Teaching is widely recognized as a profession with high emotional and cognitive demands, often leading to chronic stress and burnout (Paudel et al., 2022). Within this context, mindfulness—defined as receptive attention to present-moment experiences—functions as a critical psychological resource by promoting cognitive decentering and emotional non-reactivity (Bian and Jiang, 2025; Levit-Binnun et al., 2021). This metacognitive capacity allows teachers to disengage from automatic negative thought patterns, thereby attenuating the perception and impact of occupational stressors (Querstret et al., 2020; Taylor et al., 2021). For example, a mindful teacher may acknowledge feelings of being overwhelmed by workload demands without becoming entangled in cycles of rumination or anxiety. Empirical evidence consistently demonstrates a negative correlation between mindfulness and perceived stress across diverse professional groups, including educators (Kriakous et al., 2021; Hazlett-Stevens, 2020). Thus, we hypothesize:

H1: Mindfulness is negatively associated with occupational stress.

The harmful effects of chronic occupational stress on mental health are well-documented in educational settings (Tran et al., 2024; Ansley et al., 2021). Prolonged exposure to stressors depletes psychological resources, culminating in symptoms of anxiety, depression, and emotional exhaustion, which are core components of burnout and psychological distress (Belinda and Wei, 2024). By mitigating the perceived intensity of occupational stressors, mindfulness is theorized to indirectly promote better mental health outcomes. Consequently, we propose occupational stress as a pivotal mediating mechanism:

H2: Occupational stress is positively associated with poor mental health.

H3: Occupational stress mediates the relationship between mindfulness and mental health.

Theoretical and empirical evidence suggests that mindfulness and CSE constitute distinct yet complementary psychological resources (Eikenhout et al., 2022; McGee et al., 2020). Mindfulness cultivates awareness and nonjudgmental attention to present experiences, whereas CSE reflects an individual’s confidence in their ability to execute effective coping strategies when facing challenges (Chesney et al., 2006). A teacher with high mindfulness may clearly recognize a stressful situation without experiencing immediate emotional dysregulation. However, without adequate CSE, the individual may lack the confidence to apply effective coping strategies, allowing the stressor to persist and contribute to overall occupational stress (Çiçek et al., 2025). Conversely, high CSE enables a mindful teacher to translate their calm awareness into concrete, effective actions, thereby neutralizing stressors more successfully (Rojas et al., 2022). Accordingly, we hypothesize that CSE amplifies the stress-buffering effect of mindfulness:

H4: CSE moderates the relationship between mindfulness and occupational stress, such that the negative relationship is stronger for teachers with high CSE.

Integrating H3 and H4, we propose a moderated mediation model (Hayes, 2018) wherein the strength of the indirect effect of mindfulness on mental health (via occupational stress) is contingent upon the level of CSE. The protective effect of mindfulness is expected to be strongest among teachers who combine mindful awareness with high confidence in their coping abilities.

H5: The indirect effect of mindfulness on mental health through occupational stress is moderated by CSE, such that this indirect effect is stronger for teachers with high CSE.

Following ethical approval from the Institutional Review Board of Wuhan Wudong Hospital (Approval: 230630), data were collected between March and May 2023. Teachers from public junior high schools across five cities of Henan Province, China, were recruited using a convenience sampling method. School principals were contacted first to obtain permission for the study. Teachers were subsequently invited to participate through internal communication channels. Volunteers accessed an online questionnaire hosted on the Qualtrics platform, which began with a detailed information sheet and an electronic informed consent form. Participants were assured of the confidentiality and anonymity of their responses. The survey required approximately 20–25 minutes to complete.

An a priori power analysis was conducted using G*Power (version 3.1.9.7; Heinrich Heine University Düsseldorf, Düsseldorf, Germany) (Faul et al., 2009) for a multiple linear regression model with interaction terms, consistent with our planned moderated mediation framework. Specifically, the analysis accounted for five predictors in the most complex regression step. We set an effect size of f² = 0.15 (medium, consistent with typical effect sizes for interaction effects in psychological and medical research), $\alpha$ = 0.05, and power = 0.95. This analysis indicated a minimum required sample size of 138. Our final sample of 462 substantially exceeds this threshold, even for detecting interaction effects. Thus, our sample size provides ample statistical power to detect not only main effects but also the critical interaction effect between mindfulness and CSE.

The inclusion criteria for the study participants were as follows: (1) being a teacher in mainland China; (2) serving as an in-service subject teachers at the public junior high school level; and (3) being currently employed. The exclusion criteria were as follows: (1) educational administrators (e.g., school principals, deans, or department heads whose primary responsibilities are administrative rather than teaching); (2) retired teachers; and (3) temporary or substitute teachers. Initial screening of 485 submitted responses led to the removal of 23 cases: 15 for being substantially incomplete ($>$20% missing data) and 8 for clear straight-line or patterned responding. The final analytical sample consisted of 462 teachers. Demographic characteristics are presented in Table 1. The sample was predominantly female (84.6%), which is representative of the gender distribution in junior high school teaching in China (Gu, 2023). The age and experience distribution was varied, with a slight majority being over 40 years old and having more than 11 years of experience.

All measures were administered in their previously validated Chinese versions. We proactively implemented both procedural and statistical strategies to minimize the potential for common method variance (CMV), as recommended by Podsakoff et al. (2003). Procedural strategies included ensuring respondent anonymity, reducing evaluation apprehension, and carefully counterbalancing the order of scale items to minimize bias.

Data analyses were conducted using SPSS 27.0 (IBM Corp., Armonk, NY, USA) and AMOS 27.0 (IBM Corp., Armonk, NY, USA).

Preliminary Analysis: Descriptive statistics (means and standard deviations) were examined, and univariate normality was assessed by confirming that skewness and kurtosis values for all scale scores fell within acceptable limits (skewness $|$2$|$, kurtosis $|$7$|$) (West et al., 1995). Independent samples t-tests were conducted to examine gender differences in the key study variables.

CMV: Given the cross-sectional, self-report design, CMV was assessed statistically using Harman’s single-factor test. All items from all scales were entered into an exploratory factor analysis (EFA) using principal axis factoring without rotation.

Measurement Model Validation: Before testing the structural hypotheses, a full CFA was run with all four latent constructs (Mindfulness, Occupational Stress, CSE, Mental Health) to assess the overall fit of the measurement model and to establish convergent and discriminant validity. Convergent validity was confirmed if Average Variance Extracted (AVE) values were above 0.50 and Composite Reliability (CR) was above 0.70. Discriminant validity was assessed using the Fornell-Larcker (1981) criterion: the square root of the AVE for each construct should be greater than its correlations with any other construct.

Multicollinearity Check: Before running regression analyses, the Variance Inflation Factors (VIFs) were checked for all predictor variables in the proposed models. A VIF value greater than 3.3 (or 5, using a more conservative threshold) would indicate problematic multicollinearity (Kock and Lynn, 2012).

Hypotheses Testing: The moderated mediation hypotheses were tested using the PROCESS Macro for SPSS (Version 4.2, Model 7; The Ohio State University, Columbus, OH, USA) for SPSS. The Ohio State University, Columbus, OH, USA developed by Hayes (2018), with 10,000 bootstrap samples to generate bias-corrected confidence intervals (CIs). Mindfulness was entered as the independent variable (X), occupational stress as the mediator (M), mental health (GSI) as the dependent variable (Y), and CSE as the moderator (W) of the X$\to$M path. A significant interaction term (X*W) and a significant index of moderated mediation would support our hypotheses.

Descriptive statistics and zero-order correlations for all study variables are presented in Table 2. All variables were significantly correlated in the expected directions (p 0.01). As hypothesized, mindfulness was strongly negatively correlated with occupational stress (r = –0.76) and psychological distress (r = –0.88), and positively correlated with CSE (r = 0.85). Occupational stress was strongly positively correlated with psychological distress (r = 0.83) and negatively with CSE (r = –0.69). CSE was negatively correlated with psychological distress (r = –0.78). The high magnitude of some correlations prompted a check for multicollinearity. In the regression models, all VIF values were well below 2.5, indicating that multicollinearity did not pose a concern for interpreting the results.

Harman’s single-factor test revealed that the unrotated factor solution extracted multiple factors, with the first factor accounting for 28.7% of the total variance. As this is well below the critical threshold of 50%, we concluded that CMV did not pose a serious threat to the validity of our findings.

Confirmatory factor analysis of the four-factor measurement model demonstrated excellent fit to the data: $\chi$²(984) = 2150.82, $\chi$²/df = 2.19, comparative fit index (CFI) = 0.93, tucker-lewis index (TLI) = 0.92, RMSEA = 0.051 (95% CI: [0.049, 0.054]), SRMR = 0.045. All standardized factor loadings were statistically significant (p 0.001) and exceeded 0.60, indicating good indicator reliability.

As shown in Table 3, CR for each construct ranged from 0.83 to 0.96, far exceeding the recommended threshold of 0.70, which confirms strong internal consistency. The AVE for each construct ranged from 0.58 to 0.62, exceeding the 0.50 criterion and thereby supporting the convergent validity of the measurement model.

For discriminant validity, we applied the Fornell-Larcker criterion, comparing the square roots of the AVEs for each construct (shown in bold on the diagonal in Table 3) with their correlations with other constructs. Most constructs met this criterion, except for three pairs with high correlations: (1) The square root of the AVE for Mindfulness (Construct 1) is 0.76, whereas its correlations with Occupational Stress (Construct 2) is –0.81 (absolute value = 0.81) and with Mental Health (GSI, Construct 4) is –0.80 (absolute value = 0.80), both exceeding 0.76; (2) The square root of the AVE for Occupational Stress (Construct 2) is 0.79, while its correlation with Mental Health (Construct 4) is 0.83, exceeding 0.79; and, (3) The square root of the AVE for Mental Health (Construct 4) is 0.78, while its correlation with Occupational Stress (Construct 2) is 0.83, exceeding 0.78.

To further assess discriminant validity, particularly given the high absolute correlations between certain constructs, we additionally calculated the Heterotrait-Monotrait Ratio (HTMT) of correlations, a more stringent indicator for discriminant validity. All HTMT values ranged from 0.72 to 0.87, remaining below the conservative threshold of 0.90 and the more lenient threshold of 0.95. Specifically, the HTMT values for the highly correlated pairs were well below 0.90, indicating that these constructs are empirically distinct despite their strong bivariate correlations. This supplementary HTMT analysis, combined with the Fornell-Larcker criterion, provides robust evidence for the discriminant validity of the measurement model.

The results of the moderated mediation analysis using PROCESS Model 7 are summarized in Fig. 1 and Table 4.

Fig. 1.

Final moderated mediation model with standardized path coefficients. SCL-90, Symptom Checklist-90.

Direct and Total Effects: The direct effect of mindfulness on mental health, controlling for the mediator, was significant ($\beta$ = –0.38, SE = 0.05, p 0.001, 95% CI: [–0.48, –0.28]). The total effect of mindfulness on mental health was also significant ($\beta$ = –0.79, SE = 0.04, p 0.001, 95% CI: [–0.87, –0.71]).

Simple slope analysis showed that at low level of CSE (W = –1 SD), mindfulness significantly negatively predicted occupational stress ($\beta$ = –0.61, SE = 0.05, t (458) = –12.20, p 0.001, 95% CI: [–0.71, –0.51]). At high levels of CSE (W = +1 SD), the negative predictive effect of mindfulness on occupational stress was stronger ($\beta$ = –0.89, SE = 0.06, t = –14.83, p 0.001, 95% CI: [–1.01, –0.77]).

Our results strongly support the Transactional Model of Stress, indicating that occupational stress serves as a pivotal mechanism linking mindfulness to mental health outcomes. The significant negative association between mindfulness and occupational stress (H1) suggests that teachers who practice mindfulness are less likely to perceive work demands as threatening or overwhelming. This finding aligns with recent intervention research conducted in similar contexts. For example, Bian and Jiang (2025) demonstrated that MBIs significantly reduced occupational stress among Chinese junior high school teachers by fostering cognitive reappraisal and emotional regulation. Similarly, Taylor et al. (2021) found that even brief mindfulness training enabled teachers to reframe stressors and reduce emotional exhaustion. The significant positive association between occupational stress and psychological distress (H2) further corroborates the well-established detrimental impact of chronic stress on teacher well-being, as recently documented by Tran et al. (2024) in their study of burnout among English as a Foreign Language (EFL) teachers. Most importantly, the significant indirect effect (H3) reveals that a substantial portion of mindfulness’s beneficial impact is mediated through a reduction in perceived occupational stress. This mechanistic pathway suggests that mindfulness training benefits teachers primarily by transforming their subjective relationship to job demands themselves, thereby mitigating their stressful impact. This finding is consistent with Querstret et al. (2020), whose meta-analysis highlighted the role of mindfulness in altering stress appraisal processes.

The most novel and significant contribution of this study lies in identifying CSE as a crucial moderator in the mindfulness-stress relationship. The significant interaction effect (H4) demonstrates that the stress-buffering effect of mindfulness is not uniform across all teachers but is significantly stronger among those with high CSE. This finding can be powerfully interpreted through the lens of Social Cognitive Theory and aligns with emerging research across professions. Mindfulness fosters awareness and cognitive space for effective primary appraisal, while CSE enhances confidence in one’s ability to implement effective coping strategies during secondary appraisal. This synergistic effect aligns with Eikenhout et al. (2022), who found among Dutch police officers that CSE strengthened the negative relationship between chronic stressors and burnout. Similarly, Lange and Kayser (2022) showed that self-efficacy mitigated the impact of work-related stress during remote work arrangements. A teacher with both high mindfulness and high CSE can calmly observe challenging situations while confidently applying problem-solving strategies, seeking appropriate support, and regulating emotions effectively, leading to substantially greater stress reduction. Conversely, as Brink et al. (2020) observed in adolescents, high mindfulness without corresponding coping confidence can increase awareness of stressors without providing the capacity to manage them, potentially diminishing the protective effects of mindfulness. Our finding thus clarifies inconsistent results in previous literature by framing CSE not merely as a parallel mediator or outcome but as a critical boundary condition that determines the effectiveness of mindfulness in reducing occupational stress.

Theoretically, this study contributes by integrating two major psychological theories to explain teacher well-being. It demonstrates that the Transactional Model’s appraisal process is shaped by both a mindful disposition (affecting primary appraisal) and self-efficacy beliefs (affecting secondary appraisal). This integrated framework provides a more comprehensive understanding of the psychological resources required to thrive in demanding environments.

The empirical support for the moderated mediation model in this study has significant implications for designing and implementing evidence-based interventions to enhance teacher well-being. The findings challenge the effectiveness of generic, one-size-fits-all mindfulness training programs common in educational settings. Instead, they advocate for the development of dual-focused interventions that are theoretically aligned and pragmatically structured to cultivate mindful awareness and build domain-specific CSE.

The rationale for this integrated approach stems from the core mechanisms identified by our analysis. While mindfulness training alone supports non-judgmental awareness and reduces reactive stress appraisals (primary appraisal), it may provide an incomplete solution. It equips educators with awareness of their stressors but may not fully provide them with the “agentic capacity” or confidence to act effectively on that awareness. CSE becomes critical at this stage, as it directly shapes the secondary appraisal process—the individual’s evaluation of available coping resources. A teacher with high mindfulness but low CSE may accurately perceive a stressful situation yet feel powerless or ineffective in managing it, thereby limiting the potential benefits of mindful observation. Consequently, interventions must be designed to address both components of this psychological sequence. Therefore, effective teacher support programs must be comprehensively designed to include not only foundational mindfulness practices (e.g., focused-attention meditation, body scans, mindful breathing) but also explicit, skill-based modules aimed at enhancing confidence in implementing concrete coping strategies.

By combining mindfulness with efficacy-building activities, interventions can achieve greater effectiveness and sustainability. School administrators and policymakers should recognize that promoting teacher well-being requires investment in comprehensive, multifaceted professional development programs rather than seeking simplistic solutions.

Several limitations should be acknowledged. First, the cross-sectional design prevents definitive causal inferences. Although our model is theoretically grounded in stress and social cognitive theories, reverse causality cannot be excluded. For example, poor mental health may reduce engagement with mindfulness practices or lower perceived CSE. Future research should employ longitudinal designs (e.g., 6-month or 1-year follow-ups) to track changes in variables over time, or experimental designs to establish causal relationships. Second, the use of self-report measures, although common in psychological research on well-being, introduces the potential for social desirability bias and CMV. Although Harman’s single-factor test and HTMT analysis suggested that these issues were not severe, they still pose a potential threat to internal validity. Future studies could incorporate multi-method assessments, such as peer ratings of teacher stress, supervisor evaluations of mindfulness-related behaviors, or physiological measures, to triangulate self-report data. Third, the sample was drawn from a single province in China using convenience sampling, which limits the generalizability of the findings to other regions, such as rural versus urban areas or eastern versus western China, and to other cultural contexts. Additionally, while the gender imbalance (84.6% female) reflects the demographic reality of Chinese junior high school teachers, it may restrict the generalizability of CSE’s moderating effect across genders. Future research should strive for geographically diverse, representative samples—ideally including multi-province data or national surveys—to strengthen external validity. Fourth, a key limitation concerns the measurement tools. The original Chinese scales used to assess “Mindfulness”, “Coping Self-Efficacy”, and “Mental Health” were not specifically developed for public junior high school teachers. For example, the mindfulness scale does not address teacher-specific mindfulness contexts, and the CSE scale lacks items tailored to the coping challenges faced by teachers. This lack of population-specific validation may introduce response bias, as teachers may find it difficult to relate scale items to their unique experiences, potentially inflating or deflating scores. While the scales showed acceptable reliability in the current sample (Cronbach’s $\alpha$$>$0.91), the potential for construct underrepresentation or misalignment with teacher experiences should be considered when interpreting results.

To address these limitations and extend the current findings, several future research directions are proposed. First, as noted, longitudinal or experimental designs are necessary to establish causality. For example, a longitudinal study could examine how changes in mindfulness predict subsequent changes in occupational stress, CSE, and mental health. An experimental design could randomize teachers to a MBI group or a control group to test whether mindfulness training reduces stress and whether this effect is stronger for teachers with high baseline CSE. Second, future studies should develop or adapt measurement scales specifically for public junior high school teachers. For instance, modifying the CSE scale to include items like “I am confident in my ability to cope with student behavioral problems” or “I can effectively manage stress from Zhongkao-related workloads” would improve construct validity and reduce response bias. Validating these population-specific scales across diverse Chinese teacher samples would further strengthen their utility. Third, expanding the range of moderators and mediators could provide a deeper understanding of the model. For example, school climate may moderate the CSE-stress relationship, with high CSE having a greater impact in positive school climates. Similarly, emotional regulation could serve as an additional mediator, as mindfulness may improve emotional regulation, which in turn reduces stress. Exploring these variables would provide a more comprehensive understanding of the pathways to teacher well-being. Fourth, testing the model in diverse teacher populations would help clarify its generalizability. For example, high school teachers face different stressors (e.g., Gaokao pressure) compared with junior high school teachers, which may affect the strength of mediation or moderation effects. Comparing results across career stages, such as early-career versus veteran teachers, could also reveal whether the model’s pathways differ with professional experience. Ultimately, translating these findings into practical applications is crucial. Future research could partner with schools to develop targeted interventions—for example, combining mindfulness training to reduce stress with CSE workshops—and evaluate their effectiveness in reducing teacher burnout and improving mental health. Such applied research would bridge the gap between theory and practice, providing tangible benefits for Chinese public school teachers.