Burnout Measurement and Work-Factor Modeling among Secondary Science Teachers: A Pilot from China with a Cross-National Expansion Plan

• In Event: Digital Dilemmas and Opportunities: Teacher Agency, Burnout, and Competence in the Tech Era

Proposal

1. Introduction
Teacher burnout, manifested in ongoing exhaustion, rising cynicism, and declining professional efficacy, undermines work quality and increases turnover intentions across occupations, including teaching (Maslach & Leiter, 2016). Recent Chinese surveys report high prevalence, with 61 percent of teachers scoring in the emotional exhaustion range of clinical concern (He et al., 2025). Science teachers face distinct risks that include laboratory safety, fast-evolving curricula, and high-stakes examinations.
This proposal presents a pilot study that (a) validates the measurement of secondary-science teacher burnout using the Maslach Burnout Inventory-General Survey (MBI-GS; Maslach et al., 1997) with Rasch modeling and (b) models associations between burnout and five work factors derived from the Job Demands-Resources (JD-R) model and Conservation of Resources (COR) theory: (1) Workload, (2) Student-related Challenges, (3) Collegiality, (4) Administrative Support, (5) Working Environment (Demerouti et al., 2001; Hobfoll, 1989). We then articulate a further cross-national plan that extends this measurement-plus-modeling pipeline to larger and contrasting systems. The work directly addresses teacher preparation, induction, and school-level support, which motivates the proposed cross-national design.
2. Pilot Context and Methods
Data were collected from 78 secondary science in-service teachers across China, including 26 chemistry, 23 physics, 20 biology, and 9 integrated science teachers. The MBI-GS (15 items) was translated and back-translated following best-practice guidelines (Beaton et al., 2000). Responses used a 0-6 frequency scale (0 = never, 6 = every day). Professional Efficacy was reverse-scored so higher scores indicate greater burnout. To reduce respondent burden in a school-day setting, work-factor ratings used a three-point impact scale (1 = low, 3 = high).
Measurement. The internal structure of the MBI-GS was analyzed using Rasch modeling via BASS (BEAR Assessment System Software; Wilson & Sloane, 2000), a platform for psychometric validation (Wilson, 2005, 2023). We examined item fit (target infit MNSQ ≈ 0.70-1.30; Bond & Fox, 2013), category ordering, person-item targeting, and information functions.
Modeling. We estimated four linear models predicting (1) overall burnout and the subscales (2) Exhaustion, (3) Cynicism, and (4) Professional Efficacy from the five work factors, reporting unstandardized/standardized effects and model fit at α = .05. The pilot was approved as minimal-risk survey research.
3. Preliminary Findings from the Pilot
3.1 Teacher Burnout
Measurement validity. All items showed acceptable Rasch fit (Exhaustion: 0.75-0.79, Cynicism: ≈0.85-0.90, Professional Efficacy: 1.06-1.17; No misfit |t| > 2.0). Thresholds were ordered, and the test information function peaked in the moderate-burnout range, where early identification and support are most actionable. The Wright map indicated good person-item targeting, with limited floor or ceiling effects. These results support the use of Rasch-scaled person measures for subsequent comparative analyses.
Burnout profile. Teachers reported moderate overall burnout with M = 2.77. Exhaustion was most pronounced with M = 3.54. Professional Efficacy showed a moderate decline with M = 2.94. Cynicism was relatively low with M = 1.84. The pattern suggests emotional strain rather than pervasive detachment. Support should therefore prioritize strategies that reduce exhaustion.
3.2 Work-factor effects
Perceived impact ratings showed Workload and Student-related Challenges were highest (each 21 teachers rated “high”), whereas Collegiality, Administrative Support, and Working Environment were mostly “low” (66, 62, 59 teachers). Regression results showed a significant overall model, F(5, 72) = 9.88, p < .001, adjusted R² = .366. Workload (B = 9.23, p < .001) and Student-related Challenges (B = 4.61, p = .013) significantly predicted overall burnout. By subscale, for Exhaustion (model fit: adj. R² = .355), both Workload (B = 0.86, p = .002) and Student-related Challenges (B = 0.79, p = .001) were significant. Cynicism (adj. R² = .348) was similarly predicted by Workload (B = 0.69, p = .008), Student-related Challenges (B = 0.49, p = .033), and Collegiality (B = 1.30, p = .003). The positive collegiality effect suggests peer cultures may normalize pessimism or intensify social comparison. Professional Efficacy was positively associated with Working Environment (B = 0.99, p = .001) and negatively associated with Student-related Challenges (B = −0.58, p = .032), with an adjusted R² of approximately .200. Across models, Workload and Student-related Challenges were the most consistent predictors. The results capture both quantitative demands related to time and effort and emotional demands related to student behavior. Resource-rich environments aligned with higher efficacy, consistent with JD-R, COR, and school-climate research (Collie et al., 2012).
4. Future Directions: Scaling and Cross-National Comparison
Our proposal scales the pilot’s instruments, calibration procedures, and modeling strategy to larger and contrasting systems. The pilot serves as proof of concept and informs design choices in sampling, adaptation, and analytic extensions.
4.1 Sampling and sites.
Purposeful sampling of secondary-science teachers across at least two contrasting systems, targeting N ≈ 600-800 to support multi-group estimation, subgroup contrasts by subject, career stage, and school socioeconomic status, and providing adequate power for DIF detection.
4.2 Analytic extensions.
Within each site, we will conduct Rasch calibration, confirm threshold ordering and item fit, and test differential item functioning by country or region, subject, and career stage. Anchoring and scale linking will place all sites on a common logit metric, and if minor non-invariance emerges, we will apply alignment or MIMIC approaches to preserve comparability. Next, we will fit multi-group SEM to compare work-factors-to-burnout pathways across systems. We will also test moderated effects, in which Working Environment and Administrative Support buffer the impact of Workload and Student-related Challenges on Exhaustion and Cynicism.
5. Conclusion
This pilot establishes a feasible, valid, and portable measurement-modeling approach to science-teacher burnout. Rasch scaling provides sample-independent interval measures with strong targeting in the moderate-burnout range, precisely where supports can be most effective. Empirically, Workload and Student-related Challenges consistently elevate Exhaustion and Cynicism. In contrast, resource-rich working environments bolster Professional Efficacy—findings that translate directly into teacher preparation, induction, and school-embedded professional learning.
Building on this foundation, the proposed cross-national expansion will apply more robust sampling and analytics, including invariance testing, multi-group SEM, and moderated path modeling. By comparing systems, we aim to identify both universal and context-specific predictors of burnout, as well as scalable support practices that can be embedded across the teacher career span—from preservice preparation to induction and ongoing professional learning.

Authors

• Linwei Yu, The University of Hong Kong

• Yueru Li, Faculty of Education，University of Macau