The dataset used in this study was drawn from the 2018 National Survey of the Rural Teaching Workforce. This survey was launched as part of the Ministry of Education’s initiative to evaluate the Rural Teacher Support Plan. The plan aimed to strengthen the teaching workforce in rural and remote areas of China by improving recruitment, training, and working conditions etc. The 2018 survey was the most comprehensive national effort to assess rural teaching conditions during the implementation of this plan. Following the COVID-19 outbreak and growing pressure to reduce teachers’ non-teaching burdens, this type of large-scale national survey was not continued in subsequent years.

A stratified random sampling strategy was used to select participants from 35 counties across 18 provinces in China. Within each county, approximately half of the towns were chosen, and all lower secondary school teachers in those towns were invited to complete an online survey via the Wenjuanxing platform,¹ one of the leading online survey platforms in China. The survey was administered between April and July of 2018.

In total, 26,531 teachers from 351 schools were invited, and 20,858 teachers from 341 schools completed the survey. For the purpose of this study, in-field English teachers were defined as those who held a degree in English and exclusively taught English. In contrast, out-of-field English teachers were defined as those with a degree in a different subject area but who were assigned to teach English only. The final sample included 2,119 in-field English teachers and 356 out-of-field English teachers.

The teacher questionnaire consisted of two parts: fixed-response items for demographics and professional details, and Likert-scale items measuring job demands, job resources, personal resources, and burnout. Demographics items included gender, age, ethnicity, and marital status, while professional background covered educational attainment, college major, years of teaching, subject assignments, and professional title. Complete demographic and professional characteristics are presented in Table 1. Descriptive statistics and Cronbach’s alpha coefficients for each scale are provided in Table 2. All scale items were adapted from validated English-language instruments, translated into Chinese, back-translated, and revised by bilingual experts to ensure cultural and linguistic appropriateness.

In this research, we employed a network analytic approach to examine the relationships among job demands, job resources, personal resources, and burnout variables among in-field and out-of-field English teachers, using R (version 4.3.2 in RStudio 2023.12.0+369). The analysis consisted of four main steps: (1) network estimation and visualization, (2) centrality indices calculation, (3) network accuracy and stability evaluation, and (4) network comparison across the two teacher groups.

We first modeled the network structure using a Gaussian Graphical Model (GGM) (50). To control for the potential impact of demographic and professional covariates (age, sex, ethnicity, educational background, marital status, and professional title) on the 14 key variables, each variable was first regressed on all covariates, and the resulting residuals were used for network estimation. To correct for data non-normality, we applied a non-paranormal transformation using the huge package (51). The GGM was then estimated with the bootnet package (50), applying the graphical least absolute shrinkage and selection operator (GLASSO) (52) for regularization and selecting the optimal model using the extended Bayesian Information Criterion (EBIC) (53). The resulting networks were visualized using the qgraph package (54), where edge thickness indicates the strength of associations, with blue edges representing positive relationships and red edges representing negative ones.

We then computed node centrality using the expected influence (EI) metric, which represents the sum of signed edge weights and captures a node’s cumulative influence on its directly connected neighbors (41). EI was selected over strength because many edges were negative in both networks. To examine how job demands and resources connect to burnout symptoms, we also calculated bridge expected influence, which quantifies the standardized sum of a node’s edges directed to the nodes of another community (42). Node and bridge centrality indices were computed using the qgraph (50) and networktools (55) packages.

To assess the accuracy and stability of the estimated networks, we conducted a series of non-parametric bootstrapping procedures using the bootnet package (50). First, we evaluated the precision of edge weights through 3,000 iterations of non-parametric bootstrapping to calculate 95% confidence intervals (CIs) for each edge. Second, we assessed the stability of centrality indices using case-dropping bootstrapping. This procedure involves repeatedly re-estimating centrality metrics after randomly removing increasing proportions of the sample. The correlation stability coefficient (CS coefficient) quantifies the maximum proportion of the sample that can be dropped while still retaining a correlation at least 0.7 between the original and bootstrap subset centrality indices (50). CS-coefficients greater than 0.25, and ideally exceeding 0.5, indicate robust estimates (50).

Finally, to compare the networks of in-field and out-of-field English teachers, we conducted network comparison tests (NCTs) using the NetworkComparisonTest package to explore differences in network structure and global strength (56). In addition, we tested for differences in individual edge weights between the two networks using 3,000 permutations and Holm-Bonferroni correction to adjust for multiple comparisons. This allowed us to assess whether any specific pairwise associations between variables differed significantly between the two groups.

Table 1 presents the means and standard deviations for all measured variables. Higher scores reflect greater levels of the corresponding constructs. Figures 1, 2 present the zero-order correlation matrices for in-field and out-of-field English teachers, respectively, showing distinct patterns of association among job demands, job resources, personal resources, and burnout dimensions.

Our network analysis revealed that emotional exhaustion emerged as the most central symptom in both the in-field and out-of-field teacher burnout networks, based on EI. Traditional burnout literature has long emphasized emotional exhaustion and depersonalization as the core components of burnout (60–62), and our findings confirmed the centrality of emotional exhaustion. Although conceptually related to the other dimensions, emotional exhaustion often represents the entry point into the burnout process, with the potential to trigger both depersonalization and diminished personal accomplishment (18). In the rural Chinese context, where teachers experience consistently high workload stress (58), emotional exhaustion likely plays an influential role in intensifying burnout through its strong interconnections within the burnout cluster.

Beyond burnout symptoms, our network analysis identified workload stress as a highly central variable in the JD-R cluster, particularly for out-of-field teachers. In the out-of-field network, workload stress was the second most central node after emotional exhaustion, and it was strongly connected to emotional exhaustion. This aligns with the core proposition of JD-R model that heavy work demands deplete energy (17). In contrast, workload stress showed negligible associations with depersonalization and diminished personal accomplishment, suggesting that workload stress is primarily associated with emotional exhaustion rather than with other burnout symptoms. Among out-of-field teachers, who must often teach subjects beyond their training, workload stress is likely intensified by a lack of familiarity with course content and preparation demands (63). Although workload stress also played a meaningful role for in-field teachers, its centrality was comparatively lower.

Bridge centrality analysis revealed that workload stress had the highest bridge EI in both the in-field and out-of-field teacher networks, indicating its key role in connecting the JD-R cluster to the burnout cluster. Specifically, workload stress showed a strong positive edge connection with emotional exhaustion, indicating that excessive workload is strongly associated with emotional strain, which may subsequently trigger other symptoms of burnout.

Other job demand variables, such as student management stress and average teaching hours per week, also demonstrated high bridge EI values. These findings align with the central idea of the JD-R model that high job demands increase the risk of burnout (17). In rural China, teachers often teach multiple subjects and grade levels, especially in deep poverty areas (64). Additionally, they undertake numerous administrative and non-instructional duties, including supervising boarding students and conducting poverty-related documentation (34, 65), which further exacerbate their workload.

On the protective side, job satisfaction served as the most influential negative bridge, suggesting that it is associated with lower levels of burnout symptoms. Specifically, job satisfaction exhibited strong negative edges with depersonalization, suggesting that higher satisfaction is associated with less negative attitudes toward teaching. This aligns with the JD-R model’s motivational process, where resources like satisfaction foster engagement and buffer against burnout (17, 66). For instance, rural teachers who perceive their work as meaningful, despite heavy workloads, may be better protected from burnout (65). However, limited professional development opportunities, as seen in deep poverty areas and western China, may constrain its effectiveness by reducing teachers’ sense of growth and autonomy (34, 64). Aside from job satisfaction, other job resource variables (e.g., organizational justice, school resources, teacher-student relationship) exhibited near-zero bridge EI values and only weak connections with the burnout cluster, suggesting limited protective associations with burnout in this context.

Although in-field and out-of-field teachers differed in their training backgrounds, the network comparisons showed no statistically significant differences in structure or global strength (56). Nevertheless, small variations, such as the relatively higher bridge EI of workload stress in the out-of-field group, suggest that out-of-field teaching may influence burnout primarily by amplifying workload stress, which strongly connects to emotional exhaustion. This finding contrasts with earlier studies that identify out-of-field teaching as a direct stressor (63), where insufficient subject content is assumed to increase burnout risk. Within rural Chinese schools, however, heavy workloads, multi-subject teaching, and administrative duties are structural norms for both in-field and out-of-field teachers (64), potentially masking any direct effect of out-of-field teaching. Thus, out-of-field teaching may exacerbate existing demands, rather than introduce new stressors, highlighting the need for context-sensitive and group-specific interventions.

The present findings have several important implications for educational policy and practice. First, network analysis extends the JD-R model by using bridge EI to identify variables that can activate or deactivate burnout, allowing researchers and practitioners to prioritize interventions at these nodes with higher impact, rather than addressing every demand or resource indiscriminately. In our study, workload stress emerges as the primary bridge connecting the JD-R cluster to the burnout cluster, highlighting the importance of workload reduction as the priority for mitigating emotional exhaustion. Additionally, job satisfaction uniquely exhibited a substantial negative bridge influence, being associated with lower depersonalization. To foster genuine value in their roles, interventions could focus on enhancing professional identity and recognition, such as providing more opportunities for professional development tailored to rural contexts.

Second, it is important to recognize that out-of-field teaching amplifies existing demands rather than introducing new ones. Our findings suggest that these effects may operate indirectly, by increasing workload-related stress, which in turn is associated with emotional exhaustion. Consequently, policies that focus solely on retraining or certification may be insufficient if they do not address the broader issue of excessive workload. Therefore, effective interventions should combine upskilling initiatives (e.g., subject-specific training for out-of-field teachers) with strategies to reduce workload, such as hiring additional staff for administrative tasks or redistributing responsibilities within schools.

Third, job satisfaction emerged as a key protective factor against burnout, particularly in reducing depersonalization. Although rural policy efforts in China have improved teacher retention through financial incentives and promotion policies, these “hard” measures may not fully address the emotional and psychological dimensions of burnout. Long-term stress without corresponding psychological fulfillment can make material rewards insufficient. To promote job satisfaction more effectively, it is crucial to introduce “soft” support systems. Initiatives such as local honors (e.g., “Rural Education Contribution Award”) and programs that foster peer support and collaborative school cultures can reinforce teachers’ sense of belonging and professional identity, providing emotional resilience against burnout. Furthermore, the broader societal promotion of the “spirit of the educator” in China emphasizes both public respect for the teaching profession and the encouragement for teachers to recognize the intrinsic value of teaching.

Finally, professional characteristics such as work experience and professional titles, which reflect career development, warrant further exploration in relation to burnout, particularly in the Chinese context. For instance, Wu et al. (67) found higher emotional exhaustion among elementary school teachers in their fifth to tenth year of service, while middle school teachers showed greater burnout in their eleventh and twentieth year. Similarly, Sang et al. (68) found that teachers with over 20 years of experience reported lower levels of emotional exhaustion compared to those with 6–10 years, suggesting a non-linear relationship between experience and burnout. While we controlled for these variables, future research could examine them as moderators to explore how career trajectories interact with job demands and resources in shaping burnout patterns.

Several limitations of this study must be acknowledged. First the cross-sectional nature of this study limits our ability to make causal inferences about the relationships in the burnout network. Future longitudinal research with multiple waves of data collection is needed to better understand directionality and the temporal development of these networks.

Second, our study is situated within rural China and the contextual factors that shape burnout and out-of-field teaching may differ substantially from those in urban areas or other countries. In many Chinese rural schools, the teacher workforce is stabilized through the tenure-track system, which greatly reduces turnover. However, this long-term job security may not mitigate emotional exhaustion and may even lead to sustained burnout (32). Moreover, rural schools often exhibit a “dumbbell-shaped” teacher demographic pattern, with a concentration of older teachers close to their retirement and younger teachers at the start of their careers (33). These gender dynamics can influence how teaching assignments, including out-of-field roles are distributed, and may exacerbate burnout in specific subgroups (e.g., young teachers with limited support or senior teachers assigned new responsibilities). These structural realities are distinct from many Western or urban settings, where higher turnover rates (69) may alter the burnout and out-of-field teaching relationship. Therefore, caution is warranted when generalizing these findings to other national or international contexts.

Third, the reliance on self-reported data may introduce potential biases, such as social desirability and recall bias, which could affect the validity of our findings. Social desirability bias could lead teachers to overstate their confidence or competence in teaching English, potentially underestimating the challenges associated with out-of-field teaching in rural schools. Similarly, recall bias has an impact on the accuracy of reported workload or perceived stress. These biases may either reduce or increase the associations between certain variables, therefore influencing the strength of the estimated edges and the identification of central and bridge nodes in the network. Although anonymity and standardized instruments were employed to address these concerns, caution is warranted when interpreting self-reported data in burnout research.

Fourth, our network analysis included only a subset of job demands, job resources, and personal resources. Other potential factors, such as teacher resilience or coping strategies, may also have an impact on the burnout network. Future studies should incorporate a broader range of variables into the network to better understand how burnout is exacerbated or mitigated.

Fifth, while the data provide valuable insights into the structure of teacher burnout, we acknowledge that China’s 2020–2024 curriculum reform and the COVID-19 pandemic (2020–2022) may have affected teachers’ working conditions and stress levels. These changes could influence burnout dynamics in ways not reflected in the 2018 data. Therefore, we advocate for new nationwide surveys to assess the current state of teacher burnout amid these changing educational and social conditions.

Finally, this study focused solely on Maslach’s negative burnout dimensions, emotional exhaustion, depersonalization, and reduced personal accomplishment, excluding positive trajectories such as increased work engagement and enhanced flourishing. Although this is consistent with a large body of prior research, this limits a comprehensive view of teacher well-being. Future studies could integrate positive indicators to provide a more holistic understanding of the dynamic processes underlying teacher burnout and recovery, particularly when using methods like network analysis to model reciprocal changes in burnout and recovery.