Improvements in Mindfulness Facets Mediate the Alleviation of Burnout Dimensions

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Improvements in Mindfulness Facets Mediate the Alleviation of Burnout Dimensions © The Author(s) 2020 Published version Kinnunen, Sanna M.; Puolakanaho, Anne; Tolvanen, Asko; Mäkikangas, Anne; Lappalainen, Raimo Kinnunen, S. M., Puolakanaho, A., Tolvanen, A., Mäkikangas, A., & Lappalainen, R. (2020). Improvements in Mindfulness Facets Mediate the Alleviation of Burnout Dimensions. Mindfulness, 11(12), 2779-2792. https://doi.org/10.1007/s12671-020-01490-8 2020 ORIGINAL PAPER Improvements in Mindfulness Facets Mediate the Alleviation of Burnout Dimensions Sanna M. Kinnunen 1 &Anne Puolakanaho 1 &Asko Tolvanen 1 &Anne Mäkikangas 2 &Raimo Lappalainen 1 #The Author(s) 2020 Abstract Objectives While interventions using mindfulness have been effective in treating burnout, the mechanisms of change need more research. This study investigated which of five mindfulness facets (observing, describing, acting with awareness, non-judging, and non-reacting) mediated the intervention effects on three burnout dimensions (exhaustion, cynicism, and reduced professional efficacy) during an 8-week mindfulness-, acceptance-, and value-based (MAV) intervention and a 10-month follow-up. Methods The participants were a heterogeneous sample of employees suffering from burnout (n= 202, 80% women, mean age = 47.5 years). Latent change score modeling was conducted for each combination of the mindfulness facets and the burnout dimensions. Confidence intervals were calculated for the coefficients in the models. Results The modeling results showed that mindfulness improvement during the intervention mediated burnout alleviation during both the intervention and the 10-month follow-up. A large spread of mindfulness facets mediated changes in all the burnout dimensions during the intervention (all for cynicism, all except describing for exhaustion, and all except observing for reduced professional efficacy). The improvement in non-judging skills mediated the reductions in all burnout dimensions during the follow-up. For exhaustion, it was the only significant mediator during the follow-up, whereas for cynicism and reduced professional efficacy, describing and observing were additional mediators. Conclusions Improving mindfulness facets using a MAV intervention had significant long-term effects on burnout in this study. Non-judging is possibly the most important mindfulness facet to improve in burnout interventions, given that it mediated the changes in all burnout dimensions during both the intervention and 10-month follow-up. Keywords Mindfulness .Burnout .Mediation .Intervention .Acceptance and commitment therapy Burnout has been associated with increased mental and physical health problems, and negative organizational outcomes, such as increased employee turnover, sickness absences, and impaired job performance (Ahola et al. 2017; Morse et al. 2012). Burnout is a persistent, job-related state of ill-being that is characterized by exhaustion, cynicism, and reduced professional efficacy (Leiter et al. 2014;Näätänenetal. 2003). Exhaustion refers to feelings of physical and emotional fatigue, cynicism to distancing oneself from work and questioning the meaningfulness of one’s job, and reduced professional efficacy to perceiving one’s capabilities as inadequate for satisfactory job performance. Burnout has been reluctant to change without intervention (Mäkikangas and Kinnunen 2016). Therefore, it is essential to develop effective treatments to mitigate its adverse effects. It is also important to understand by which processes these interventions work. By emphasizing the intervention processes that are the most likely to alleviate burnout, intervention effectiveness could likely be improved, and resources used cost-effectively. Process-based intervention research is a way to increase process knowledge since it focuses on studying theory-based and empirically supported processes responsible for positive intervention outcomes (Hofmann and Hayes 2019). Process-based research is interested in whether the manipulation of a certain process Electronic supplementary material The online version of this article (https://doi.org/10.1007/s12671-020-01490-8) contains supplementary material, which is available to authorized users. *Sanna M. Kinnunen [email protected] 1 Faculty of Education and Psychology, University of Jyväskylä, PO Box 35, FI-40014 Jyväskylä, Finland 2 Faculty of Social Sciences (Psychology), University of Tampere, 33014 Tampere, Finland Mindfulness https://doi.org/10.1007/s12671-020-01490-8 is efficacious in reducing the targeted symptoms and whether the used intervention can manipulate the process intendedly (Hayes and Hofmann 2017; Hofmann and Hayes 2019). Hofmann and Hayes (2019) suggested that different procedures could influence the same process. Hence, diverse interventions could lead to similar outcomes. In interventions using mindfulness, it is a central, theoretically expected mechanism of change (Crane et al. 2017; Hayes et al. 2012). Fletcher and Hayes (2005) offered a functional definition of mindfulness that includes elements from several conceptualizations of mindfulness and integrates diverse mindfulness processes. The functional mindfulness definition comprises processes of staying in the present moment, taking an open and non-judgmental stance towards one’sexperiences, detaching from the guidance of one’sthoughtsand feelings, and transcending the conceptual description of oneself. In empirical research, a measure that is well-suited for studying different mindfulness processes is the Five Facet Mindfulness Questionnaire (FFMQ, Baer et al. 2006). It evaluates five mindfulness facets, namely observing, describing, acting with awareness, non-judging, and non-reacting. Observing refers to noticing inner and outer stimuli, such as sensations, thoughts, and feelings; describing is the ability to describe observed stimuli with words; acting with awareness refers to being aware of one’s situation and acting with conscious intention rather than reacting automatically; nonjudging means refraining from evaluating one’s thoughts, feelings, and sensations; and non-reactivity is the ability to let thoughts and feelings come and go without becoming attached or impulsively reacting to them. Concerning the functional definition of mindfulness by Fletcher and Hayes (2005), observing and describing depict the skills to be in contact with the present moment, while non-judging reflects the ability to take an accepting stance towards the experiences at the moment. Non-reacting and acting with awareness depict the skills to defuse from the behavioral guidance of thoughts and feelings and instead act according to one’s personal goals. Recently, interventions using mindfulness have shown promise in burnout treatment (Luken and Sammons 2016; Lloyd et al. 2013). Mindfulness has also been the mediator of burnout change in these interventions (Roeser et al. 2013). However, recent cross-sectional studies have indicated that certain mindfulness processes could be more relevant to positive intervention outcomes than others. Yang et al. (2017) found that, while all five mindfulness facets of the FFMQ had a negative association with exhaustion and disengagement (comparable with cynicism), the strongest association was found for acting with awareness. Kriakous et al. (2019) also observed that higher levels of acting with awareness were associated to lower levels of emotional exhaustion and depersonalization. Taylor and Millear (2016) separated between the burnout dimensions and found that non-judging and non-reacting were associated with emotional exhaustion, acting with awareness, and non-judging with cynicism, and observing with reduced professional efficacy. For each facet, higher mindfulness was associated with lower burnout. These results indicate that, instead of studying mindfulness as a unified mediator, researchers should study its processes as separate mechanisms of change in the alleviation of burnout. Although the research on the associations between mindfulness facets and burnout is mostly cross-sectional, there is evidence of the differential role of mindfulness facets in the changes of other well-being indicators during the intervention. Either non-judging or acting with awareness predicted positive intervention outcomes in terms of stress, anxiety, work-related rumination, and fatigue (Bergman et al. 2016; Querstret et al. 2016,2018). However, with some well-being indicators, simultaneous improvements in observation-related (observing or describing) and reaction-related (non-judging or non-reacting) facets were needed to induce positive change. This need was noticed for psychological distress, psychological symptoms, and depression (Heeren et al. 2015; Querstret et al. 2018; Waters et al. 2018). More process research is needed to understand better which combination of mindfulness facets would show the most beneficial results in burnout treatment. Furthermore, the findings of Taylor and Millear (2016) suggest that different mindfulness facets could be important for the alleviation of separate burnout dimensions, indicating that it is warranted to study the associations between mindfulness processes and burnout separately for each dimension. The current study aimed to understand better how interventions using mindfulness achieve their positive effects. This study explored whether the five facets of mindfulness, according to the FFMQ (observing, describing, acting with awareness, non-judging, and non-reacting), were mediators of burnout change during an 8-week intervention and a 10-month follow-up. Furthermore, this study investigated the differences between the burnout dimensions (exhaustion, cynicism, and reduced professional efficacy) regarding how the mindfulness facets mediated their changes. Mindfulness facets were expected to mediate the intervention effects on the burnout dimensions, but no specific hypotheses were made about the associations between the separate mindfulness facets and burnout dimensions. Method Participants 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 Mindfulness 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 premeasurement 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). Sample Attrition Initially, there were 266 participants (133 each in the intervention and control groups). Before completing the post-measurement questionnaire, 63 participants withdrew from the study. A total of 41% of these dropouts (n= 26) came from the intervention group, and 59% (n= 37) were from the control group. In addition, one participant from the intervention group was excluded from the analyses because their BBI score dropped significantly between enrollment and premeasurement (randomization was completed in the enrollment phase, when the BBI score matched the inclusion criteria). An independent samples ttest revealed no significant differences betweenthosewhodroppedout and those who remained in terms of sex, age, education, or initial level of reduced professional efficacy. However, those who dropped out experienced more exhaustion and cynicism at the enrollment phase than those who stayed. 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 preand 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 ttest. 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% Mindfulness rated it as rather good, 29% described it as rather tight, and 4% considered it very tight. Procedures Intervention The 8-week intervention combined weekly group meetings with the use of a web program. The main structure and most homework assignments were based on the mindfulness intervention by Williams and Penman (2011). Theoretically, the intervention was founded on the principles of ACT (Hayes et al. 2012). In line with this, the informatory content was formed based on ACT. Thus, the present intervention combined elements of traditional mindfulness-based intervention and ACT to increase the mindfulness and acceptance skills and promote the valuebased actions of the participants. Therefore, the program was described as mindfulness-, acceptance-, and value-based (MAV) intervention. The intervention included weekly themes related to identifying the factors contributing to one’s burnout, changing routines to support one’s well-being, and learning new ways to relate to one’s situation. The participants practiced self-compassion and a non-judging stance towards their experiences. They also practiced letting go of attempts to control their thoughts and feelings. The participants were instructed to perform formal mindfulness-acceptance practices (e.g., 10–15 min body scan or loving-kindness meditation) twice a day for 6 days a week. The formal practices also included a short breathing space that was to be done once a day. In addition, the participants were instructed to mindfully carry out informal activities, such as daily tasks, and change their routines. They were further advised to perform valuebased actions in their daily lives. Homework assignments and voluntary material were available through the web program. The intervention was standardized and delivered by two psychologists with experience and training related to the practices used. Adherence to the protocol was relatively high in terms of participation in group meetings and completion of homework. A detailed description of the contents and adherence is presented in the supplementary material (Appendixes Band C). TAU In the research project to which this study belongs, the purpose was to investigate whether adding a MAV intervention to TAU would have superior effects compared with those of only TAU in burnout alleviation. Therefore, both the intervention and control group participants could use TAU services. Furthermore, for ethical and practical reasons, the researchers did not advise the intervention participants not to use the services. The control group participants were especially encouraged to use the available services, but they were not directed to any service by the research group. Furthermore, the control group did not receive any intervention from the research group, but they were promised access to the web program after the 12-month study period was over. In Finland, several approaches are used to alleviate burnout symptoms, such as therapeutic conversations in employee health care, rehabilitation, sick leaves, medication, support from employers, and changes in job conditions. Sixty-three percent of the intervention and 62% of the control participants utilized at least one support form during the intervention. During the follow-up, 70% of the intervention and 80% of the control participants used TAU services. Details about the used services are presented in the supplementary material (Appendix C). Measures 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’mgood 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’salphasfrompre-to fup10 measurements for burnout and mindfulness are presented in Table 1. Data Analyses Descriptive statistics (means, standard deviations, correlations, and reliability calculations) were computed with SPSS Statistics 24, and other analyses were completed with Mplus 8 Mindfulness Table 1 Means, standard deviations, and correlations of the study variables (n= 177–202) VariableMSD123456789101112 1 OB1 3.15 0.67 0.81 2 OB2 3.43 0.65 0.57** 0.85 3 OB3 3.51 0.67 0.59** 0.71** 0.86 4 DE1 3.49 0.75 0.28** 0.15* 0.21** 0.92 5 DE2 3.64 0.70 0.24** 0.30** 0.32** 0.76** 0.93 6 DE3 3.75 0.76 0.26** 0.27** 0.43** 0.74** 0.79** 0.95 7 AC1 2.97 0.68 0.26** 0.08 0.12 0.17* 0.10 0.15 0.89 8 AC2 3.15 0.60 0.19** 0.30** 0.15* 0.07 0.28** 0.21** 0.53** 0.87 9 AC3 3.20 0.66 0.18* 0.22** 0.36** 0.10 0.22** 0.38** 0.54** 0.63** 0.91 10 NJ1 3.54 0.71 0.08 0.06 0.05 0.10 0.03 0.11 0.26** 0.13 0.07 0.87 11 NJ2 3.82 0.69 −0.00 0.03 0.07 0.15* 0.24** 0.25** 0.00 0.22** 0.18** 0.57** 0.89 12 NJ3 3.84 0.71 0.05 0.13 0.20** 0.17* 0.29** 0.34** 0.05 0.19* 0.29** 0.46** 0.71** 0.90 13 NR1 2.94 0.62 0.33** 0.14 0.12 0.33** 0.18** 0.16* 0.34** 0.22** 0.13 0.28** 0.17* 0.14 14 NR2 3.21 0.58 0.21** 0.35** 0.20** 0.21** 0.34** 0.27** 0.14* 0.42** 0.22** 0.11 0.33** 0.28** 15 NR3 3.32 0.63 0.23** 0.29** 0.46** 0.25** 0.34** 0.46** 0.20** 0.33** 0.50** 0.13 0.32** 0.42** 16 EX1 4.20 0.93 −0.08 0.03 0.10 −0.02 0.06 0.07 −0.18* −0.07 −0.03 −0.15* −0.06 −0.01 17 EX2 3.65 1.06 0.12 −0.01 −0.04 −0.03 −0.06 −0.13 0.02 −0.14 −0.11 −0.06 −0.20** −0.15* 18 EX3 3.31 1.12 −0.03 −0.16* −0.24** −0.08 −0.14 −0.29** −0.06 −0.15* −0.28** 0.06 −0.17** −0.20** 19 CY1 3.34 0.97 −0.07 −0.03 0.05 −0.03 −0.07 0.00 −0.29** −0.28** −0.14 −0.25** −0.15* −0.20** 20 CY2 2.82 0.99 0.06 −0.02 −0.12 −0.03 −0.16* −0.17* −0.10 −0.32** −0.24** −0.15* −0.34** −0.35** 21 CY3 2.67 1.07 −0.01 −0.15* −0.29** −0.10 −0.23** −0.28** −0.15* −0.30** −0.40** −0.03 −0.27** −0.38** 22 RE1 3.92 0.89 −0.00 0.02 0.02 −0.07 −0.04 −0.02 −0.28** −0.25** −0.18* −0.32** −0.20** −0.20** 23 RE2 3.38 1.13 0.11 0.03 −0.08 −0.10 −0.18* −0.22** −0.06 −0.30** −0.23** −0.23** −0.34** −0.28** 24 RE3 3.20 1.19 0.04 −0.17* −0.29** −0.13 −0.28** −0.31** −0.06 −0.28** −0.32** −0.09 −0.29** −0.38** Variable 13 14 15 16 17 18 19 20 21 22 23 24 1OB1 2OB2 3OB3 4DE1 5DE2 6DE3 7AC1 8AC2 9AC3 10 NJ1 11 NJ2 12 NJ3 13 NR1 0.83 14 NR2 0.56** 0.84 Mindfulness (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’salphas) 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 postand fup10 measurements. At the premeasurement, 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-tosample 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 Yis described at a time t, and ΔY t is defined as the change in Yfrom t–1tot (McArdle 2009). In the current study, the change scores were calculated for the factors instead of observed variables; hence, Yhere 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 preto post-measurement and from postto fup10 measurement in the mindfulness facets and burnout dimensions (marked D in the figure). Modeling was conducted, adjusting Table 1 (continued) 15 NR3 0.47** 0.59** 0.88 16 EX1 −0.16* −0.05 0.01 0.75 17 EX2 0.05 −0.06 −0.04 0.57** 0.81 18 EX3 0.08 −0.07 −0.19* 0.47** 0.70** 0.82 19 CY1 −0.31** −0.20** −0.17* 0.28** 0.23** 0.16* 0.79 20 CY2 −0.11 −0.23** −0.28** 0.17** 0.52** 0.38** 0.59** 0.83 21 CY3 −0.07 −0.13 −0.36** 0.14 0.45** 0.64** 0.39** 0.64** 0.85 22 RE1 −0.24** −0.10 −0.10 0.30** 0.25** 0.21** 0.66** 0.46** 0.35** 0.65 23 RE2 −0.06 −0.13 −0.21** 0.20** 0.55** 0.38** 0.41** 0.76** 0.55** 0.58** 0.81 24 RE3 −0.08 −0.15* −0.33** 0.12 0.47** 0.60** 0.31** 0.57** 0.84** 0.40** 0.63** 0.83 OB observing, DE describing, AC acting with awareness, NJ non-judging, NR non-reacting, EX exhaustion, CY cynicism, RE reduced professional efficacy, 1pre-measurement, 2post-measurement, 3 fup10 measurement. Cronbach’s alphas are italicized **p<0.01 *p<0.05 Mindfulness 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 (χ 2 ), 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), β 2 ,β 9 , and β 10 were the most relevant coefficients to the research questions about the mediation effects. β 2 showed whether the intervention affected the change from preto postmeasurement in each mindfulness facet. β 9 and β 10 showed whether the change in a mindfulness facet from preto postmeasurement predicted the change in a burnout dimension from preto post-measurement (β 9 ) and from postto fup10 measurement (β 10 ). Furthermore, the pathways β 2 ×β 9 and β 2 ×β 10 showed whether the effects of the intervention on the burnout dimensions were mediated by the changes in the mindfulness facets (indirect effects). β 2 ×β 9 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, β 2 ×β 10 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. β 3 showed whether the intervention had an additional direct effect on the change from preto Fig. 1 Latent change score model of mindfulness (x) and burnout (y). The same model was used separately for each combination of mindfulness facets and burnout dimensions. The essential pathways are bolded. A = pre, L = post, E = fup10, D = latent difference score, Le = individual level. Previously identified structures of the five mindfulness facets (Baer et al. 2006) and three burnout dimensions (Näätänen et al. 2003)wereutilized as bases for parcel creation, and the items of both measures were divided into three parcels Mindfulness 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 (β 2 ×β 9 ,β 2 ×β 10 ,β 2 ,β 9 ,andβ 10 ) 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. Results The means, standard deviations, and correlations for the variables are shown in Table 1. In Table 2, the results of the fit indexes of the change score models are presented for each pair of mindfulness facets and burnout dimensions. All the models had a satisfactory fit with the data. Mindfulness Facets Mediated Intervention Effects on Burnout Dimensions Tables 3,4,5,6,and7show the βvalues 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, β 2 ,β 3 ,β 9 , β 10 ,β 2 ×β 9 ,andβ 2 ×β 10 . 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 (β 2 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 β 2 ,β 9 , β 10 ,β 2 ×β 9 ,andβ 2 ×β 10 and differences thereof). Exhaustion Improvements in observing, acting with awareness, non-judging, and non-reacting during the intervention mediated the alleviation of exhaustion during the intervention (significant β 2 ,β 9 ,andβ 2 ×β 9 ). None of the significant four facets was more important than the others for the alleviation of exhaustion during the intervention (the confidence intervals overlapped between the facets). Only the improvement in non-judging during the intervention mediated the additional alleviation of exhaustion during the follow-up (significant β 2 , β 10 ,andβ 2 ×β 10 ). The intervention had an additional direct effect on the alleviation of exhaustion, after controlling for the mediators (significant β 3 ). Cynicism Improvements in all five mindfulness facets during the intervention mediated the alleviation of cynicism during the intervention. Improvements in describing and non-judging Table 2 Model fits for change score models of mindfulness facets and burnout dimensions Mindfulness Burnout χ 2 pvalue RMSEA CFI TLI SRMR OB EX 215.78 .0004 0.047 0.967 0.962 0.072 DE EX 232.02 < .0001 0.052 0.969 0.964 0.076 AC EX 247.60 < .0001 0.057 0.955 0.949 0.070 NJ EX 182.47 .0365 0.033 0.985 0.983 0.055 NR EX 183.17 .0338 0.033 0.982 0.980 0.056 OB CY 194.58 .0084 0.038 0.978 0.975 0.065 DE CY 237.31 < .0001 0.054 0.967 0.963 0.086 AC CY 228.35 < .0001 0.051 0.965 0.960 0.084 NJ CY 170.23 .1236 0.026 0.991 0.990 0.066 NR CY 155.74 .3574 0.014 0.997 0.997 0.054 OB RE 172.54 .1004 0.027 0.987 0.985 0.065 DE RE 222.71 .0001 0.049 0.970 0.966 0.075 AC RE 195.36 .0075 0.039 0.977 0.973 0.065 NJ RE 200.56 .0037 0.041 0.975 0.972 0.066 NR RE 142.20 .6627 0.000 10.00 10.01 0.053 OB observing, DE describing, AC acting with awareness, NJ non-judging, NR non-reacting, EX exhaustion, CY cynicism, RE reduced professional efficacy, χ 2 chi-square test, RMSEA root mean square error of approximation, CFI comparative fit index, TLI Tucker-Lewis index, SRMR standardized root mean square residual A good model fit was assumed when CFI and TLI were close to 0.95, the SRMR was close to 0.08, and the RMSEA was close to 0.06 Mindfulness