themselves and situations in a way that cynicism and reduced professional efficacy alleviate. The importance of joint development of observation-related facets (observing and describing) and reaction-related facets (non-judging and non-reacting) have also been noticed with depression (Heeren et al. 2015; Querstret et al. 2018).

Only non-judging (measured here with reverse-scored items, e.g., “I tell myself I shouldn’t be feeling the way I’m feeling.” and “I make judgments about whether my thoughts are good or bad.”) systematically mediated the changes in all burnout dimensions during both the intervention and the 10-month follow-up. Hence, improvement in non-judging skills could be the key to a long-term change in one’s way of seeing themselves and adjusting their behavior to support burnout alleviation. There is often a discrepancy between the personal standards and perceived performance of people developing burnout symptoms (Ozbilir et al. 2015). Furthermore, low self-rated professional efficacy could lead to a situation where one refuses to rest and works even longer hours to compensate for one’s perceived shortcomings. The improvement of non-judging could help one process their expectations and develop a less guilt-inducing attitude towards themselves. When non-judging skills are improved, it could be easier to recognize the over-demanding expectations and change one’s behavior accordingly. Better non-judging skills could also help to adopt a more self-compassionate way to act with these expectations. One could acknowledge the feelings of exhaustion and give space for recovery in daily living. Non-judging stance towards experiences could also prevent automatic negative evaluations of job conditions and thereby alleviate cynicism. In line with the present findings, high non-judging has been linked to lower levels of exhaustion and cynicism in cross-sectional research (Taylor and Millear 2016).

Interestingly, in the current study, the role of acting with awareness was not as pronounced as in cross-sectional burnout studies (Kriakous et al. 2019; Yang et al. 2017). Compared with the present sample that included only those with high initial burnout, the participants in the previous studies experienced varying levels of burnout symptoms (Kriakous et al. 2019; Yang et al. 2017). Acting with awareness could be an important skill to mitigate the risk factors for developing severe burnout. Furthermore, when burnout symptoms are mild, there could be more resources for acting with conscious intention to improve one’s well-being. However, when exhaustion has depleted resources of the employee, and when the evaluations of one’s capabilities to handle job demands have become more negative, taking conscious action to change the situation could be overwhelming. Thus, when severe burnout has developed, improving non-judging skills could be more beneficial in helping the person give time for rest and observe themselves and the situation with less judgment and negativity. The current findings indicated that with a MAV intervention, non-judging skills could be improved to a level that yields positive long-term effects on burnout. This result is encouraging for the further study of cost-effective ways to increase relevant skills for burnout alleviation.

More research is needed to generate conclusions on the associations between mindfulness facets and burnout dimensions. In this study, mindfulness and burnout were investigated with self-report measures, which raised the possibility of common method bias. Nevertheless, these phenomena are best evaluated by using self-reports, while the use of longitudinal data diminishes the likelihood of common method bias (Doty and Glick 1998). The differences between burnout dimensions regarding essential mindfulness facets were evaluated based on confidence intervals, which showed maximums and minimums for the mediation effects on the present sample. With a larger sample size, the non-significant mediators could become significant. Before the post-measurement, people experiencing more exhaustion and cynicism were more likely to drop out, which raised the question of whether the intervention asked too much of the participants. However, all the participants had high initial burnout, as evaluated by the BBI (Näätänen et al. 2003). Most completed the intervention, and the dropout rate during the follow-up was comparable to that of Roeser et al. (2013) with shorter follow-up. The generalizability of the results is restricted because the sample mainly comprised highly educated women. This group could have been more motivated to participate in an intervention that contained a substantial amount of homework and group participation. However, the investigated intervention was not an exception in the requirement of homework and participation (Luken and Sammons 2016). More research is needed on whether interventions similar to that investigated in the current study are viable, for example, for less educated or male working populations. Further study is needed to determine whether the present mediation results could be replicated in other interventions using mindfulness. Furthermore, other processes affecting MAV intervention results should be studied besides mindfulness processes since the present intervention had an additional direct effect on exhaustion, after controlling for the mediators.