Although social work is a very rewarding profession, from which practitioners often attain very high rates of personal accomplishment (McFadden, 2015), it is also a very challenging profession which can lead to social workers experiencing concerningly high rates of stress and burnout (Evans et al., 2006; Ravalier et al., 2021). In a large-scale study, Ravalier et al. (2021) (n = 3,421) found that social workers in the United Kingdom (UK) experienced much higher rates of stress than a general English population (Warttig et al., 2013), with 60% of their sample highlighting that they planned to leave their current position. In another large-scale study, McFadden (2015) (n = 1,359) found, that in comparison to other health and social care workers, social workers in the UK experienced very high rates of emotional exhaustion and depersonalisation. Evans et al. (2006) (n = 237) found that mental health social workers in England and Wales experience higher rates of emotional exhaustion than consultant psychiatrists, and rates of emotional exhaustion that are three times higher other mental health practitioners. Practitioner stress and burnout can impact the social work profession negatively, at both a micro, and macro level. On a micro level, it can directly impact on service user care by reducing the social worker’s capacity for attentiveness, empathy, optimism, collaboration, and capacity to connect emotionally with, or risk assess what the service user’s needs might be (Salyers et al., 2017; Savaya et al., 2016). On a macro level, social worker stress and burnout has been linked to poor job performance, reduced organisational commitment (Savaya et al., 2021), productivity, increased training costs and disruption of services, increased employee sick days, and turnover (Ravalier et al., 2021). Turley et al. (2021) highlight that the average length of a social work career in the United Kingdom is less than 8 years, which is significantly shorter than that of nurses (16 years) and medical doctors (25 years).

The evidence that Mindfulness-based programmes (MBP) could help social workers to adapt to and recover from feelings of stress and burnout is growing (Craigie et al., 2016; Kinman et al., 2020; Lynn & Mensinga, 2015; Miller et al., 2020). In a randomised controlled trial (RCT) of the Mindfulness-based Social Work and Self-Care programme (MBSWSC), Maddock et al. (2023) (n = 62) found that MBSWSC participation led to social workers experiencing large significant reductions in stress and emotional exhaustion versus a mindfulness and self-compassion control group (MSC). MBSWSC group participants also experienced reduced depersonalisation, but there was no significant difference in this outcome versus MSC (Maddock et al. 2023). In a follow-up study aiming to replicate these findings among a broader range of professionals, including social work managers, Maddock et al. (2024) (n = 60) found that MBSWSC participants experienced large significant reductions in stress, emotional exhaustion, and a moderate significant reduction in depersonalisation versus MSC.

It is currently unclear how MBSWSC improved these outcomes, and over the last two decades, the scientific mindfulness literature has consistently highlighted the need to verify the most important mechanisms of MBPs, which influence changes in mental health and well-being outcomes, such as stress and burnout (Gu et al., 2015; Maddock & Blair, 2023; Van der Velden et al., 2015). The identification of the key mechanisms of action of MBPs would enhance our understanding of how they work for specific groups, such as social workers (Kazdin, 2009). It could also support the development of other effective programmes of support, through the enhancement of the most potent mechanisms of action e.g., if self-compassion was found to be a key mechanism, self-compassion activities could be utilised to improve stress and burnout in social work (Miller et al., 2020; Svendsen et al., 2017). The empirical evaluation of theories, which help to explain the onset, maintenance, and recurrence of stress and burnout in social workers, and how the deleterious effects of these outcomes can be ameliorated through MBP participation, could be an important step in more clearly identifying how MBPs lead to beneficial effects for social workers (Brown, 2015; Carlson, 2015; Kinman et al., 2020; Miller et al., 2020). There is only one theory within social work which focusses on how MBPs might improve stress and burnout in social workers, the clinically modified Buddhist psychological model for social work practice and self-care (CBPM) (Maddock, 2023). The CBPM is presented in full elsewhere (Maddock, 2023). In short, it focusses on the effects of six mindfulness mechanisms of action (CBPM domains): mindfulness, self-compassion, attention regulation, acceptance, non-aversion, and non-attachment, and how they can function both individually, and collectively, to reduce repeated negative thinking patterns (worry and rumination) which often present as cognitive avoidant coping strategies. The development of each CBPM domain leads to increased tendencies in social workers to approach the sources of their stress (e.g., through the fuller engagement with, and processing of difficult emotions which can result from social work practice), which leads to reduced feelings of stress and burnout (emotional exhaustion, depersonalisation, and increased personal accomplishment), both directly, and indirectly through reduced worry and rumination (Maddock, 2023).

This study has two aims: (1) To provide a greater understanding of how MBPs improve stress and burnout in social workers, using pre and post RCT data from a sample of social workers who had completed a MBP in Maddock et al. (2023) and Maddock et al. (2024), using the CBPM as a theoretical framework; (2) investigate if changes in the CBPM domain and mediating variable scores after engaging in a MBP predicted the stress and burnout levels of social workers. To achieve these aims, the study tested hypotheses related to the CBPM's structural equation models' fit and the direct and indirect effects of specific mindfulness mechanisms on stress and burnout.

Method

Participants

The participants were from two randomised controlled trials examining the effects of mindfulness-based programmes (MBPs) on stress and burnout among social workers in Northern Ireland (Maddock et al., 2023; Maddock et al., 2024). To examine the changes in each CBPM domain, mediating and outcome variable (i.e., stress and burnout) more rigorously in each MBP, only participants (n = 94) who completed both the experimental MBSWSC (n = 56) and control MSC (n = 38) programmes in Maddock et al. (2023) and Maddock et al. (2024) and provided post programme data were included in this study. A total of 83 female (88%) (51 in the MBSWSC programme, and 32 in the MSC programme), and 11 male (12%) (5 in the MBSWSC, and 6 in the MSC programme) social work professionals completed the MBP that they were allocated to, and the study’s measures, across both RCTs. This gender split is in line with the overall gender profile of social workers in Northern Ireland (Hughes, 2022). The sample included social work professionals in Northern Ireland who engaged with service users as part of their role, including social workers, senior social work practitioners, and social work service managers. This sample size is sufficient to power structural equation models of fit, as the sample contained more than ten participants per estimated variable (Bentler & Chou, 1987). The sample size for this study also means that Hypothesis 2 and 3 are powered sufficiently to detect large moderated mediation effects at a power of 80% (Fritz & Mackinnon, 2007).

Procedure

MBSWSC and MSC follow a protocol, and were delivered by the same facilitators in both Maddock et al. (2023) and Maddock et al. (2024). Fidelity to each MBPs protocol was assessed after each session, with an adapted treatment fidelity tool, which is based on Kechter et al. (2019) recommendations for reporting treatment fidelity in MBP trials.

MBSWSC is based on the CBPM (Maddock, 2023), and theories of cognitive and emotion regulation and stress coping. MBSWSC is a unique group-based, online MBP (delivered via MS Teams), in that it focusses on social work practice and self-care, with a key goal of the programme being to support social workers to reduce their levels of stress and burnout, through increased introspective awareness of the interconnections between thoughts, emotions, physical sensations, and behaviours. MBSWSC does this through three main teaching modalities, delivered each week, over a 6-week period: (1) psychoeducation on what key concepts such as mindfulness, attention regulation, acceptance, self-compassion, non-aversion, non-attachment, worry, rumination, avoidant and approach stress coping are, and what their potential roles are in reducing the onset, maintenance and recurrence of stress and burnout, (2) experiential learning through engagement with, and reflection on participation in CBPM domain focussed body scan exercises. The goal being to allow participants to learn experientially, and to better integrate and embody the programme’s didactic content, and (3) apply their conceptual and experiential learning to social work case contexts through the engagement with role play scenarios which are common in social work e.g., dealing with a challenging service user. The participants engaged with several different body scan meditations over the 6-week duration of the programme. The goal of these body scan mediations was to support participants to develop a personal mindfulness practice which focussed on improving each CBPM domain, with a view to reducing repeated negative thinking (worry and/or rumination), stress, and burnout, in line with the CBPM theory (Maddock, 2023). These were: (1) a general mindfulness body scan meditation, (2) a attention regulation body scan, (3) a acceptance body scan meditation, (4) a non-aversion and non-attachment body scan, and (5) a self-compassion body scan meditation (including the writing of a self-compassionate letter) to help participants develop, refine and practice mindfulness skills/techniques. To support participants to develop, refine and practice these mindfulness skills/techniques in their day-to-day lives and social work practice, participants were requested each week to complete homework activities, which incorporated these body scans. This homework took 20-30 min to complete, 6 out of the 7 days per week over the programme’s 6-week duration. The MBSWSC programme was facilitated by two professionally qualified social workers, who are trained mindfulness facilitators. More details on the MBSWSC programme, and its delivery are published in Maddock et al. (2023).

The mindfulness and self-compassion programme (MSC) was adapted from the Mindfulness Based Living Course programme, and delivered in an online group format (via MS Teams) over the same 6-week period (MBLC: Choden & Regan-Addis, 2018). The aim of the MSC programme was to introduce participants to the concepts of mindfulness and self-compassion, and related experiential practices. Sessions included key learning about: (1) how body scans can shift the attention away from the mind to the body, (2) that this process can support the development of embodied reflexivity, supporting social work practice, (3) mindfulness, and (4) self-compassion’s potential role in reducing stress, burnout, and improving social work practice. The participants were introduced to three experiential practices which were deemed to be relevant to social work practice: (1) the body scan, (2) the 3-min breathing space (3MBS), and (3) the self-compassion break (SCB). In order to embed the key learning from these exercises, and support their use in the participants everyday lives, visual mnemonics for the 3MBS and SCB were provided. To support participants to develop, refine and practice these mindfulness skills/techniques, they were requested to complete homework activities. These homework activities also took approximately 20–30 min to complete, 6 days per week over the same 6-week period. The MSC programme was also facilitated by two professionally qualified social workers, who were trained mindfulness facilitators. More details on the MSC programme, and its delivery is published in Maddock et al. (2023).

Measures

Demographic information was collected at the baseline of both RCTs. In order to empirically examine the relationships set out in the CBPM, self-report measures of each CBPM domain, mediating and outcome variable were collected pre and post MBP completion. Scale reliabilities were calculated using the responses provided.

The Perceived Stress Scale (PSS) is a widely used valid and reliable 10-item measure of perceived stress (Cohen et al., 1983). The PSS is scored on a 5-point Likert scale (0=never to 4=very often), with higher scores indicating higher perceived stress. The reliability of the scores on the PSS in both RCTs with social workers was deemed to be acceptable (Cronbach’s α = 0.88) (Maddock et al., 2023), and (Cronbach’s α = 0.87) (Maddock et al., 2024). McDonald's omega for the PSS in this study was 0.86 (Hayes & Coutts, 2020).

The Maslach Burnout Inventory (MBI) is reliable, valid, and the most widely used measure of occupational burnout (Maslach et al., 1996; Maslach et al., 2001). The MBI is a 22-item measure and is scored on a 7-point Likert scale (0=never to 6=everyday), with 3 subscales measuring: emotional exhaustion, depersonalisation/loss of empathy, and personal accomplishment. Maddock et al. (2023) and Maddock et al. (2024) confirmed moderate to high reliability on the MBI subscales reporting Cronbach’s α = 0.91, 0.69 and 0.75, and α = 0.92, 0.81 and 0.84 for emotional exhaustion, depersonalisation, and personal accomplishment, respectively. McDonald's omega for the emotional exhaustion, depersonalisation, and personal accomplishment in the subscales in this study were 0.90, 0.64, and 0.75, respectively (Hayes & Coutts, 2020).

The Southampton Mindfulness Questionnaire (SMQ) is a valid and reliable 16-item measure that assesses the degree to which a person can remain mindful in response to unpleasant thoughts and images (Chadwick et al., 2008). The SMQ contains 8 positively and negatively worded items, on a 7-point Likert scale (0=disagree totally to 6=agree totally). Scores on this measure range from 0-96. High scores on this measure indicate higher levels of mindfulness. Maddock et al. (2023) and Maddock et al. (2024) reported good internal consistency among social workers (α=0.87, and 0.87, respectively). This scale also delves into the components of mindfulness comprising 2 subscales which measure components of the CBPM; Letting Go/Non-attachment (SMQ-LG); Non-aversion (SMQ-AV). Maddock et al. (2023) and Maddock et al. (2024) reported acceptable Cronbach’s alpha of 0.76 and 0.74 for the SMQ-LG, and good to acceptable Cronbach’s α of 0.82 and 0.60 for the SMQ-AV, respectively. McDonald's omega for the SMQ in this study was 0.60 for SMQ-AV, and 0.70 for SMQ-LG (Hayes & Coutts, 2020).

The Experiences Questionnaire – Decentering (EQ-D; Fresco et al., 2007) is a valid and reliable 11-item measure of the capacity to regulate attention through decentering i.e., an individual’s ability to objectively observe their thoughts and emotions (Fresco et al., 2007). The EQ-D is scored on a 5-point Likert scale (1=never; 5=always), and scores range from 11-55, with higher scores indicative of higher levels of attention regulation. The reliability of the scores on the EQ-D in both RCTs with social workers was deemed to be acceptable (Cronbach’s α = 0.84) (Maddock et al., 2023), and (Cronbach’s α = 0.84) (Maddock et al., 2024). McDonald's omega for the EQ-D in this study was 0.88 (Hayes & Coutts, 2020).

The Philadelphia Mindfulness – Acceptance Subscale (PHLSMS-A) is a valid and reliable 10-item measure of acceptance (Cardaciotto et al., 2008). It is scored on a 5-point Likert scale (1=never; 5=very often); with total scores ranging from 10-50. Higher levels of acceptance are indicated by lower scores. Maddock et al. (2023) and Maddock et al. (2024) reported good scale reliabilities among social workers, with a Cronbach’s alphas of 0.89 and 0.88 respectively. McDonald's omega for the PHLMS-A in this study was 0.89 (Hayes & Coutts, 2020).

The Self-Compassion Scale-Short Form (SCS-SF; Raes et al., 2011) is a valid and reliable 12-item measure of self-compassion, which is scored on a 5-point Likert scale (1=almost never; 5=almost always). Higher scores indicate higher levels of self-compassion. The reliability of the scores on the SCS was found to be acceptable in both Maddock et al. (2023) and Maddock et al. (2024) (Cronbach’s α = 0.87, and 0.86). McDonald's omega for the SCS in this study was 0.86 (Hayes & Coutts, 2020).

The Penn State Worry Questionnaire (PSWQ; Meyer et al., 1990) is a valid and reliable 16-item measure of the intensity, pervasiveness, and uncontrollability of worry (Startup & Erickson, 2006). PSWQ scores range from 16-80, with higher scores indicating higher levels of worry (Startup & Erickson, 2006). The reliability of the scores on the PSWQ in in both Maddock et al. (2023) and Maddock et al. (2024) was found to be high (Cronbach’s α = 0.95, and 0.94). McDonald's omega for the PSWQ in this study was 0.94 (Hayes & Coutts, 2020).

The Rumination Reflection Questionnaire (RRQ; Trapnell & Campbell, 1999) is valid and reliable 12-item measure of rumination. Scores range from 12-60 on this 5-point Likert scale (1=strongly disagree; 5=strongly agree), with higher scores indicating higher rates of rumination. Maddock et al. (2023) and Maddock et al. (2024) both highlighted the scales high internal consistency with social workers (Cronbach’s α = 0.94, and 0.94). McDonald's omega for the RRQ in this study was 0.92 (Hayes & Coutts, 2020).

Data Analyses

To test if mediated effect, and direct and mediated effect CBPM models of stress, emotional exhaustion, depersonalisation, and personal accomplishment would be good fits to the data, structural equation modelling (SEM) was used. SEM is a multicomponent statistical process which can allow researchers to construct theoretical models, such as the CBPM, and test to see if the relationships set out in such theories fit a dataset, providing evidence for, or against, the theory as being a useful explanatory framework (Malaeb et al., 2000). To ensure that the SEM analyses were statistically valid, through having enough participants to support the number of variables included in each SEM model, the total score for the SMQ was used to assess levels of mindfulness. This means that each CBPM SEM model included four (mindfulness, self-compassion, attention regulation, and acceptance) of the six CBPM domains. Data were screened for errors, missing values, and outliers using the interquartile rule (Hoaglin et al., 1986) on SPSS 27 (IBM, Armonk, NY), with whisker and box plots. A complete dataset is needed to construct SEMs, so any missing data was dealt with using the expectation-maximization method (Schumacker & Lomax, 2016). Measurement and structural models based on the CBPM were developed (model specification and identification). Four mediated effect, and four direct and mediated effect models representing different potential explanatory frameworks of how the CBPM might help to explain changes in stress, emotional exhaustion, depersonalisation, and personal accomplishment were specified and estimated using LISREL 10.3.4.4 (Joreskog & Sorbom, 2009). A covariance and an asymptotic weight matrix were computed with the parameters being estimated using maximum likelihood (Joreskog & Sorbom, 2009). The fit of each model to the data was then assessed (model estimation and testing) (Schumacker & Lomax, 2016). When reporting the assessment of model fit in SEM, it is important to use several different indices, as there are no consensus rules within the literature that researchers can follow (Crowley & Fan, 1997; Hooper et al., 2008). In line with Hooper et al. (2008) and Kline (2005), this study includes the Chi-Square statistic (where a non-significant chi-square indicates model fit), its degrees of freedom and p value. It also includes a number of relative fit indices: the Standardised Root Mean Square Residual (SRMR) (< 0.05, Hooper et al., 2008), the Comparative Fit Index (CFI) (≥ 0.95, Hu & Bentler, 1999), and the parsimony fit index - PNFI (< 0.05, Hooper et al., 2008), and the RMSEA, which is an absolute fit index, where values which are ≥0.06 are taken indicate good model fit (Hu & Bentler, 1999). The Akaike Information Criterion (AIC; Akaike, 1987) was used for the purposes of model comparison, with the smallest value being indicative of the most parsimonious model (Byrne, 2012). In this study, if the Chi-Square test, more than one relative fit index (SRMR, CFI, PNFI), and the RMSEA are deemed to indicate model fit in this study, the model will be deemed to be a good fit to the data. If only one fit index (the chi-square, more than one relative fit index such as SRMR, CFI, PNFI, or the RMSEA) deems the model to fit the data, the model will be deemed to be an acceptable fit. If none deem the model to fit the data, the model will be deemed to be a poor model fit.

To examine the direct and indirect effect (through worry and/or rumination) of each CBPM domain (mindfulness, attention regulation, acceptance, self-compassion, non-attachment, non-aversion) on social worker stress and burnout (emotional exhaustion, depersonalisation, and personal accomplishment), several conditional indirect and direct effect models were tested. These analyses were performed using SPSS 27.0 (IBM, Armonk, NY), with the SPSS PROCESS macro (Hayes, 2018). Preacher and Hayes’ bias-corrected non-parametric bootstrapping techniques were used, with 5000 bootstrapped samples (Hayes, 2018). The models were considered statistically significant if the point estimates of the 95% confidence intervals (CIs) did not contain zero (Furr, 2018; Hayes, 2018). To further examine that any significant effects found where due to MBP participation and not merely associations between the variables of interest, these tests were also run on the participants baseline data, and compared. If significant differences were found i.e., significant direct or indirect effects were found on the change score dataset, and not in the baseline dataset, these significant effects were attributed to MBP participation.

Results

SEM Model Fit Statistics

The model fit statistics for all 8 mediated, and direct and mediated CBPM models are presented in Table 1.

Table 1 SEM model fit statistics
Full size table

Stress

The fit indices in Table 1 indicate that the mediated effects CBPM model has a good model fit to the data on stress change scores. The Chi-Square is non-significant (p = 0.76), which is > 0.05; the RMSEA = 0.00, which is < 0.06 (Hu & Bentler, 1999), and the CFI = 0.99, which is > 0.95 (Hu & Bentler, 1999), and the SRMR = 0.02, which is < 0.05 (Hooper et al., 2008). The direct and mediated effects CBPM model was also found to be a good model fit to the data on stress change scores. The Chi-Square is non-significant (p = 0.40), which is > 0.05; the RMSEA = 0.00, which is < 0.06 (Hu & Bentler, 1999), and the CFI = 0.99, which is > 0.95 (Hu & Bentler, 1999), and the SRMR = 0.01, which is < 0.05 (Hooper et al., 2008).

Burnout-Emotional Exhaustion

The fit indices in Table 1 indicate that the mediated effects CBPM model was not a good model fit to the data on Burnout-emotional exhaustion change scores. The direct and mediated effects CBPM model was found to be a good model fit to the data on Burnout-emotional exhaustion change scores. The Chi-Square is non-significant (p = 0.39), the RMSEA = 0.00, the CFI = 0.99, and the SRMR = 0.01.

Burnout-Depersonalisation

The fit indices in Table 1 indicate that the mediated effects CBPM model has a good model fit to the data on Burnout-depersonalisation. The Chi-Square is non-significant (p = 0.2), the RMSEA = 0.07, the CFI = 0.99, and the SRMR = 0.04. The direct and mediated effects CBPM model was found to be a good fit to the data on Burnout-depersonalisation change scores. The Chi-Square is non-significant (p = 0.4), the RMSEA = 0.00, the CFI = 0.99, and the SRMR = 0.01.

Burnout-Personal Accomplishment

The fit indices in Table 1 indicate that the mediated effects CBPM model has a good model fit to the data on Burnout-personal accomplishment change scores. The Chi-Square is non-significant (p = 0.58), the RMSEA = 0.00, the CFI = 0.99, and the SRMR = 0.03. The direct and mediated effects CBPM model was found to be a good fit to the data on Burnout-personal accomplishment change scores. The Chi-Square is non-significant (p = 0.40), the RMSEA = 0.00, the CFI = 0.99, and the SRMR = 0.01.

Conditional Direct and Indirect Effects

The conditional direct effects of the CBPM domains on the stress and burnout (emotional exhaustion, depersonalisation, personal accomplishment) of social workers is presented in Table 2, with the significant direct effects italicised. Changes in mindfulness (β = -0.26, SE = 0.07: 95% CI -0.40, -0.11), attention regulation (β = -0.38, SE = 0.14: 95% CI -0.66, -0.09), and non-aversion (β = -0.58, SE = 0.19: 95% CI -0.96, -0.19) scores post MBP participation were found to have a conditional direct effect on emotional exhaustion. Changes in mindfulness (β = -0.15, SE = 0.06: 95% CI -0.3, -0.03), self-compassion (β = -0.18, SE = 0.09: 95% CI -0.36, -0.01), and non-aversion (β = -0.35, SE = 0.18: 95% CI -0.7, -0.1), scores post MBP participation were found to have a conditional direct effect on depersonalisation. Changes in mindfulness scores were also found to have significant conditional direct effect on both stress and personal accomplishment, however these significant relationships were also present within the baseline data set. Changes in attention regulation were also found to have significant conditional direct effect on both depersonalisation and personal accomplishment, however these significant relationships were also present at baseline. Changes in non-attachment scores were found to have significant conditional direct effect on personal accomplishment, however this significant relationship was also found within the baseline dataset.

Table 2 Moderated mediation analysis - conditional direct effects
Full size table

The conditional indirect effects of the CBPM domains on the stress and burnout (emotional exhaustion, depersonalisation, personal accomplishment) of social workers, when mediated by worry or rumination are presented in Table 3, with the significant indirect effects italicised. Changes in self-compassion (β = 0.08, SE = 0.05: 95% CI 0.01, 0.18) scores were significantly associated with perceived stress when mediated by reduced rumination scores. Changes in non-attachment, self-compassion, mindfulness, attention regulation, and acceptance scores were significantly associated with perceived stress when mediated by reduced worry scores, however these significant relationships were present within the baseline dataset. Changes in non-attachment, non-aversion, and attention regulation were significantly associated with perceived stress when mediated by reduced rumination scores, however these significant relationships were found in the dataset at baseline.

Table 3 Moderated mediation analysis - conditional indirect effects
Full size table

There were no significant indirect effects found on emotional exhaustion due to MBP participation alone. Self-compassion, and acceptance were significantly associated with Burnout-emotional exhaustion when mediated by reduced worry scores, but these relationships were also found at baseline.

Changes in attention regulation (β = -0.34, SE = 0.11: 95% CI -0.56, -0.11) was significantly associated with Burnout-depersonalisation when mediated by reduced worry scores. Self-compassion (β = -0.07, SE = 0.04: 95% CI -0.16, -0.01), and mindfulness change scores (β = -0.09, SE = 0.04: 95% CI -0.18, -0.01) were significantly associated with Burnout-depersonalisation when mediated by reduced rumination scores. Non-attachment, non-aversion, and acceptance were significantly associated with Burnout-depersonalisation when mediated by reduced rumination scores, but these relationships were also present in the baseline data.

Non-aversion (β = 0.17, SE = 0.09: 95% CI 0.02, 0.36), self-compassion (β = 0.08, SE = 0.04: 95% CI 0.01, 0.17), attention regulation (β = 0.17, SE = 0.09: 95% CI 0.02, 0.36), acceptance (β = -0.05, SE = 0.03: 95% CI -0.11, -0.01) were significantly associated with Burnout-personal accomplishment when mediated by reduced rumination scores.

Discussion

One of the main aims of this paper was to provide a greater understanding of how MBPs alleviate stress and burnout in social workers, using pre and post RCT data from a sample of social work professionals who had completed a MBP, using the CBPM as a guiding theoretical framework (Maddock, 2023). This paper provides promising preliminary evidence that the mediated effects, and direct and mediated effects CBPM models of stress, depersonalisation and personal accomplishment are good explanatory frameworks of how engaging in a MBP might improve these outcomes for social workers. This paper also provides initial evidence that the direct and mediated effects CBPM model is a good explanatory framework of how changes in emotional exhaustion may occur for social workers due to MBP participation.

The other key aim of this paper is to investigate if changes in each CBPM domain (mindfulness, attention regulation, acceptance, self-compassion, non-attachment, and non-aversion) and mediating variable (worry and rumination) both directly and indirectly predicted the stress and burnout levels of the social workers who completed the MBPs in Maddock et al. (2023) and Maddock et al. (2024). Several of the significant direct and indirect effects found in the analysis of the change scores from Maddock et al. (2023) and Maddock et al. (2024) were also found within the baseline data, meaning that we cannot attribute these significant effects to MBP participation. Several other significant direct and indirect effects were found to exist in the change score data, which were not found in the baseline data, meaning that we can attribute these effects to MBP participation. These significant findings are in line with the results of several systematic reviews (Alsubaie et al., 2017; Maddock & Blair, 2023; Van der Velden et al., 2015), which have highlighted the potentially complex direct and mediated change processes in outcomes such as stress, emotional exhaustion, depersonalisation, and personal accomplishment following MBP participation. The results from this paper highlighted how different outcomes (e.g., emotional exhaustion) were significantly predicted directly by changes in individual CBPM domains (e.g., attention regulation), that individual outcomes were also directly predicted by multiple CBPM domains (e.g., changes in non-aversion, and mindfulness, and self-compassion scores directly predicted changes in depersonalisation scores), and that a number of CBPM domains predicted changes in a number of different outcomes (e.g., changes in mindfulness and non-aversion predicted changes in both emotional exhaustion and depersonalisation). It is also clear that this complexity applies to the results of the conditional indirect effect analyses, where improvements in individual CBPM domains (e.g., self-compassion) were found to have a significant indirect effect through reduced rumination on individual outcomes (e.g., stress), and changes in a number of CBPM domains (e.g., attention regulation, self-compassion) were found to significantly predict changes in a number of CBPM domains when mediated by either change in worry and/or rumination (e.g., depersonalisation and personal accomplishment). These findings support Kazdin (2009) who highlighted how multiple outcomes, such as stress, emotional exhaustion, depersonalisation, and personal accomplishment can be improved through multiple pathways i.e., that by improving self-compassion, stress might improve both directly and indirectly through reduced rumination.

The results from this study indicate that MBP participation appears to reduce the emotional exhaustion of social workers through improvements in mindfulness, attention regulation, and non-aversion. The finding that attention regulation had a conditional direct effect on emotional exhaustion is supported by Crowder and Sears (2017), who in a small-scale mixed methods exploratory study examining the effects of MBP participation on social workers, highlighted the potential for improved attention regulation skills to be an important protective factor against burnout in social work. In an RCT study, Hülsheger et al. (2013) found that increased mindfulness post MBP participation predicted reduced emotional exhaustion in Dutch public sector employees, supporting the results from this study. The findings that non-aversion (which in Western mental health literature most closely resembles experiential avoidance) significantly directly predicted emotional exhaustion is in line with Iglesias et al. (2010) who found that experiential avoidance was negatively associated with the emotional exhaustion of critical care nurses.

Depersonalisation in social workers appears to reduce through both direct and mediated pathways post MBP participation. Supporting this paper’s findings, Maddock and McCusker (2022) found that changes in mindfulness was a significant predictor of change in depersonalisation in social work students who completed an MBSWSC programme. The findings that non-aversion significantly directly predicted depersonalisation is also in line with Maddock and McCusker (2022), and with Iglesias et al. (2010) who both found that experiential avoidance was negatively associated with the depersonalisation of critical care nurses. The finding that changes in self-compassion post MBP participation had a conditional direct effect on the levels of depersonalisation experienced by the social workers in this study is also supported by Maddock and McCusker (2022). The findings relating to the potential roles that mindfulness, self-compassion, and non-aversion may play in reducing depersonalisation not only has important implications for our understanding of how social worker burnout may be reduced through MBP participation, but they could also help us to understand how MBP participation may improve social work practice. Depersonalisation, often conceptualised as a loss of empathy, can lead to social workers developing negative attitudes towards their service users, which can lead to increased cynicism and insensitivity being shown to the recipients of their service (Frieiro Padin et al., 2021). The results from this study indicate that MBP participation improves social worker’s competencies in being mindful, self-compassionate, and their ability to observe their thinking processes non-judgementally and without defence. It appears that if social workers were to engage in an MBP, the development of these three CBPM domains would allow them to develop increased feelings of empathy for their service users, which is a core therapeutic process in social work (Rogers, 1967; Tanner, 2020), and a key ingredient of effective social work practice (Gerdes & Segal, 2011).

The finding that changes in four (self-compassion, attention regulation, acceptance, and non-aversion) out of the six CBPM domains post MBP participation had conditional indirect effects on the personal accomplishment of the social workers when mediated by rumination, and three (self-compassion, mindfulness, and attention regulation) out of the six CBPM domains post MBP participation had conditional indirect effects on the depersonalisation of the social workers when mediated by either rumination or worry is in line with the CBPM (Maddock, 2023). These results provide further empirical support that the CBPM could be a useful explanatory framework for how experiences of depersonalisation and feelings of personal accomplishment may change due to MBP participation (Maddock, 2023). The important role that reducing rumination may have played in reducing the stress, depersonalisation and personal accomplishment of the social workers in this study, is in line with systematic reviews conducted by Gu et al. (2015) and Alsubaie et al. (2017) with clinical and non-clinical populations respectively, which highlighted the key role that reduced rumination post MBP participation appears to play in reducing the risk of onset, maintenance and recurrence of stress and burnout.

This paper provides some initial preliminary evidence, and greater theoretical clarity on how MBPs might support improvements in stress and burnout in social workers. This theoretical clarity is particularly important, as there is an increasing awareness about the need reduce social worker stress, burnout and subsequent attrition within the profession (Kinman et al., 2020; Romero-Martín et al., 2024; Turley et al., 2021), and the potentially important role that MBPs could play in social work education and practice to meet this need (Beer et al., 2020; Beer et al., 2021; Maddock, 2023). This literature highlights that MBPs could help to reduce stress and burnout, by supporting social workers to develop self-care competencies (Beer et al., 2020, 2021; Maddock, 2023; Romero-Martín et al., 2024) which are increasingly becoming recognised by social work regulatory bodies and international social work associations as key standards of proficiency that help to ensure safe and effective social work practice (NASW, 2021; NISCC, 2015; SWRB, 2019). It has been argued that the significant time commitments that accompany participation in MBSR and MBCT may not be suitable for health and social care professionals (Craigie et al., 2016). There has been calls in this literature for tailored MBPS, which are refined and adapted to meet different professional occupational needs (Calcagni et al., 2021). Social work focussed MBPs, which are refined and adapted to meet the needs of already busy and stressed social workers are needed (Beer et al., 2020; Maddock, 2023). The identification of each significant conditional direct and indirect effect of changes in each CBPM domain and mediating variable on the stress and burnout of social workers post MBP programme in this study, helps to highlight which mechanisms of action could be focussed on, or intensified when MBPs for social workers are developed or modified.

The development of each CBPM domain, through MBP participation, would likely support improved social work practice through the enhancement of a number of social work skills and competencies, including: (1) increased resilience and self-awareness, through reduced fear of the negative thoughts and emotions that are commonly avoided in social work e.g., shame (Gibson, 2016). The development of each CBPM domain would support an increased capacity to approach and recognise when difficult thoughts and emotions are triggered in practice, and how to process these adaptively, using mindfulness-based skills and practices e.g., acceptance and self-compassion (Maddock, 2023); (2) increased empathy and compassion, which would allow social workers to more clearly understand what the service user’s needs and wishes are, facilitating improved assessment and relationship-based practice (Gibson, 2016; Klinger et al., 2012); (3) improved anti-oppressive practice through increased skills in reflection (Schon, 1983), and reflexivity (McCusker, 2022). The development of each CBPM domain would support social workers to develop a clearer understanding of how negative thoughts and emotions arise e.g., shame, and corrosive impact they can have on their own lives, if they are not regulated. This would likely support the still incomplete, but more accurate assessment of the impact of the service user’s cultural and structural environment on their lived experience (Thompson, 2012). For example, if a service user in recovery from a substance use problem is experiencing feelings of shame due to stigma and/or social exclusion, through enhanced capacities in reflection and reflexivity, the social worker may be in a better position to recognise that there is an increased risk of the service user re-initiating substance use, particularly if they are not given the opportunity to emotionally ventilate (Elison et al., 2006).

Limitations and Future Research

The results are preliminary and should be interpreted with caution due to several limitations. The external validity of this study’s findings is limited by self-selection bias, as the social workers in this study wanted to take part in this study, and were randomly allocated to and completed one of the MBPs. This makes it difficult to establish how representative the participants in this study are of wider populations of social workers. The use of a single purposive and convenience sample also means that these findings cannot be generalised to a larger population of social workers (Unrau & Grinnell, 2011). The nature of SEM means that the well-fitting CBPM models for each outcome may be one of many possible models that also fit the data. The patterns of relationships in the data appear to be consistent with the theoretical relationships set out in the CBPM, but do not definitively prove that the relationships exist (Schumacker & Lomax, 2016). Both the MBSWSC programme and the MSC control group occurred within supportive group environments with trained facilitators. The impact of these environments, in which participants could share and learn from each other, and have a positive social experience is a potential confounding variable that may have impacted both MBSWSC’s and MSC’s effects. The use of one data collection point, which the change scores function as, means that causality cannot be asserted (Kazdin, 2007; Mathieu & Taylor, 2006). This study is only powered to control for type II error for large moderated mediation effects. This study was underpowered to detect for small to medium moderated mediation effects (Fritz & MacKinnon, 2007). The number of pathways tested likely reduced the study’s power further, increasing the potential risk of type II error, and of potentially statistically significant pathways not being detected. The results from this study indicate that future research that aims to investigate potential mindfulness mechanisms of action, particularly those within social work research, might benefit from using the CBPM as a guiding theoretical framework. The CBPM would benefit from having its validity and reliability established over time, across other groups of social workers, and in different cultural contexts.

In conclusion, this study provides some promising initial preliminary evidence for the CBPM as being a useful explanatory framework of how social worker stress and burnout might be improved through MBP participation. The study’s results also suggest that if social workers engage in MBPs which positively change each CBPM domain and mediating variable, they are likely to experience reduced stress and burnout.