How Strongly Is Teachers’ Burnout Related to Teacher Absenteeism,Teacher–Student Interactions,and Student Motivation and Achievement: A Meta-Analysis and Systematic Review

The Concept of Burnout

Burnout is defined as a work-related syndrome with three symptoms, namely, emotional exhaustion, depersonalization, and reduced personal accomplishment (Maslach et al., 2001; World Health Organization [WHO], 2021). The key symptom, emotional exhaustion, describes feelings of emotional fatigue, a depletion of individual resources, and overexertion; it forms the individual stress component of burnout and is most frequently investigated as the core symptom of burnout (Maslach et al., 2001). However, burnout encompasses more than a lack of sufficient energy (Maslach et al., 2018; Schaufeli et al., 2020). The second symptom, depersonalization, is defined as a callous, aloof, or indifferent attitude and behavior toward recipients on the job and reflects the interpersonal dimension of burnout because it represents an emotional withdrawal from recipients (Demerouti et al., 2001). In the context of the teaching profession, depersonalization is typically investigated as a detached attitude toward students (Maslach et al., 2018). The third symptom, reduced personal accomplishment, refers to the feeling of incompetence, reduced efficacy, and insufficient ability to succeed in the job (Maslach & Leiter, 1999; Maslach et al., 2001). Although burnout is closely related to work-related depression and substantially overlaps with depressive symptomatology (Bianchi & Schonfeld, 2025), it often has been conceptualized as a distinct psychological syndrome in previous research comprising emotional exhaustion, depersonalization, and reduced personal accomplishment. These three symptoms, which are illustrated on the left-hand side of the heuristic working model in Figure 1, also have been identified in empirical work testing the structure of burnout (Aboagye et al., 2018; Byrne, 1994) and were adopted in the ICD-11 (11th revision of the International classification of diseases; WHO, 2021). In line with this, other definitions also emphasize the multidimensionality of the burnout syndrome but additionally incorporate physical and cognitive symptoms of exhaustion (Demerouti et al., 2001; Schaufeli et al., 2020).

OPEN IN VIEWER

Figure 1.

Heuristic working model.

Although burnout is classified as the presence of all three of these symptoms in the ICD-11 (WHO, 2021), it is important to consider the different severities of emotional exhaustion, depersonalization, and reduced personal accomplishment (Maslach et al., 2001; Nadon et al., 2022). Current research has drawn on the underlying assumption that individuals can experience different patterns of the three burnout symptoms, which may develop differently over time (Herman et al., 2018; Mäkikangas & Kinnunen, 2016).

Furthermore, research on antecedents of burnout has suggested differential associations, with emotional exhaustion being most strongly related to general indicators of high job demands (e.g., time pressure) and depersonalization resulting more from social stressors, such as highly demanding interactions and value dissonance (Byrne, 1994; Skaalvik & Skaalvik, 2017; van Droogenbroeck et al., 2014). In contrast, reduced personal accomplishment seems to be most strongly related to self-efficacy (Brouwers & Tomic, 2000). Accordingly, differential associations with theoretically postulated consequences of the individual burnout symptoms are also feasible. Therefore, our meta-analyses considered all three burnout symptoms differentially.

Theoretical Consequences of Teacher Burnout: Constructs and Theoretical Models

Various theoretical approaches have provided explanations for how burnout may influence workplace behavior through different psychological pathways; these can be motivational, social, cognitive, or physical (Fredrickson, 2001; Hobfoll, 1989; Lazarus & Folkman, 1984; Maslach & Leiter, 1999). Numerous studies have documented these relations with individual behavior in terms of reduced professional performance and high rates of absenteeism across different occupational fields (Bakker & Costa, 2014; Salvagioni et al., 2017; Taris, 2006). In light of this, examining the strength of the association between burnout and professional consequences among teachers is especially important. In the following, we first define what we mean by the professional consequences. Subsequently, we elaborate on the theoretically and empirically suggested associations between burnout and each theoretical consequence for the teaching profession (see Figure 1).

Literature Search

To identify relevant studies on the relationship between teacher burnout and its professional correlates, a systematic literature search was carried out in November 2020 and updated in October 2021 and April 2025 using several strategies, which are illustrated in Figure 2. We combined an electronic database search (i.e., PsycINFO, Web of Science, and OpenGrey) with additional manual search strategies to identify both published and unpublished work on teacher burnout. Because we were only interested in teachers currently teaching at preschools, elementary schools, and secondary schools, we further narrowed the search to these types of schools. The exact search terms and the respective search strategies that were employed in the different databases are described in detail in the Online Supplement (Supplementary Table S1) in the online version of the journal.

OPEN IN VIEWER

Figure 2.

PRISMA (Preferred Reporting Items for Systematic Reviews and Meta Analyses) flow diagram for the literature search process.

To generate a comprehensive literature search, we only combined the thesaurus search terms for teacher with the thesaurus terms for burnout in PsycINFO, leaving the correlate of interest out in the first step. This search identified 3,361 results. For the electronic search using Web of Science, we confined the search and included the correlate of interest. This resulted in 635 additional records after removing duplicates. To search specifically for unpublished work, theses, and dissertations, we conducted an additional search in OpenGrey, which revealed 339 additional records.

In the second step, we conducted both a backward search based on the reference lists of the identified studies and a citation search. To make sure that we had not missed any relevant studies, we reviewed previous meta-analyses that either comprised some of our variables or concerned similar relationships observed in mixed occupational groups. A further 165 records were identified with these additional strategies.

Inclusion and Exclusion Criteria

Records identified by the aforementioned search strategies were first screened by title and abstract, followed by a full-text coding of eligible studies. The prescreening was conducted by the first and second authors, with the support of three student research assistants, and was based on the following criteria. We included studies for the subsequent full-text coding that (a) assessed teacher burnout; (b) reported on professional correlates in terms of absenteeism, teacher–student interactions, and student motivation and achievement; (c) investigated a sample of teachers currently teaching at pre-, elementary, and secondary schools; (4) provided quantitative data; and (5) were written in a language with a Latin alphabet. This resulted in a decision to either include studies that met (coded as 1) or excluded studies that did not meet all the inclusion criteria (coded as 0). Before prescreening, all coders completed a training session based on the coding manual. Intercoder agreement regarding the decision to include or exclude records for the subsequent full-text coding was established based on the double coding of 35% of the identified records (k = 1,377) with κ = .90. On the basis of the elucidated criteria and as illustrated in the PRISMA flow diagram (Page et al., 2021; Pigott & Polanin, 2020) in Figure 2, a total of 3,761 records were excluded after screening the titles and abstracts, leaving 339 studies for further coding. In the second coding step, we decided to exclude or include studies in the systematic review based on the following refined criteria.

Coding

The first and second authors conducted the full-text coding and extracted the following information from primary studies. First, we retrieved the main study characteristics (e.g., the names of the authors, the publication year, and the type of publication). Second, regarding the sample characteristics, we recorded the sample size, the percentage of female teachers, the mean teaching experience, the grade level taught (i.e., elementary, secondary, or mixed), the school type (i.e., private, public, or mixed), and the country. Third, we captured the following information for teacher burnout: the burnout symptom examined, the instrument used, and the reliability. Regarding the correlates, we coded the investigated construct, the reliability, and information on whether teacher self-reports, student reports, observational measures, or standardized tests were applied. Finally, we retrieved the effect size and coded the following additional information: the level of significance and information on whether manifest or latent correlations were retrieved.

The agreement between coders was assessed by double coding 50% (n = 43) of the 86 studies included. The first author re-coded studies previously coded by the second author, and a trained student research assistant re-coded studies originally coded by the first author. Regarding the categorization of burnout symptoms, the interrater reliability was κ = .88, and regarding the classification of the correlates into the associated coding categories, the interrater reliability was κ = .80. In addition to the double coding, studies for which uncertainties arose during the coding process or for which the coding decisions were not entirely clear were discussed by the research team to ensure that no study was incorrectly coded. The data are published open access at PsychArchive (https://doi.org/10.23668/psycharchives.21918).

Data Processing

We conducted several preliminary steps before aggregating the effect sizes meta-analytically. First, we converted effect sizes other than correlation coefficients (Borenstein et al., 2009; Thalheimer & Cook, 2002). Second, we re-coded effect sizes for the sake of interpretability so that negative correlations reflected findings where higher levels of burnout were associated with less favorable professional correlates. This was necessary because primary studies differed with regard to whether they reported the effect sizes for positively or negatively scored constructs (e.g., personal accomplishment vs reduced personal accomplishment; student misbehavior vs lack of student misbehavior).

Third, although most studies reported manifest effect sizes, four studies reported only latent correlations. To increase the comparability of the effect sizes, we estimated the uncorrected correlation for these four studies, following the formula of Hunter and Schmidt (2004, p. 96). Finally, we considered the following three types of dependencies that emerged in the data of the primary studies: (a) dependent subsamples examining the same relationship, (b) multiple effect sizes for the same association due to the use of multiple subscales of the respective construct, and (c) self-reports and other reports of the observed construct being used within one sample. Ignoring the dependency of multiple effect sizes would entail the risk of a biased estimation of the summary effect in the sense of an overestimation of the precision of the mean relationship if the dependent effect sizes were positively correlated (Borenstein et al., 2009; Hunter & Schmidt, 2004; Rodgers & Pustejovsky, 2019). To address the problem of dependent samples, we did not include multiple reports of the same relationship that were based on a dependent sample. Therefore, we retained the respective effect sizes from the most comprehensive study and excluded dependent subsamples. To account for dependent effect sizes due to the use of multiple subscales, we aggregated the correlation coefficients that referred to the same construct into a single correlation for the (composite) construct by using the procedures recommended by Hunter and Schmidt (2004, pp. 435–439). Likewise, we computed the reliability of the composite correlation using the Spearman–Brown formula (Hunter & Schmidt, 2004, p. 438). This approach retains as much information as possible and avoids the dependence on effect sizes caused by the use of multiple subscales in the same sample (Borenstein et al., 2009). Because we aimed to investigate potential differences in effect sizes arising from rater perspective, self-reports and other reports were not combined. This applied to four studies. Although the error in the aggregated effect sizes is reasonably small if the number of effect sizes that are based on the same sample is minor compared with the total number of effect sizes included in the synthesis (Hunter & Schmidt, 2004), we applied robust variance estimation to address the dependencies in sampling errors within these studies (Hedges et al., 2010).

Analyses

To investigate the research question, we conducted a univariate meta-analysis for each combination of the three burnout symptoms and the correlates. Against the background of the expected variability in the true effect sizes between studies in addition to heterogeneity caused by sampling error, we specified random-effects models throughout (Borenstein et al., 2009; Raudenbush, 2009). To this end, we used the Hunter–Schmidt estimator for the between-study variance in all models and analyzed the untransformed correlation coefficients weighted by sample size using the R package metafor (Field, 2001; Viechtbauer, 2010). We performed robust variance estimation to consider the dependencies caused by multiple effect sizes within studies that were based on self-reports and other reports (Hedges et al., 2010). Additionally, we conducted psychometric meta-analyses with the R package psychmeta to further adjust the results for the attenuation in effect sizes caused by measurement error (Dahlke & Wiernik, 2021). For this purpose, we corrected the correlations for measurement error by means of artifact distribution to obtain more precise estimates of the effect sizes (Borenstein et al., 2009; Hunter & Schmidt, 2004). Detailed results for the psychometric meta-analyses are reported in Supplementary Table S5 in the online version of the journal, which resulted in effect sizes that followed roughly the same pattern but tended to be larger in magnitude. We did not conduct a meta-analysis for associations with fewer than five primary studies (Higgins et al., 2009), but we did systematically review the respective studies.

The total amount of heterogeneity was estimated based on τ². Additionally, we considered the Q test for heterogeneity and I² for the proportion of variance between studies. To investigate the amount of heterogeneity in the effect sizes that could be attributed to study and sample characteristics, we estimated meta-regression models with the following moderators: publication year, type of publication, teacher gender, years of teaching experience, grade level (i.e., elementary, secondary, or mixed), and the operationalization of the correlates (i.e., self-reports or other). Categorical moderators were dummy coded for these analyses. Additional subgroup analyses were conducted to further explore the differences between teacher self-reports and external reports for teacher–student interactions.

Meta-analyses are susceptible to publication bias, which can lead to distorted estimates of the average effect size (Borenstein et al., 2009; Carter et al., 2019). For this reason, we investigated publication bias both visually, using funnel plots, and statistically, using Egger’s tests and by reporting precision-adjusted effect-size estimates (Egger et al., 1997). For an overview of a similar methodologic approach, see Wartenberg et al. (2023).

Study Characteristics

The 86 studies reported results for 90 independent samples, representing a total sample size of 38,457 teachers. The studies included were published between 1986 and 2025. However, most studies were conducted after 2015 (66%), indicating a significant increase in research on the correlates of teacher burnout in the past 10 years. Of the 86 studies included, 80 represented published journal articles, whereas six studies referred to unpublished work, including dissertations (k = 5) and one master’s thesis. Most of the studies included were cross-sectional, with only 23 studies being longitudinal, of which 15 represented intervention studies. Regarding the country of origin, 26 of the studies investigated U.S. samples, 17 were conducted in Germany, six in Norway, five in the Netherlands, and four in Belgium. The remaining studies included other European countries, Ghana, Israel, Canada, China, Taiwan, and the Philippines. The sample size ranged from five to 4,567 teachers. On average, 74.72% of participants per sample were female teachers (SD = 19.13), with an overall mean teaching experience of 14.19 years (SD = 4.98) across samples.

Most of the samples included measures of emotional exhaustion (k = 86), confirming the tendency in research to investigate emotional exhaustion as the central symptom of burnout. Nevertheless, 37 samples reported results for depersonalization and reduced personal accomplishment, respectively. Almost all studies included applied the Maslach Burnout Inventory (Maslach et al., 2018). Only 10 studies adopted other instruments, including, among others, the Emotional Exhaustion Subscale of the Friedman Burnout Scale (Friedman & Farber, 1992), the Tedium Scale (Pines & Kafry, 1978), and the Burnout Clinical Subtype Questionnaire (Montero-Marín & García-Campayo, 2010).

Overall Effects for the Relationship Between Teacher Burnout Symptoms and Professional Correlates

In this section we describe the meta-analytic findings on the relationship between the three burnout symptoms and the respective correlates. In line with Cohen’s (1988) recommendations, we interpreted correlation coefficients ≥.10 as small, ≥.30 as moderate, and ≥.50 as large. Subsequently, we supplemented the meta-analytic findings with those of a systematic review of the primary study results from which a calculation of overall effect sizes was not possible. This concerns effect sizes where fewer than five studies were available, providing an insufficient empirical basis for meta-analytic methods (Higgins et al., 2009). The overall effect sizes for the remaining correlations are reported in Table 1, together with information about the sample size and heterogeneity and the effect sizes that were adjusted for measurement error. Forest plots illustrating the effects and the 95% confidence intervals for each study are depicted in Supplementary Figures S1–S17 in the online version of the journal.

OPEN IN VIEWER

Table 1

Association Between Teachers’ Burnout and Professional Correlates

Factor	k	N	r	95% CI r		Heterogeneity						95% CI r adj
				Lower	Upper	τ 2	I 2	Q H	df	p H	r adj	Lower	Upper
Emotional exhaustion
Absenteeism	10	3,246	.18	0.08	0.29	0.017	84.24	67.31	(9)	<.001	.20	−0.08	0.31
Emotional support	48	17,540	−.13	−0.18	−0.09	0.010	77.39	225.08	(47)	<.001	−.17	−0.21	−0.13
Classroom management	41	19,820	−.23	−0.28	−0.17	0.012	86.14	303.95	(40)	<.001	−.27	−0.32	−0.23
Instructional support	24	5,217	−.13	−0.16	−0.10	0.001	23.42	31.6	(23)	.109	−.17	−0.22	−0.12
General interactions	16	3,792	−.12	−0.17	−0.07	0.003	41.38	28.16	(15)	.021	−.15	−0.21	−0.09
Student motivation	10	6,108	−.23	−0.28	−0.17	0.005	77.57	46.37	(9)	<.001	−.27	−0.34	−0.19
Student achievement	10	5,614	−.05	−0.11	0.01	0.005	70.95	36.15	(9)	<.001	−.06	−0.13	0.01
Depersonalization
Absenteeism	8	2,613	.14	0.04	0.24	0.010	75.64	35.64	(7)	<.001	.16	0.04	0.28
Emotional support	16	7,816	−.27	−0.38	−0.17	0.008	75.90	90.20	(15)	<.001	−.36	−0.43	−0.29
Classroom management	16	5,489	−.26	−0.32	−0.20	0.007	71.66	59.07	(15)	<.001	−.34	−0.41	−0.27
Instructional support	10	2,848	−.27	−0.30	−0.23	0.000	0.00	7.81	(9)	.553	−.40	−0.44	−0.36
General interactions	7	1,651	−.21	−0.33	−0.09	0.008	61.03	22.06	(6)	.001	−.28	−0.43	−0.14
Student motivation	4	2,020
Student achievement	2	42
Reduced personal accomplishment
Absenteeism	8	3,032	.08	0.03	0.12	0.001	30.42	11.73	(7)	.110	.08	0.02	0.15
Emotional support	15	2,503	−.22	−0.32	−0.12	0.022	79.59	75.52	(14)	<.001	−.32	−0.44	−0.21
Classroom management	15	4,567	−.19	−0.28	−0.10	0.014	81.02	84.07	(14)	<.001	−.24	−0.34	−0.14
Instructional support	11	2,480	−.30	−0.43	−0.18	0.027	87.29	91.68	(10)	<.001	−.43	−0.63	−0.24
General interactions	9	852	−.21	−0.37	−0.05	0.031	74.67	37.77	(8)	<.001	−.28	−0.52	−0.04
Student motivation	2	1,157
Student achievement	2	166

Note. k = number of effect sizes; N = total sample size; r = pooled effect size (correlation); 95% CI _r  = 95% confidence interval of the mean effect size; τ² = between-study variance; I² = ratio of true heterogeneity to total observed variation; Q_H = Q test of heterogeneity; df = degrees of freedom for Q test of heterogeneity; p_H = p value for Q test; r_adj = mean effect size corrected for measurement error; 95% CI _{r adj} = confidence interval of the mean effect size corrected for measurement error.

Heterogeneity and Meta-Regression

As Table 1 illustrates, there was substantial heterogeneity between studies, which indicated that the true effects varied between studies. Considering τ², the greatest total amount of variability among the true effects was observed for emotional support (τ² = 0.022), instructional support (τ² = 0.027), and general interactions (τ² = 0.031) in the context of reduced personal accomplishment. Additionally, for most relationships, I² indicated that a moderate to large proportion of the total variability (I² = 41.38–87.29) was based on the true variance of effect sizes between studies (Higgins & Thompson, 2002). There were exceptions for the following three relationships. There was no significant heterogeneity in effect sizes between studies regarding the association between emotional exhaustion and instructional support (Q = 31.60; p = .109), between depersonalization and instructional support (Q = 7.81; p = .553), or between reduced personal accomplishment and absenteeism (Q = 11.73; p = .110). We explored several study and sample characteristics to identify factors that contribute to effect-size heterogeneity and moderate the relationships between teacher burnout and professional correlates. In the following subsections, we describe the results for one study characteristic that proved to be a significant moderator for several relationships, namely the rater perspective (Table 2). The remaining study and sample characteristics that we analyzed are listed in Supplementary Table S6 in the online version of the journal. In general, the findings from the meta-regression should be interpreted cautiously given that some analyses relied on a limited number of effect sizes.

OPEN IN VIEWER

Table 2

Results from Meta-Regression Analysis Examining Rater Perspective as a Moderator

Professional correlates	Moderator	Emotional exhaustion						Depersonalization						Reduced personal accomplishment
		β	p	k	Q M (df)	p M	τ 2	β	p	k	Q M (df)	p M	τ 2	β	p	k	Q M (df)	p M	τ 2
Absenteeism	Self-report	−.09	.349	10	0.88 (1)	.349	.015	−.16	.035	8	4.46 (1)	.035	.007	−.05	.279	8	1.17 (1)	.279	.001
Emotional support	Self-report	−.07	.036	48	4.38 (1)	.036	.008	.03	.786	16	0.07 (1)	.786	.008	−.13	.126	15	2.34 (1)	.126	.016
Classroom management	Self-report	−.14	.000	41	14.37 (1)	<.001	.008	−.54	.536	16	0.38 (1)	.536	.007	−.15	.040	15	4.24 (1)	.040	.010
Instructional support	Self-report	−.04	.247	24	1.34 (1)	.247	.001	−.08	.554	10	0.35 (1)	.554	.000	−.37	<.001	11	84.22 (1)	<.001	.000
General interactions	Self-report	−.11	.031	16	4.67 (1)	.031	.002	−.07	.769	7	0.09 (1)	.769	.008	−.15	.258	9	1.28 (1)	.258	.021
Student motivation	Self-report	−.15	.000	10	12.25 (1)	<.001	.002
Student achievement	Achievement measure	.19	.139	10	2.19 (1)	.139	.004

Note. β = estimated regression coefficient; p = p value for meta-regression; k = number of effect sizes; Q_M = Q test of moderator; df = degrees of freedom for Q test of heterogeneity; p_M = p value for Q test of moderator; τ² = between-study variance. Statistically significant coefficients are printed in boldface.

Regarding the use of teacher self-reports to measure the correlate of interest, meta-regressions revealed that an assessment of teacher emotional support (β = −.07; p = .036) and classroom management (β = −.14; p = .000) via self-reports was related to higher negative correlations with teacher emotional exhaustion (r _ES  = −.16; 95% CI _ES  = −0.24 to −0.08; r_CM =−.26; (%% CI _CM  = −0.31 to −0.21). The same was true for the relationship between emotional exhaustion and student motivation (β = −.15; p = .003), which was stronger for teacher self-reports (r_SM = −.26; 95% CI _SM  = −0.31 to −0.21). The effect also was evident when emotional support (r_ES = −.08; 95% CI_ES = −0.12 to −0.03), classroom management (r_CM = −.10; 95% CI_CM = −0.15 to −0.05), and student motivation (r_SM = −.12; 95% CI_SM = −0.16 to −0.07) were reported by others, although the relationships were smaller. This was not true for the relationship between reduced personal accomplishment and instructional support (β = −.35; p < .001), which was not significant for external reports (r_IS = −.03; 95% CI _IS  = −0.11–0.04), whereas teacher self-reports of instructional support (r_IS = −.41; 95% CI _IS  = −0.45 to −0.37) revealed significantly larger effect sizes. We did not find any significant differences in the effect sizes for the other relationships between burnout symptoms and types of teacher–student interactions depending on the rater perspective, although, overall, the correlations tended to be smaller for external reports. For a detailed presentation of the subgroup analyses and the separate effect sizes for self-reports and other reports, see Supplementary Figures S21–S31 in the online version of the journal. Considering student achievement, there was no difference in the size of the relationship between teacher emotional exhaustion and student achievement depending on whether standardized tests or teacher evaluations were used to assess student achievement (β = .19; p = .139).

Longitudinal Associations Between Teacher Burnout and Professional Correlates

Of the 31 longitudinal studies that we identified in the literature search, 13 were intervention studies. Of these intervention studies, only eight reported longitudinal effect sizes for the control group or controlled for possible intervention effects. Most of the remaining longitudinal studies did not focus on the consequences of burnout. As a result, either no longitudinal correlation was reported (e.g., Taxer et al., 2019) or burnout was recorded as an outcome at a later measurement point (e.g., Dicke et al., 2014). However, 11 studies indicated longitudinal correlations between burnout dimensions at a prior measurement point and at least one professional outcome at the subsequent measurement point. The vast amount of longitudinal studies investigated dimensions of teacher–student interactions. Supplementary Table S7 in the online version of the journal provides an additional overview of the 20 studies that reported effect sizes for the longitudinal association. Overall, 13 studies implemented two measurement points with a mean time interval of 223.82 days (SD = 140.61; range = 122–933), two studies involved three measurement points with a mean time interval of 91.5 days (SD = 28.5; range = 63–120), and one study implemented four survey dates with 4 months between the first three measurement points and a fourth measurement point after 7 years.

Taking a closer look at the 10 studies that explicitly investigated the consequences of teacher burnout, a mixed picture emerges. For instance, Ansari et al. (2020) examined the relationship between teacher emotional exhaustion at the beginning of the school year and the quality of teacher–student interactions throughout the year in a sample of 117 preschool teachers. Their findings showed a significant negative association between emotional exhaustion and the quality of emotional support, classroom management, and instructional support. In line with this, Benita et al. (2018) also reported that teachers who had experienced higher burnout (i.e., depersonalization) early in the school year saw more classroom disturbances at the end of the school year, a perception shared by their students. Seiz et al. (2015) studied 205 secondary school teachers and found that higher pedagogic/psychological knowledge improved classroom management only when teachers had low emotional exhaustion, enabling them to apply their knowledge effectively.

Although these studies suggest a longitudinal relationship, they did not account for prior instructional quality and do not allow any conclusions to be drawn about the extent to which teacher burnout explains changes in the consequences considered. However, Shen et al. (2015) found that teacher emotional exhaustion at the beginning of the school year was associated with a reduction in ninth and tenth grade students’ perceptions of teacher emotional support and their autonomous motivation by the end of the first school term, based on a sample of 33 teachers and 1,452 students. In line with this, Elzie and Downer (2025) described a positive association between teacher emotional exhaustion (N = 101) at the beginning of the school year and observed student disruptive behavior at the end of the term when controlling for teacher levels of depersonalization and personal accomplishment. This indicates that exhausted teachers are likely to exhibit less effective classroom management strategies over the course of one school term. Hoglund et al. (2015) further demonstrated that teacher emotional exhaustion at the beginning of the second school term was related to less growth in elementary students’ literacy skills throughout that term, as rated by teachers. However, using multilevel growth models with 65 high-needs elementary teachers, they found no significant links between burnout and emotional or instructional support, but unexpectedly, burnout was positively associated with classroom management over time (Hoglund et al., 2015). Similarly, Sandilos et al. (2020) and Doyle et al. (2022) did not find significant associations between teacher emotional exhaustion at the beginning of the school year and changes in the quality of teacher–student interactions (i.e., emotional support, classroom management, and instructional support) by the end of the year when controlling for the effects of professional development training. These findings are confirmed by cross-lagged panel models, which found no significant association between teacher level of emotional exhaustion and personal accomplishment and later teacher–student interactions, whereas teachers’ perceived closeness and conflict with their students predicted changes in the experience of teacher burnout across one school year when controlling for intervention effects (Alamos et al., 2022). Similarly, following a sample of 1,141 Croatian secondary school teachers across 1 school year, Burić et al. (2024) found that student levels of behavioral and emotional engagement and disaffection were more predictive of teachers’ experience of emotional exhaustion than vice versa in a four-wave cross-lagged panel model.

In summary, there is little longitudinal evidence to verify the cross-sectional associations between teacher burnout symptoms and the professional correlates that we considered. The few longitudinal studies that have investigated them, however, suggest longitudinal correlations that partly appear to confirm the effects found for absenteeism, emotional support, classroom management, and student motivation but not for instructional support and student achievement, for which the reported correlations tended to be smaller.

Publication Bias

A visual inspection of the funnel plots (see Supplementary Figures S18–S20 in the online version of the journal) revealed possible publication bias for some relationships because the distributions were not completely symmetric (Sterne et al., 2005). Based on this visual inspection, it appeared that studies with smaller samples sometimes yielded predominantly smaller or larger correlations than studies with larger samples. Additionally, we analyzed possible publication bias by conducting Egger’s regression test for funnel plot asymmetry and computing precision-adjusted effect sizes (Egger et al., 1997; Stanley & Doucouliagos, 2014). The results (Table 3) indicated no evidence of publication bias.

OPEN IN VIEWER

Table 3

Results of Egger’s Test for Symmetry of Funnel Plots

	k	β	p	r PB adj	95% CI r PB adj
					Lower	Upper
Emotional exhaustion
Absenteeism	10	1.65	.291	.10	−0.12	0.32
Emotional support	48	0.78	.413	−.17	−0.30	−0.03
Classroom management	41	0.39	.638	−.24	−0.34	−0.14
Instructional support	24	−0.84	.091	−.08	−0.14	−0.01
General interactions	16	0.08	.909	−.13	−0.23	−0.02
Student motivation	10	2.10	.066	−.30	−0.41	−0.19
Student achievement	10	−0.89	.367	−.02	−0.13	0.09
Depersonalization
Absenteeism	8	1.74	.167	.05	−0.13	0.23
Emotional support	16	−0.70	.586	−.25	−0.43	−0.07
Classroom management	16	0.77	.346	−.30	−0.41	−0.19
Instructional support	10	−0.21	.764	−.26	−0.33	−0.18
General interactions	7	−1.67	.147	−.12	−0.31	0.07
Student motivation	4	0.84	.529	−.31	−0.47	−0.16
Student achievement	2
Personal accomplishment
Absenteeism	8	−0.71	.527	.11	−0.01	0.23
Emotional support	15	−0.24	.844	−.20	−0.42	−0.02
Classroom management	15	1.27	.267	−.25	−0.42	−0.08
Instructional support	11	1.51	.345	−.39	−0.65	−0.13
General interactions	9	0.77	.564	−.28	−0.62	0.05
Student motivation	2
Student achievement	2

Note. k = number of effect sizes; β = estimated regression coefficient of Egger’s test; p = p value for Egger’s test; r_{PB adj} = precision-adjusted effect-size estimates (correlation); 95% CI _{r PB adj}  = 95% confidence interval of the precision-adjusted effect size.

Magnitude and Meaning of Teacher Burnout

Our meta-analyses broadly support the idea that teacher burnout symptoms are associated with higher absenteeism, a lower quality of teacher–student interactions, and reduced student motivation and achievement. These results are in line with other research syntheses focusing on teacher samples, which reported comparable results for the association between burnout, turnover intentions, student misbehavior (as a specific indicator of classroom management), and student educational development, although turnover intentions seem to be more strongly related to teacher burnout than to absenteeism (Aloe et al., 2014; Madigan & Kim, 2021a, 2021b). Interestingly, meta-analyses in other occupational groups have found effect sizes for absenteeism and general job performance that were comparable but smaller than ours (Sabagh et al., 2018; Salvagioni et al., 2017).

In this respect, the findings suggest that depersonalization and reduced personal accomplishment, in general, may be no less detrimental than emotional exhaustion. This is in line with research on the prevalence of burnout symptoms, which has illustrated that depersonalization and reduced personal accomplishment are no less significant and occur even more frequently than emotional exhaustion in some teacher samples (Garcia-Carmona et al., 2019). Consistent with substantive theory (Maslach & Leiter, 1999), our results suggest on a descriptive level that emotional exhaustion (as the stress-related symptom of burnout) is most strongly related to absenteeism. By contrast, the role that emotional exhaustion plays in teacher–student interactions appears to be less important. For depersonalization and reduced personal accomplishment, our results seem indicative of a stronger relationship with emotional support, classroom management, and instructional support. These findings support the idea that emotional exhaustion tends to be more decisive for the individual, whereas depersonalization (as the social dimension of burnout) and reduced personal accomplishment (as the self-evaluation dimension) tend to be particularly detrimental to the quality of teacher–student interactions and teaching processes (Maslach & Leiter, 1999).

Primary studies that investigated the association between teacher burnout symptoms and student motivation or achievement are scarce, and they provide conflicting results, as seen in our meta-analyses. Although the results of the few studies that considered student motivation and achievement suggest that students are less intrinsically motivated, engaged, and interested if their teachers experience burnout, evidence regarding student achievement is inconclusive, and our meta-analyses did not find any evidence for a significant association with teacher emotional exhaustion. Nonetheless, it might be too early to dismiss the idea that teacher burnout is associated with student achievement. Three studies that were methodologically especially sophisticated (i.e., they had large samples and used standardized tests to measure student achievement) found significant, albeit small, correlations with teacher emotional exhaustion (Arens & Morin, 2016; Klusmann & Richter, 2014; Klusmann et al., 2016). Because other constructs, especially individual student characteristics such as intelligence, prior knowledge, and conscientiousness (Grasby et al., 2020; Levpušček et al., 2013), determine achievement to a large extent, thereby reducing the potential of teacher characteristics to influence student achievement (Bardach & Klassen, 2021), the smaller effect sizes that we found are not surprising.

Limitations

This study aimed to provide a comprehensive summary of previous findings on teacher burnout. However, some limitations still need to be considered. First, restricting database search strategies (e.g., in PsycINFO) to specific teacher samples increases the risk of missing relevant studies. We sought to mitigate this risk through an extensive supplementary search, including backward and forward citation tracking. This targeted approach was adopted to keep the number of database records manageable. In future research, the challenge of very large numbers of search results also could be addressed by combining a more targeted database search with a broader, machine-assisted search strategy (Campos et al., 2024).

Relatedly, the decision to differentially investigate the three burnout symptoms led to the exclusion of some studies that applied instruments other than the Maslach Burnout Inventory (Maslach et al., 2018) because those studies only reported results for a total burnout score. Although the fact that most studies used the Maslach Burnout Inventory (Maslach et al., 2018) increased the comparability of the effect sizes, including studies that used other instruments could have contributed to a broader understanding of the correlates of teacher burnout.

Second, although the inclusion of various correlates of burnout at different levels of manifestation represents another major strength of our meta-analyses, it may reduce the comparability of studies due to different operationalization methods at the same time. The included operationalization methods differed not only in the source of the reports but also in the type of reporting measure used. Absenteeism was mostly measured by retrospective teacher self-reports, with substantial variation in the periods referenced and the structure of the response options. Considering teacher–student interactions, some studies asked for the specific actions of the teacher (e.g., providing individual support for students), whereas others used student behavior as an indicator of teacher–student interactions (e.g., classroom disturbances). For instance, the dimension of classroom management included both teacher strategies to prevent or address student misbehavior and student discipline behavior. For student motivation (e.g., autonomous motivation, interest, basic needs satisfaction, and engagement), we included different related constructs due to the paucity of primary studies that considered these variables. The same was true for student achievement, for which we combined grades and standardized test scores. In contrast to standardized test scores, grades represent a rather subjective measure based on teacher judgment that encompasses more than the mere achievement of students (e.g., also their general behavior in class). Accordingly, it is not conclusively clear whether all studies included in the meta-analyses measured the same underlying construct.

Third, the studies included in the meta-analyses varied not only in terms of their underlying conceptualization but also in terms of the approach they used to address the multilevel data structure regarding student reports of teacher–student interactions, student motivation, and student achievement; most studies estimated classroom-level correlations on the basis of the manifest means of student-level responses, whereas some studies used the latent representations of the classroom-level constructs. The literature on aggregating multilevel data shows that the two approaches can yield different estimates of classroom-level effects depending on the number of students per classroom, the intraclass correlations of the variables, and the existence and size of contextual effects (Lüdtke et al., 2008). Unfortunately, we were not able to further align the effect sizes, but instead we had to use the correlations provided by the primary studies. Likewise, certain studies considered the school level in their analyses and investigated the relationship between teacher burnout and the professional correlates at the school level. However, because we were specifically interested in this relationship at the teacher level, we excluded the few studies that reported results at the school level.

Fourth, although the decision to exclude special education teachers from the analyses enabled us to provide well-informed conclusions about the population of general education teachers, it prevented us from differentiated statements about the consequences of burnout within the important group of special education teachers, which is confronted with specific challenges and which is particularly at risk of experiencing burnout (Bettini et al., 2017; Brunsting et al., 2022). It would have been interesting to consider teacher orientation (i.e., general vs. special education) as a moderator for the size of the associations (for a comprehensive research synthesis on the correlates of special education teacher burnout, see Brunsting et al., 2026; Park & Shin, 2020). Unfortunately, the sample of studies including only special education teachers was too small to systematically investigate heterogeneity depending on teacher orientation using meta-regression analyses (e.g., Brunsting et al., 2026). Additionally, meta-analyses only provide very limited options for investigating potential (study-level) sources of heterogeneity. For example, meta-regression would allow one to infer the extent to which studies with higher proportions of special education teachers tend to yield higher or lower values of certain correlation coefficients. However, these associations are at the study level, and they would not be informative about whether special education teachers (as individuals) are more or less prone to burnout-related outcomes. A more focused design, such as individual-participant-data meta-analysis, would be an approach to answer this question.

Finally, we exclusively considered cross-sectional correlations in the meta-analyses. Including longitudinal studies significantly reduces the comparability across studies and impedes the interpretation of effect sizes, because some longitudinal studies control for different characteristics at baseline, whereas cross-sectional studies do not. Additionally, the length of the intervals between measurement points varies greatly across studies. However, with nonexperimental cross-sectional data, a causal interpretation is not possible. Because we exclusively considered these kinds of data, the results of our meta-analyses should not be interpreted in terms of causal associations. A reciprocal relationship is also conceivable, in particular for teacher–student interactions and for student motivation and achievement. For example, established theoretical models and recent research have discussed low student motivation, engagement, and performance and disruptive student behavior as stressors that might impede teacher–student interactions and further promote the development of teacher burnout symptoms (Aloe et al., 2014; Jennings & Greenberg, 2009; de Ruiter et al., 2020). Moreover, results from intervention studies also have suggested a reciprocal relationship between teacher burnout and teacher–student interactions. Interventions that aimed to reduce teacher burnout found positive effects for the quality of teacher–student interactions (Roeser et al., 2022), whereas interventions that focused on the improvement of interactional processes showed a positive effect for teacher burnout (Dicke et al., 2015). However, the longitudinal studies reviewed in our research synthesis offer initial insights into the ongoing relationship between teacher burnout and professional consequences, showing effects across various timeframes. Although smaller, they partly appear to confirm reciprocal associations with emotional support, classroom management, and student motivation.

Implications for Future Research

In general, previous research focused predominantly on the symptom of emotional exhaustion, reflecting the tendency to consider it as the central burnout symptom. However, the results of our meta-analyses illustrate the importance of considering all three burnout symptoms in future research to better understand their differential effects (Maslach & Leiter, 1999). In particular, depersonalization and reduced accomplishment should not be ignored because these symptoms seem to be at least as detrimental as emotional exhaustion. In addition, it could be interesting to investigate the interplay between the three burnout symptoms. For instance, prior research has indicated that the appraisal of one’s own abilities as sufficient to meet demands can mitigate negative consequences for students (Herman et al., 2018). This suggests that experiencing personal accomplishment could possibly buffer the experience of other burnout symptoms. To gain a deeper understanding of the interplay among the three burnout symptoms, adopting a person-centered approach would be particularly valuable (Howard & Hoffman, 2018).

Although the dimensional perspective on burnout provides important insights into its structure and consequences, the broader conceptual debate about its distinction from depressive symptoms remains unresolved (Bianchi & Schonfeld, 2025). Rather than reflecting two clearly separable phenomena, burnout and depressive symptoms may partly represent manifestations of a shared underlying factor related to general health impairment. Advancing this debate will require research syntheses that jointly consider burnout and depressive symptoms, ideally supported by more accurate diagnostic tools and valid cut-off values for burnout, as proposed by Schaufeli et al. (2020).

In contrast to the abundant literature on teacher–student interactions, research on how teacher burnout relates to student cognitive-motivational development is scarce, making it difficult to draw reliable conclusions. Although the primary studies included in our meta-analyses partially suggested a negative link between teacher burnout and student motivation and achievement, more research is imperative on these associations. In our review, we focused on student motivation and achievement because these variables represent relevant educational outcomes that have been studied most frequently. However, investigating the association between teacher burnout and other student characteristics, such as affective experiences or social-emotional competencies, also would be worth considering. For instance, close relationships in particular have been assumed when the same construct is considered on the teacher and student levels (Bardach & Klassen, 2021). Thus, looking at the emotional exhaustion of teachers and student emotions (e.g., enjoyment and anxiety) might be promising. Likewise, a negative association between teacher depersonalization and student social-emotional competencies and prosocial behavior is conceivable (Hoglund et al., 2015). Students might adopt the cynical and detached behavior of their teachers because teachers function as role models (Oberle & Schonert-Reichl, 2016).

Building on the univariate associations investigated in our meta-analyses, a model-based meta-analysis could be a promising approach to further investigate the correlates of teacher burnout (Becker & Aloe, 2019; Cheung & Chan, 2005). For instance, such an analysis would make it possible to investigate the potential mechanisms that underlie the relationships between teacher burnout, teacher–student interactions, and student motivation and achievement that are postulated by established theoretical models (Jennings & Greenberg, 2009; Maslach et al., 2001). Similarly, model-based meta-analyses would make it possible to investigate more complex associations and to consider moderating variables such as contextual factors.

To obtain a better understanding of the contextual conditions under which the relationships between teacher burnout, teacher–student interactions, and student motivation and achievement are particularly close, future research should consider two avenues. First, it would be interesting to examine other contextual variables that might explain the variation in the relationships between studies. In particular, the socioeconomic, migration, and ethnic backgrounds of students seem to be relevant in several respects. Generally, teachers are more important to students who have less favorable demographic and socioeconomic backgrounds (Hamre & Pianta, 2005; Klusmann et al., 2016). Consequently, there may be a closer relationship between teacher burnout, the quality of teacher–student interactions, and student motivation and achievement in these contexts (Klusmann et al., 2016). It is plausible that, especially for students at risk of school failure (Sirin, 2005), there is a crucial difference in whether their teachers who are experiencing burnout teach them in a minor subject or in multiple core subjects or are their class teacher.

Second, it would be interesting to consider further contextual factors, such as the school climate, school leadership, social support, and societal recognition, because these factors might help to explain variability in effect sizes across different school contexts (Dutta & Sahney, 2016; Schulze-Hagenest et al., 2026; Zheng et al., 2017). Among others, challenging working conditions, societal recognition, and unrealistic expectations, the classroom composition and a shared history of relationships between teachers and their students both inside and outside the classroom seem relevant characteristics that might influence the size of the relationship between the experiences of burnout, absenteeism, and teacher–student interactions among teachers and student motivation and achievement (e.g., Aldrup et al., 2018; Schulze-Hagenest et al., 2026; van Droogenbroeck et al., 2014). In line with this, research also should investigate the context specificity of burnout and consider the school level in the multilevel structure. So far, only a few studies have investigated the correlates of teacher burnout at the school level. Those studies found only little variation in the levels of burnout between schools among teachers, with the quality of cooperation being particularly important (e.g., Klusmann et al., 2008; Schulze-Hagenest et al., 2026; Zheng et al., 2017). However, initial evidence exists that highlights the importance of school leadership not only for teacher burnout and teacher attrition but also for student development (Zheng et al., 2017).

Finally, future research should focus more on longitudinal or experimental studies. Longitudinal studies appear especially promising to uncover potentially reciprocal relationships and to provide insights regarding the stability and variability of burnout symptoms depending on contextual factors. For instance, a vicious cycle in which a low-achieving, disengaged group of students, high levels of teacher burnout, and ineffective, less supportive teacher–student interactions reinforce each other might be especially pronounced in school contexts with little collegial support or a negative school climate. This knowledge could be particularly important in identifying the most promising target variables for interventions, that is, those that are mutable and relevant. However, our research synthesis points out that extensive longitudinal designs to investigate the consequences of teacher burnout are scarce. Ideally, longitudinal studies should comprise multiple measurement time points within 1 school year (for an example of such a study, see Burić et al., 2024) because teachers changing classes from one school year to the next could impair the interpretability of the results. In addition to longitudinal studies, experiments represent an important avenue for future research. For example, Jennings et al. (2013) conducted a randomized, controlled trial to assess the effectiveness of a mindfulness-based professional-development program. Implementing a similar design would make it possible to test whether a reduction in teacher burnout due to an intervention program represents the mediating mechanism with which the quality of teacher–student interactions can be increased.

Practical Implications

Although causality cannot be determined with certainty from our analyses, it is important to provide teachers with the necessary resources and competencies to prevent a vicious cycle (i.e., teacher burnout, teaching processes, and student experiences and behavior) from developing or to break it at an early stage. Teacher experiences of burnout seem to change more strongly in response to internal and external stressors, whereas other teacher characteristics remain relatively stable (Voss et al., 2023). Accordingly, interventions to prevent or mitigate burnout should be prioritized, particularly in light of the increasing demands placed on teachers. For instance, the research synthesis of Iancu et al. (2018) on the effectiveness of different intervention programs suggested that building knowledge and competencies on how to cope with stress and strengthening resources are generally effective in reducing emotional exhaustion and enhancing personal accomplishment. Regarding the reduction of depersonalization, however, established interventions seem to be less effective (Iancu et al., 2018). This is especially disadvantageous in light of the fact that depersonalization is negatively related to social interactions in particular and poses a threat to the individuals involved in those interactions (i.e., students). Accordingly, it is imperative to focus on the development and implementation of effective interventions that specifically address depersonalization.

In addition to specific interventions, prevention programs also should be generally offered to teachers. Improving the school climate and the work environment should be an integral part of burnout-prevention efforts because intervention approaches that target individual experience can only be effective if the environment is supportive with regard to the newly learned experiences (Walton & Yeager, 2020). In this sense, a school climate in which teachers are supported in their individual needs, demands, and personal development might prevent the occurrence of burnout because it might provide the basis for teachers to develop empowering skills and characteristics. For instance, self-efficacy, social-emotional competencies, and basic need satisfaction, as well as support from colleagues and supervisors, represent important factors to help teachers deal with job demands and high workloads and to protect teachers from experiencing burnout (Aldrup et al., 2017; Aloe et al., 2014; Jennings et al., 2017).