A Comparison of Measurement of Stability and Predictors of Special Education Burnout and Work Engagement

Abstract

Special education teacher (SET) stress and burnout is a significant problem. A total of 490 special education teachers were surveyed across the United States. The purpose of this study was to (a) assess and compare three measures of burnout/work engagement, the Maslach Burnout Inventory (MBI), the Oldenburg Burnout Inventory (OLBI), and the Utrecht Work Engagement Scale (UWES) and (b) examine change over the course of a school year that coincided with the COVID-19 pandemic. Significant measurement quality issues were observed for the MBI and OLBI, including questionable convergent validity. Burnout of SETs was found to be highly stable for the MBI and OLBI. Teachers experienced little mean change in burnout over the school year, and perceptions of the effects of COVID and demographic and school variables were generally not predictive of change in burnout for any measure. Implications are discussed.

Keywords

special education teacher staff development/in-service burnout

Attrition is the long-term consequence of chronic and unaddressed stress and burnout (Brunsting et al., 2014)—an issue especially pronounced for special education teachers (SETs) that has only worsened with the COVID-19 pandemic (Brunsting et al., 2023; McGrew et al., 2023). A nationwide survey of 468 SETs conducted in the winter of 2020 revealed that diagnostic prevalence rates of generalized anxiety disorder and major depressive disorder were 12.4 and 5.6 times the rate observed in normative samples in the United States, respectively (Cormier, McGrew, et al., 2022). Burnout is strongly related to depressive symptomatology (Bianchi et al., 2018), and when unaddressed, teacher burnout leads to poor teacher–student relationships, lower instructional quality, and most importantly, worse student educational outcomes (Covell et al., 2009; Dykstra et al., 2013; Jennings & Greenberg, 2009; Madigan & Kim, 2021; Maricuţoiu et al., 2023). For a workforce that was already experiencing significant issues of attrition (Billingsley & Bettini, 2019) with 98% of states in the United States reporting special education shortages (U.S. Department of Education, 2020), these alarmingly high rates of mental health impacts pose significant concerns that are costly for schools, teachers, and students.

Burnout has been assessed by different measures such as the Maslach Burnout Inventory (MBI; Maslach et al., 1986), the Oldenburg Burnout Inventory (OLBI; Demerouti et al., 2003), and the Utrecht Work Engagement Scale (UWES; Schaufeli et al., 2006). The MBI is the most common measure and has been used for more than 40 years representing more than 90% of studies around the world surrounding burnout (Schaufeli & Enzmann, 1998). A cursory search using the MBI as the search term in Google Scholar produced more than 25,400 results. With respect to educators, historically, the MBI is the primary measure for evaluating SET burnout (Bettini et al., 2017; Garwood et al., 2018; Hopman et al., 2018). The OLBI has been used less frequently than the MBI (Karavasilis, 2019), and the literature is even more sparse for the use of the OLBI with SETs. Studies on the UWES are available, but from other countries (c.f., Simbula et al., 2013), and none for SETs.

Burnout, as defined by the MBI (Maslach et al., 1986), is a three-dimensional construct characterized by emotional exhaustion (EE), depersonalization (DP), and reduced personal accomplishment (PA). EE refers to feelings of being emotionally overextended, DP refers to maladaptive and cynical attitudes and feelings toward others, and PA refers to the negative evaluation of oneself in relation to one’s job. The OLBI assesses exhaustion and disengagement from work. In addition to emotional aspects, exhaustion as assessed by the OLBI includes physical and cognitive aspects, which differentiates it from the MBI (Demerouti et al., 2003). The inclusion of cognitive aspects to one’s work is particularly relevant for SETs whose task demands include complex problem-solving skills involving assessment, development, and implementation of individualized student goals and education plans and collegial interaction and collaboration for support of implementation of instructional plans. The opposite of burnout, work engagement, has also been used to assess commitment to one’s work with the UWES. Work engagement represents a positive work-related experience representative of dedication, high levels of energy, and enthusiasm.

Despite the robust history of burnout research, researchers voiced concerns about its measurement. For example, questions for the MBI include the internal consistency of some of subscales (Wheeler et al., 2011), a focus on emotional vs. physical aspects related to burnout (Demerouti et al., 2010), use of the three-factor structure (Hawrot & Koniewski, 2018; Schwarzer et al., 2000; Szigeti et al., 2017), and unidirectional wording of the items (Demerouti et al., 2003, 2010; Halbesleben & Demerouti, 2005; Kristensen et al., 2005). Wheeler et al. evaluated the internal consistency reliability on the three MBI subscales. PA and DP alpha estimates fell below the levels recommended for determining burnout. Hawrot and Koniewski (2018) examined the construct validity of the MBI-Educators survey using confirmatory factor analysis. Rather than validating the oblique three-factor structure of the MBI, a bifactor model with a general burnout factor and three specific factors of PA, DP, and EE produced the best fit for the data. Further analysis suggested a unidimensional MBI total score was appropriate to use and interpret. Finally, each subscale of the MBI is worded in one direction: EE and DP are phrased negatively, while PA is phrased positively. It may be difficult for raters to endorse negatively worded statements while also agreeing with positively worded statements (Demerouti et al., 2010).

For intervention researchers, however, there are additional unknowns that may have important consequences. For example, it is not clear which burnout tool, if any, is the best outcome measure for evaluating the effectiveness of a burnout reduction intervention. If the burnout measure used is insensitive to the intervention, it has the potential to result in a Type 2 statistical error, falsely deeming a promising or even effective intervention as ineffective. Furthermore, issues of internal consistency and reliability may also mask effects and lower statistical power; studies involving SETs have frequently found low to moderate internal consistency in some burnout subdomains (e.g., Brunsting et al., 2022; Robinson et al., 2019; Squillaci, 2021). The chronicity, stability, and unvarying nature of burnout over time have been replicated (Brouwers & Tomic, 2000; Burke & Greenglass, 1995; Cece et al., 2021; Hakanen et al., 2008; Schaufeli & Enzmann, 1998; Taris et al., 2005) and for the MBI at the subscale levels for EE, DP, and PA (Brouwers & Tomic, 2000). Researchers of burnout interventions report no, small, or mixed effects with the MBI and its subscales when used as an outcome measure (Fraiman et al., 2022; Kestian, 2020). As an example, in a controlled study with classroom teachers, Montero-Marin et al. (2021) compared an instructor-less versus self-led mindfulness intervention. The frequency of use of mindfulness practice was associated with changes in several self-report measures of outcomes such as mindfulness, self-compassion, mental well-being, anxiety, depression, and stress in both conditions. The single exception was a failure to detect a change in MBI scores. In other words, while the intervention produced numerous positive effects on stress and well-being, ratings on the MBI were not one of them, a finding replicated from another test of a different mindfulness curriculum (Fraiman et al., 2022). To add further confusion, researchers may use a single score for the MBI rather than the subscale scores (Montero-Marin et al., 2021; Szigeti et al., 2017). The developers of the MBI recommend that separate scores for each of the subscales be used because of differential response patterns observed based on the burnout outcome assessed (Maslach et al., 1986). Thus, there may be a need to (re)consider the MBI as a primary intervention effectiveness outcome measure in future research and carefully use it in how the measure is applied (separate scores for each domain vs overall mean score). If burnout is truly a stable construct then negative results in intervention studies are to be expected. However, it may also be the case that the responsiveness of the burnout construct is not adequately captured by the burnout measures commonly employed in intervention studies.

Most of the research on special education teacher burnout has focused on understanding its antecedents rather than measurement. Several researchers have identified the job demands associated with teacher burnout, and several comprehensive reviews are available (Billingsley & Bettini, 2019; Brunsting et al., 2014; Park & Shin, 2020). Factors identified include student factors, school factors, working conditions, and teacher factors, including intrapersonal factors. For student factors, teachers of students who are older and with more complex diagnoses or behavioral disorders report higher burnout (Frank & McKenzie, 1993; McLeskey et al., 2004; Nichols & Sosnowsky, 2002; Singer, 1993). School factors have received the most attention and have identified teachers from high-poverty schools, high-minority schools, schools located in the southern United States, and schools located in urban areas as reporting higher burnout (Abel & Sewell, 1999; Carver-Thomas & Darling-Hammond, 2019).

Teacher factors associated with burnout include experience, race, and gender. Findings indicate that young or inexperienced special education teachers (Bettini et al., 2017; Billingsley, 2004; Zabel & Zabel, 2002) report higher burnout. Furthermore, the intersections of race and gender have been explored broadly, but they have received limited attention in special education. For example, Scott et al. (2021) conducted a literature review of attrition and retention of special education teachers. The review revealed that of 47 articles examined over a 10-year period, researchers have nearly ignored disaggregating results to explore factors of attrition and retention related to teacher ethnoracial identities. There are unique factors that exist that impact stress, burnout, and mental health among teachers of color broadly (Cormier et al., 2021), even more pressing is how these racialized experiences in schools impact special education teachers of color (Scott et al., 2022; Trainor et al., 2019). Among these are being hypervisible to take on additional roles as cultural brokers and schoolwide disciplinarians and being devalued as an educator in the field of special education (see Cormier et al., 2021), as well as issues associated with stress, burnout, and mental health of special education teachers of color compared with White special education teachers.

Given the limited research that has been conducted related to SET and burnout, the clinical significance of the study of burnout, and the equally important need for clarity of relevant and sensitive outcome measures for intervention researchers, the purpose of this study was twofold to (a) assess and compare three measures of burnout/work engagement, the MBI (Maslach et al., 1986), the OLBI (Demerouti et al., 2003), and the UWES (Bakker et al., 2003) with data collected at two timepoints during a school year coinciding with the COVID-19 pandemic, and (b) examine change in burnout/work engagement over the course of a school year that coincided with the COVID-19 pandemic and compare that change with teacher stress, teacher demographic variables, and school variables. Understanding how different burnout/work engagement measures compare may inform future research on SET burnout. Specifically, we address the following research questions:

Research Question 1: Internal Structure. Do the MBI, OLBI, and UWES scales exhibit the same psychometric properties at both timepoints?

Research Question 2: Convergent Validity Evidence. To what extent do the MBI, OLBI, and UWES measure the same burnout construct?

Research Question 3: Stability of Burnout. How stable is burnout of SETs during a school year as measured by the MBI, OLBI, and UWES?

Research Question 4: Impact of the COVID-19 Pandemic. How much do SET perceptions of the impact of the COVID-19 pandemic on their stress level and emotional exhaustion influence their experience of burnout as measured by the MBI, OLBI, and UWES?

Research Question 5: Covariates of Burnout. Which teacher demographic and school variables influence changes in burnout?

Method

Participants were SETs at public or charter schools in the United States (see Table 1). A recruitment flyer was generated and distributed to personal and professional contacts, social media, school districts, and professional teacher organizations. Across the United States, participants completed surveys of the variables over two time points: Time 1 (October-November 2020) and Time 2 (May 2021). Participants who completed Time 1 were invited to complete Time 2 with no new teachers recruited. Participants received no incentives for participation. Teacher and school demographic information and teacher intrapersonal variables were obtained at Time 1 only. Time 2 included an assessment using all three outcome measures: the MBI, the OLBI, and the UWES. The survey was administered using Qualtrics®. The study was approved by the IRB at the institution of the first author.

Table 1.

Demographic and School Variables for Time 1 and Time 2.

Demographic variable	Time 1 (n = 490)		Time 2 (n = 199)
Demographic variable	M (SD) or n	Percentage	M(SD) or n	Percentage
Age	42.94 (11.32)		43.70 (11.08)
Gender
Female	437	89.2%	177	88.9%
Male	52	10.6%	21	10.6%
Years of teaching experience	13.72 (9.28)		14.88 (9.20)
Years in Position	6.56 (6.78)		6.61 (6.96)
Race and ethnicity (choose all that apply)
White or Caucasian	429	87.6%	180	90.5%
Black or African American	47	9.6%	14	7.0%
American Indian or Alaska Native	7	1.4%	3	1.5%
Asian	3	0.6%	1	0.5%
Native Hawaiian or Pacific Islander	0	0%	0	0%
Latino/a	31	6.3%	12	6.0%
Other	18	3.7%	8	4.0%
School location
Rural	64	13.1%	22	11.1%
Suburban	170	34.7%	71	35.7%
Urban	256	52.2%	106	53.3%
Grade-level taught (choose all that apply)
Preschool and PreK	46	9.4%	23	11.6%
Early Elementary (K–2)	146	29.8%	60	30.2%
Elementary (3–5)	196	40%	79	39.7%
Middle school (6–8)	183	37.3%	66	33.2%
High school (9–12)	155	31.6%	70	35.2%
%Free/Reduced price Lunch School (Title 1)	323	65.9%	116	58.3%

Measures

Background Information

Teachers were asked to provide demographic and school background information. Information such as age, gender, race/ethnicity, type and location (state) of school, whether the school was a Title 1 school, years of experience teaching, years of experience in current position, number of IDEA category(ies) of students taught, type of classroom taught (resource, inclusive, self-contained; specialty school) and grade level taught (see Table 1).

Impact of COVID-19 on Emotional Exhaustion and Stress

To understand the impact of the pandemic on teachers, two single-item questions were asked at the end of the school year (Time 2): (a) how has the COVID-19 pandemic impacted your emotional exhaustion and (b) your stress. Participants were asked to rate these items on a 5-point Likert-type type response scale (1 = not at all to 5 = extremely).

Outcome Measures of Burnout and Engagement

Burnout

The Maslach Burnout Inventory, Educator’s Survey (MBI-ES; Maslach et al., 1986) assesses teacher burnout. The MBI-ES consists of three subscales: (a) emotional exhaustion (EE; 9 items), (b) depersonalization (DP; 5 items), and (c) personal accomplishment (PA; 8 items). Participants were asked to rate how frequently they felt burned out at work on a seven-point Likert-type type response scale (0 = never to 6 = every day). Items were averaged to produce overall subscale scores. Higher scores on the emotional exhaustion and depersonalization subscales and lower scores on the personal accomplishment subscale indicated higher levels of burnout. According to Squillaci (2020), emotional exhaustion scores below 18 are considered to represent a correct level, scores between 18 and 29 are considered at risk, and scores greater than 29 represent a dangerous level; for depersonalization, scores below 6 are correct, between 6 and 11 are at risk, and above 11 are dangerous; and for personal achievement, scores above 39 are correct, between 34 and 39 are at risk, and below 34 are dangerous. Internal consistency of the sample at Time 1 was ⍺=.899 and ⍵=.902 for the EE subscale, ⍺=.667 and ⍵=.636 for the DP subscale, and ⍺=.723 and ⍵=.719 for the PA subscale. Internal consistency of the sample at Time 2 was ⍺=.910 and ⍵=.919 for the EE subscale, ⍺=.631 and ⍵=.616 for the DP subscale, and ⍺=.708 and ⍵=.725 for the PA subscale.

The Oldenberg Burnout Inventory (OLBI) is a 16-item measure burnout (Demerouti & Bakker, 2008). The OLBI consists of two subscales, exhaustion (EXH; including affective, cognitive, and physical components) and disengagement (DIS; including distancing oneself from work goals and work content specifically and from work generally). The 16 items are distributed evenly across the two subscales and include positively and negatively worded items. Items that are positively framed are reverse-scored before creating subscale totals. Items on both subscales were summed to create total scores, with higher scores indicating higher levels of burnout. Internal consistency of the sample at Time 1 was ⍺ = .771 and ⍵ = .786 for the EXH subscale and ⍺ = .778 and ⍵ = .785 for the DIS subscale. Internal consistency of the sample at Time 2 was ⍺ = .740 and ⍵ = .783 for the EXH subscale and ⍺ = .785 and ⍵ = .759 for the DIS subscale.

Work Engagement

The abridged version of the Utrecht Work Engagement Scale was used to assess work engagement (Schaufeli et al., 2006). The UWES measures work engagement in three domains: vigor, dedication, and absorption. Participants self-report their perception of their work engagement on a five-point Likert-type response scale (0 = never to 5 = almost always). The original version of the UWES consists of 19 items, but the current study utilized a six-item version that assessed work engagement in the domains of vigor and dedication. Schaufeli and colleagues (2006) recommended using total scores for the shortened version to reduce issues of multicollinearity given the high correlations between domains. Therefore, items were averaged to produce an overall score, with higher scores indicating higher levels of work engagement. Internal consistency of the current sample was ⍺=.895 and ⍵=.888 at Time 1 and ⍺=.888 and ⍵=.881 at Time 2.

Data Analysis

Research Question 1: Internal Structure

To assess the longitudinal measurement invariance of the MBI, OLBI, and UWES, a longitudinal confirmatory factor analysis (CFA) framework was employed using Mplus (Muthén & Muthén, 2023) utilizing the MLR estimator to make standard errors and likelihood ratio tests robust to non-normality in item responses. Measurement invariance testing was performed by applying increasingly restrictive constraints to model parameters across time points. The first model tested for each instrument was a configural model, in which a standard correlated traits model was fit combining both time points. For example, the configural model for MBI had separate factors for EE, DP, and PA at each of the two time points. To account for item-specific variance, correlation between timepoints was allowed for each item individually. Model fit was evaluated by chi-square test, comparative fit index (CFI; adequate fit > .90, good fit > .95), root mean square error of approximation (RMSEA; good fit < .05), and standardized root mean square residual (SRMR; good fit < .08) (Browne & Cudeck, 1993; Hu & Bentler, 1999). In the event of a poor fit of the configural model, modifications were considered so that a well-fitting model could be used as the beginning of the measurement invariance testing procedure (Jorgensen, 2017). Modifications were considered and employed based on published empirical measurement studies, large residual correlations within the configural model, and exploratory factor analysis (EFA) models at each time point.

Once a well-fitting configural model was fit, a sequence of increasingly restrictive models was fit and their fit was compared. The first such model, the weak invariance model, adds equality constraints to factor loadings for each item across timepoints The second model, the strong invariance model, adds equality constraints to intercepts for each item across time points. The final model, the strict invariance model, adds equality constraints to residual variances for each item across time points. Model comparisons were conducted using a likelihood ratio test; models were deemed invariant if the was not significant (p > .05). Even if the likelihood ratio test was significant, measurement invariance was still concluded if differences in the root mean square error of approximation (ΔRMSEA) was ≤.015 (Chen, 2007) and the difference in comparative fit index (ΔCFI) was ≥−.01 (Cheung & Rensvold, 2002).

Research Question 2: Convergent Validity Evidence

To assess whether MBI subscales, OLBI subscales, and the UWES measure the same burnout construct, correlations among subscales at both time points were computed. In addition, an EFA was conducted in Mplus (Muthén & Muthén, 2023) using the six observed subscores as indicators. The number of factors to extract was determined by visual inspection of the scree plot (Cattell, 1966).

Research Question 3: Stability of Burnout

To answer the third question about the stability of burnout over the school year, descriptive and inferential statistics were used. Descriptive statistics for MBI subscales, OLBI subscales, and the UWES at Time 1 and Time 2 were computed. Next, repeated measures t tests for MBI subscales, OLBI subscales, and the UWES to determine whether any overall change in levels of burnout could be detected across time points were applied. As a sensitivity analysis, this t test was replicated in the strict measurement invariance models. All analyses for the third, fourth, and fifth research questions were performed using R (R Core Team, 2023).

Research Question 4: Impact of the COVID-19 Pandemic

To answer the fourth research question concerning the impact of COVID-19 on emotional exhaustion and stress, teacher responses were summarized and also correlated with both burnout at Time 2 and change in burnout (Time 2—Time 1) for each MBI, OLBI, and UWES domain. As the impact of COVID-19 items is ordinal, Kendall’s tau-b was used to assess correlations.

Research Question 5: Covariates of Burnout

For our fifth question on teacher and school predictor variables of change in burnout, we computed correlations for change scores (Time 2—Time 1) of subscales of the MBI, the OLBI, and the UWE with demographic and school-related variables. For ordinal predictor variables, Kendall’s tau-b was used to assess the relationship with change scores. For nominal predictor variables, analysis of variance was used to assess the relationship with change scores; the square root of eta-squared was reported as being a rough analog of a correlation coefficient.

Because participants were only required to respond to background questions and Time 1 MBI items to be included in the study, significant missingness for other measures at Time 1 and Time 2 was found. Time 1 UWES had 20% missing and Time 1 OLBI measures had 42% missingness. The dropout for MBI domains was 59% (total 59% missing at Time 2), the dropout for work engagement was 54% (total 63% missing at Time 2), and the dropout for OLBI domains was 32% (total 61% missing at Time 2). While substantial, this level of missingness is consistent with other large-scale longitudinal surveys (Gustavson et al., 2012). Continuers did not vary substantially from dropouts on any outcome at Time 1 (maximum Cohen’s d difference of 0.12).

Results

A total of 490 SETs completed the survey at Time 1. Participants came from all 50 states with the highest percentage from Pennsylvania (13.7), Georgia (10.8), and Illinois (8.8). Table 1 shows the participant demographics at Time 1 and Time 2; 40.6 % of participants were retained from Time 1 to Time 2. Overall, similar percentages for each of the variables were retained across the two time points for the variables. The mean age of the participants was 43, the large majority was female (89%). The mean years of teaching experience was 14 and the number of years in the current position was 7. Differences were noted between the two time points with respect to race and grade level. At Time 2, about 91% of participants were White (an increase of 3%), while the number of Black teachers was 7% (a decrease of 3%). For grade level, 11% of teachers taught pre-K and kindergarten (increase of 3% from time 1), 33% taught middle school (decrease of 4%), and 35% taught high school (increase of 3%). The highest percentage of teachers were from elementary schools; a large proportion, more than 50%, also represented Title 1 schools.

Evaluation of Time 2 MBI-EE scores indicated that 62.2% of teachers fell within the dangerous level of burnout while 27.8% more were at risk (Squillaci, 2020). According to Time 2 MBI-Depersonalization scores, 14.7% of teachers were at dangerous levels and 36.3% were at risk. Finally, according to Time 2 MBI-Personal Accomplishment scores, 24.5% of teachers were at dangerous levels and 38.6% were at risk. Note that the evaluation of teachers as falling into dangerous and at-risk categories was necessarily performed with the full items set (i.e., without removing Items 12 and 16).

Research Question 1: Internal Structure

Results of configural model fit and measurement invariance testing can be found in Table 2. Initial configural models for MBI and OLBI had poor fit and were modified to create a well-fitting configural model. The configural model for the UWES exhibited a good fit. For the MBI, EFA performed at Time 1 and Time 2 both revealed no substantial (>.30) loadings for Item 16 (“Working with people directly puts too much stress on me.”) and a stronger cross-loading of Item 12 (“I feel very energetic.”) on the EE factor than on the PA factor to which it belongs. A 2008 review of MBI measurement studies (Worley et al., 2008) indicated that significant problems with these two items were identified long ago and these items were commonly removed when EFA and CFA techniques are used. A configural CFA was refit after removing Item 12 and Item 16 but still exhibited poor fit, largely due to residual correlations between Item 5 (“I feel I treat some students as if they were impersonal objects.”) and Item 15 (“I don’t really care what happens to some students”); these residual correlations were .246 at Time 1 and .180 at Time 2. These large residual correlations were reported by numerous studies in Worley et al.’s (2008) review. After removing Item 12 and Item 16 and allowing the residuals of Item 5 and Item 15 to covary, the configural model exhibited an acceptable fit. Items 12 and 16 of the MBI were removed from subdomain scores for all subsequent analyses.

Table 2.

Measurement Invariance Testing.

Instrument	Model	χ²	df	Δp	RMSEA	ΔRMSEA	CFI	ΔCFI	SRMR
MBI	Configural	1,960.892	865		.051		.839		.086
	Modified Configural	1,165.278	693		.037		.922		.068
	Weak	1,194.412	710	.034	.037	.000	.920	−.002	.072
	Strong	1,234.717	727	.001	.038	.001	.916	−.004	.073
	Strict	1,244.912	747	.407	.037	−.001	.917	.001	.076
OLBI	Configural	859.221	442		.051		.828		.083
	Modified Configural	568.347	363		.039		.905		.071
	Weak	588.167	383	.415	.038	−.001	.905	.015	.077
	Strong	607.187	395	.088	.038	.000	.902	−.003	.077
	Strict	616.277	403	.368	.038	.000	.901	−.001	.078
UWES	Configural	60.141	43		.032		.991		.036
	Weak	63.609	48	.724	.030	−.002	.992	.001	.040
	Strong	68.132	53	.484	.028	−.002	.992	.000	.047
	Strict	72.340	59	.604	.024	−.004	.993	.001	.043

Note. MBI = Maslach Burnout Inventory; OLBI = Oldenburg Burnout Inventory; UWES = Utrecht Work Engagement Scale; df = degrees of freedom; Δp = p value associated with chi-square difference test; RMSEA = root mean square error of approximation; CFI = comparative fit index; SRMR = standardized root mean square residual.

For the OLBI, a review of measurement literature revealed that CFAs including a method factor for negatively phrased items result in substantially improved fit. A three-factor EFA model was fit at each time point which recovered the EXH, DIS, and negatively phrased method factor. However, Item 16 (“When I work, I usually feel energized.”) cross-loaded strongly (λ = .579 at Time 1 and λ = .621 at Time 2) on the DIS factor and had a small (λ = .182 and λ = .095 at Time 2) loading on the EXH factor. We did not find a replication of this finding in the literature; however, because the phenomenon was so pronounced, we removed Item 16 from the OLBI for all further analyses. A configural model for the OLBI which included a negatively phrased item method factor and from which Item 16 was removed fit adequately.

Following the fitting of acceptable configural models, longitudinal measurement invariance testing reported in Table 2 revealed that all three instruments exhibited adequate invariance across time points for all levels of invariance.

Research Question 2: Convergent Validity Evidence

A correlation matrix for all six burnout subdomains at both time points can be found in Table 3. When all six burnout subdomains were used as indicators in an EFA, inspection of the eigenvalues revealed a strong first eigenvalue (3.62 at Time 1 and 3.50 at Time 2) and very small second eigenvalue (1.01 at Time 1 and 0.93 at Time 2). Accordingly, a one-factor CFA model (equivalent to a one-factor EFA model) was fit at each time point so that residual correlations could be more easily examined. All six subdomains loaded significantly onto the general burnout factor; however MBI PA (λ = −.411 at Time 1 and λ=-.419 at Time 2) and MBI DP (λ = .644 at Time 1 and λ = .588 at Time 2) loaded less strongly than other domains (|λ| > .70). In addition, substantial residual correlations were found between OLBI EXH and MBI EE (r = .232 at Time 1 and r = .105 at Time 2) and between MBI PA and UWES (r = .171 at Time 1 and r = .150 at Time 2).

Table 3.

Correlations Among Burnout Subdomains.

Variable	MBI-EE	MBI-DP	MBI-PA	OLBI-EXH	OLBI-DIS	UWES
MMI-EE		.543	−.179	.787	.620	−.592
MBI-DP	.489		−.287	.487	.549	−.483
MBI-PA	−.228	−.346		−.158	−.389	.542
OLBI-Exh	.775	.405	−.248		.572	−.499
OLBI-Dis	.639	.527	−.453	.583		−.783
UWES	−.589	−.436	.479	−.527	−.662

Note. Correlations above the diagonal are correlations at Time 1. Correlations below the diagonal are correlations at Time 2. All correlations are significant after applying Bonferroni correction for multiplicity (i.e., p < .05/30). MBI = Maslach Burnout Inventory, EE = Emotional Exhaustion, DP = Depersonalization, PA = Personal Accomplishment; OLBI = Oldenburg Burnout Inventory, Exh = Exhaustion, Dis = Disengagement; UWES = Utrecht Work Engagement Scale.

Research Question 3: Stability of Burnout

For evaluation of stability or change in burnout of SETs over the course of the school year using the three measures, Table 4 shows mean values at Time 1 and Time 2 as well as the correlations between scores at the two time points for each burnout measure. Baseline scores for each of the subscale measures significantly correlated with the Time 2 assessment, with between 35% (MBI-PA) and 55% (UWES) of variance in common between baseline and Time 2 scores. For the related question about change in burnout over the school year, when mean change in burnout scores was evaluated using t tests (Table 5), the UWES was the only measure that showed significant change, a decrease, from Time 1 to Time 2, t(180) = 2.73, p = .007, Cohen’s d = 0.15.

Table 4.

Stability of Burnout and Work Engagement.

Variable	Time 1 Mean (SD)	Time 2 Mean (SD)	T1-T2 Correlation
MBI: EE	30.17 (9.60)	30.57 (10.64)	.72 (.77)
MBI: DP	5.81 (5.06)	6.34 (5.15)	.66 (.79)
MBI: PA	34.25 (5.31)	33.94 (5.30)	.59 (.76)
OLBI: Exh	20.61 (3.27)	20.47 (3.34)	.66 (.80)
OLBI: Dis	15.55 (4.17)	15.93 (3.85)	.71 (.77)
UWES	20.24 (4.64)	19.58 (4.42)	.74 (.79)

Note. Correlations in parentheses are latent correlations estimated in the strict measurement invariance model. All T1-T2 correlations are significant after using a Bonferroni adjustment to correct for multiplicity (i.e., p < .05/6). MBI = Maslach Burnout Inventory, EE = Emotional Exhaustion, DP = Depersonalization, PA = Personal Accomplishment; OLBI = Oldenburg Burnout Inventory, Exh = Exhaustion, Dis = Disengagement; UWES = Utrecht Work Engagement Scale.

Table 5.

t Tests of Mean Change Scores.

Variable	Mean Change (SD)	t test	p	Cohen’s d^a
MBI: EE	0.40 (7.64)	t(198) = 0.74	.461	0.05
MBI: DP	0.53 (4.19)	t(198) = 1.78	.077	0.10
MBI: PA	−0.31 (4.82)	t(198) = −0.91	.367	−0.03
OLBI: Exh	−0.13 (2.71)	t(111) = −0.52	.602	−0.02
OLBI: Dis	0.38 (3.07)	t(108) = 1.28	.203	0.09
UWES	−0.66 (3.27)	t(180) = 2.73	.007	−0.15

Note. MBI = Maslach Burnout Inventory; EE = Emotional Exhaustion; DP = Depersonalization; PA = Personal Accomplishment; OLBI = Oldenburg Burnout Inventory; Exh = Exhaustion; Dis = Disengagement; UWES = Utrecht Work Engagement Scale.

Cohen’s d was computed using pooled standard deviation as the standardizer.

Research Question 4: Impact of the COVID-19 Pandemic

To answer the fourth question about the impact of the COVID-19 pandemic on teacher emotional exhaustion and stress, 75.6% of teachers reported that the pandemic had a moderate to extreme impact on their emotional exhaustion while 87.2% of teachers reported a moderate to extreme impact on their stress. Only 4.4% of teachers reported that the COVID-19 pandemic did not impact their emotional exhaustion at all while 0.6% reported no impact on their stress. The only significant relationship between self-report of the effect of the COVID-19 pandemic and change in burnout and work engagement variables was between the effect of the COVID-19 pandemic on exhaustion and change in EE (τ = .14); these correlations are reported in Table 6. Teachers reporting higher levels of COVID-19 impact on their stress also reported significantly higher levels of EE (τ = .25) and EXH (τ = .22) and significantly lower levels of work engagement (τ = −.14), but no significant difference in DP (τ = .04), PA (τ = .02), or DIS (τ = .11) at Time 2. Teachers reporting higher levels of COVID-19 impact on their emotional exhaustion also reported significantly higher levels of EE (τ = .28) and EXH (τ = .27) and significantly lower levels of work engagement (τ = −.14), but no significant difference in DP (τ = .06), PA (τ = −.06), or DIS (τ = .11) at Time 2.

Table 6.

Correlations Between COVID-19 Factors, Teacher Demographics and School Factors with Change in Burnout and Work Engagement.

	MBI			OLBI		UWES
Variable	EE	DP	PA	Exh	Dis	UWES
COVID-19 Factors
Exhaustion(a)	.16*	.11	−.06	.12	.09	−.02
Stress(a)	.12	.03	−.03	.11	.06	−.04
Demographic Factors
Age	−.05	−.11	.12	−.13	−.14	.11
Gender(b)	.07	.06	.01	.01	.04	.05
Race(b)	.22*	.21	.18	.13	.16	.09
Years In Position	−.04	−.00	−.02	−.06	−.09	−.06
Years Taught	−.08	−.11	.04	−.10	−.16	.06
Second Job	−.01	.00	−.01	−.02	−.03	.03
Hours Worked/Week	.04	.09	−.10	−.08	−.03	−.00
School Factors
School Type(b)	.11	.07	.12	.31*	.24	.14
Grade Level(b)	.15	.06	.16	.20	.24	.19
Number IDEA Categories	.05	.11	−.04	.04	−.15	−.09
Title 1 School	.02	−.10	−.03	−.01	−.08	.04
Classroom Type(b)	.13	.20	.20	.26	.23	.08

Note. (a) For ordinal predictors, Kendall’s tau-b correlation coefficients are listed. (b) For nominal predictors, the square root of eta-squared from one-way ANOVA is listed in this table. MBI = Maslach Burnout Inventory, EE = Emotional Exhaustion, DP = Depersonalization, PA = Personal Accomplishment; OLBI = Oldenburg Burnout Inventory, Exh = Exhaustion, Dis = Disengagement; UWES = Utrecht Work Engagement Scale.

p < .05.

Research Question 5: Covariates of Burnout

For the last question about teacher and school factors associated with burnout at Time 2, only two factors were significant; all correlation estimates can be found in Table 6. For teacher variables, race (η² = .05, p = .036) was associated with a change in the MBI- Emotional Exhaustion scale. During the course of the school year, average emotional exhaustion for Black teachers decreased whereas average emotional exhaustion increased for teachers of other races. Although Time 1 scores were lower for Black teachers, there was no significant difference in MBI-Emotional Exhaustion at Time 1 between Black teachers (M=28.8) and teachers of other races (M=32.5). Further, only one additional school factor, school type, correlated with the OLBI exhaustion subscale (η² = .07, p = .042); teachers at charter schools experienced a decrease in average exhaustion during the school year whereas teachers at other schools experienced no average change.

Discussion

Rather than understanding the impact of the COVID-19 pandemic on teacher burnout, our original intent of the study was to understand the stability of burnout over the school year when measured by different instruments and the impact of demographic variables on change in burnout assessed by the different tools. However, given that data collection coincided with the impact of the COVID-19 pandemic, we gained an opportunity to understand the impact of the pandemic as it wreaked havoc on systems of care, including public schools and the students they serve (Asbury et al., 2020; Pressley, 2021). These findings expose the significant burnout special educators experienced. Nearly 75% of teachers reported that the COVID-19 pandemic had a high impact on their emotional exhaustion and even more on their stress. Further and consistent with these findings, nearly all teachers reported emotional exhaustion with the MBI at the dangerous level/at-risk levels of burnout based on the categorization described by Squillaci (2021); and more than half reported MBI-Depersonalization scores and reduced MBI-Personalization scores at dangerous/at-risk levels.

Analysis of the internal structure of the different measures revealed that the MBI had significant measurement quality issues, including factor structure problems and low reliability in DP and moderate reliability in PA. Similarly, low-to-moderate reliabilities have been previously found in other studies of burnout in SETs (Brunsting et al., 2023; Langher et al., 2017; McGrew et al., 2023; Robinson et al., 2019; Squillaci, 2021). These findings of low to moderate reliability are concerning due to the tendency of low reliability to inflate Type 2 statistical errors (Sutcliffe, 1980). Findings of non-significance, such as those within this study or the studies detailed in the introduction may possibly be due to properties of the measure of burnout used rather than of the burnout construct itself. In addition, measures with low reliability can create bias in multivariate analyses such as multiple regression or structural equation modeling (Cole & Preacher, 2014). On the positive side, all three scales demonstrated adequate longitudinal measurement invariance, suggesting that the meaning of scores remains constant across time points.

Investigation of the convergent validity of MBI, OLBI, and UWES domains revealed that while MBI-EE, OLBI-EXH, OLBI-DIS, and UWES all relate strongly to each other, MBI-DP and MBI-PA exhibit substantial unique variability. If a researcher’s theoretical model of burnout is as a singular construct, a total, unidimensional, MBI score would not likely reflect that construct. The large number of researchers employing an MBI total score may be guilty of a misconception described by Lilienfeld and Strothers (2020). They present the erroneous belief of researchers that a measure taps into the construct based on its name rather than content and that the instruments are reliable. When such assumptions are made, the replicability of psychological and educational research is at risk. By contrast, if a multidimensional theory of burnout is to be used, the MBI may be more consistent with that theory than the OLBI or UWES.

Equally concerning is the finding of the constancy of high burnout over the school year reported by teachers. Prior research dating back as far as 30 years ago indicated that burnout as measured with the MBI is highly stable (Burke & Greenglass, 1995). Not only did the results in this study support these early and more recent findings with the MBI (Cece et al., 2021), burnout as measured with the OLBI in this study was also stable across the school year and also a finding consistent with the literature (Hultell et al., 2013). Examination of the very small correlations between change in burnout and teacher and school factors combined with the very large correlations between burnout scores at Time 1 and Time 2 suggests not only (a) burnout is quite stable but also begs the question (b) to the extent burnout does change, is that change largely unrelated to background contexts and are these findings just noise? However, there was a significant change from Time 1 and Time 2 for the UWES, although the effect size was small. This result differs from a study by Schwarzer et al. (2000) and by Seppälä et al. (2009) who examined the UWES longitudinally and concluded it was stable, like the MBI and OLBI. These findings also differ from Brunsting et al. (2022) in their survey of special education teachers of students with emotional-behavioral disorders. Of the 230 teachers surveyed nationally during the COVID-19 pandemic, ratings of emotional exhaustion decreased and personal accomplishment increased significantly over the school year.

Examination of the correlations between school and teacher variables and change in burnout and work engagement generally did not match correlations reported in the literature at a single time point. For example, in a review of the literature, Brunsting et al. (2014) summarized findings of school factors such as financial support, classroom setting, and classroom composition as accounting for burnout—variables not significant in our study. The only finding for school factors was with school type (public, private, and charter). Teachers from charter schools reported less change in burnout with the OLBI exhaustion subscale compared to teachers from other types of schools. These findings suggest that, while school factors may relate to levels of burnout cross-sectionally, they have little relevance to the progression of burnout in individuals.

Perceptions of the effect of COVID-19 on stress and emotional exhaustion were mostly not predictive of change in burnout for any measure. Furthermore, demographic and school variables were mostly not predictive of change in burnout for any measure, with the exception of the MBI-EE with race. For teacher factors, again it was unexpected that only one variable correlated with burnout, and that was only with one burnout subscale. As with school factors, our findings suggest that the teacher factors investigated in our study have little influence on the progression of burnout. In contrast, Brunsting et al. (2014) reported teacher age, gender, and experience as correlating with burnout at a single time point. That is in general, older teachers, female teachers, and more experienced teachers reported lower scores of burnout. However, in contrast, the only finding in our sample was with race. MBI-Emotional exhaustion decreased for Black teachers and increased for teachers of other races. This finding was unexpected because research suggests that Black teachers experience more stress in their jobs (Scott et al., 2022; Trainor et al., 2019). However, a study by Baker et al. (2021) that took place during the first months of the pandemic showed that Black teachers reported less negative impact of stressors, better mental health, and more protective factors and impact of protective factors on coping and teaching compared to White teachers. These findings are also consistent with another study by Bottiani et al. (2019) that occurred prior to the pandemic. Bottiani and colleagues (2019) sampled 255 teachers from low-income and urban middle schools. They found that White teachers reported higher stress and burnout compared with Black teachers. Because of the low sample size of Black teachers in our study, these results should be interpreted with caution.

The research surrounding the unique racialized experiences of SETs who are from minoritized backgrounds is limited overall. Most recently several scholars explored these experiences in depth (see Boveda & Weinberg, 2022; Cormier, Scott, et al., 2022; Drame et al., 2022; Kulkarni et al., 2022; Siuty & Atwood, 2022). While the challenges of stress and burnout that SETs experience are often similar regardless of the race and ethnicity of the teachers, SETs of color experience stressful work lives in ways not experienced by White teachers (Cormier, Scott, et al., 2022, 2023) and may not be captured in our data using these measures of burnout and work engagement. Therefore, these findings suggest that future research should disaggregate results and examine the work lives of SETs by race and ethnicity to provide more robust scholarship in this area.

Limitations

There were important limitations to consider for this study that may impact the potential to generalize the information learned. First, the study was initiated during the fall of 2020 when many schools continued to experience the COVID-19 pandemic lockdown. Second, because of missing data, it may be likely that the Time 2 sample represents teachers who were more resilient with fewer stressors over the school year. Third, it is also possible that the volunteer participants who completed the survey online may be different from teachers who chose to not complete the survey or were not connected to our advertising outlets. Fourth, because the study was administered nationwide, we were unable to provide incentives. This is because one state on the West Coast of the United States required that any individual interested in the study be compensated with the same incentive structure whether they qualified for the study or not. Thus, because we were unsure of the response rate, we were unable to offer any incentives, potentially leading to higher attrition over time. Finally, since we used a convenience sample, it may be difficult to interpret the lack of significant correlations between teacher/school factors and our burnout measures. Therefore, it may be likely that teachers whose personal characteristics or school contexts had more significant association with burnout were less likely to participate in this study.

In conclusion, burnout and attrition are significant issues for public schools. It is time to move into more intervention research that addresses teacher burnout. However, these findings suggest that burnout may be difficult to change given its robustness over time and measurement issues. Thus, researchers may need to consider and explore other outcomes that are more amenable to change and sensitive to intervention.

Footnotes

Acknowledgements

The authors thank Lindsey Ogle for her help with survey development and data collection and teachers who gave their valuable time.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

This research was supported by the Institute of Education Science Grant #R324A200232.

ORCID iDs

Lisa Ruble

Christopher J. Cormier

John McGrew

References

Abel

M. H.

Sewell

(1999). Stress and burnout in rural and urban secondary school teachers. Journal of Educational Research, 92(5), 287–293. https://doi.org/10.1080/00220679909597608

Asbury

Fox

Deniz

Code

Toseeb

(2020). How is COVID-19 affecting the mental health of children with special educational needs and disabilities and their families? Journal of Autism and Developmental Disorders, 51(5), 1772–1780. https://doi.org/10.1007/s10803-020-04577-2

Baker

C. N.

Peele

Daniels

Saybe

Whalen

Overstreet

(2021). The experience of COVID-19 and its impact on teachers’ mental health, coping, and teaching. School Psychology Review, 50(4), Article 491. https://doi.org/10.1080/2372966X.2020.1855473

Bakker

Demerouti

Schaufeli

(2003). Dual processes at work in a call centre: An application of the job demands-resource model. European Journal of Work and Organizational Psychology, 12(4), 393–417. https://doi.org/10.1080/13594320344000165

Bettini

Jones

Brownell

Conroy

Park

Leite

Crockett

Benedict

(2017). Workload manageability among novice special and general educators: Relationships with emotional exhaustion and career intentions. Remedial and Special Education, 38(4), 246–256. http://doi.org/10.1177/0741932517708327

Bianchi

Schonfeld

I. S.

Laurent

(2018). Burnout syndrome and depression. In Young-Ku

(Ed.), Understanding depression (pp. 187–202). Springer. https://doi.org/10.1007/978-981-10-6577-4_14

Billingsley

B. S.

(2004). Special education teacher retention and attrition: A critical analysis of the research literature. Journal of Special Education, 38(1), 39–55. https://doi.org/10.1177/00224669040380010401

Billingsley

B. S.

Bettini

(2019). Special education teacher attrition and retention: A review of the literature. Review of Educational Research, 89(5), 697–744. https://doi.org/10.3102%2F0034654319862495

Bottiani

J. H.

Duran

C. A.

Pas

E. T.

Bradshaw

C. P.

(2019). Teacher stress and burnout in urban middle schools: Associations with job demands, resources, and effective classroom practices. Journal of School Psychology, 77, 36–51. https://doi.org/10.1016/j.jsp.2019.10.002

10.

Boveda

Weinberg

A. E.

(2022). Centering racialized educators in collaborative teacher education: The development of the intersectionally conscious collaboration protocol. Teacher Education and Special Education, 45(1), 8–26. https://doi.org/10.1177/08884064211062874

11.

Brouwers

Tomic

(2000). A longitudinal study of teacher burnout and perceived self-efficacy in classroom management. Teaching and Teacher Education, 16(2), 239–253. https://doi.org/10.1016/S0742-051X(99)00057-8

12.

Browne

M. W.

Cudeck

(1993). Alternative ways of assessing model fit. In Bollen

K. A.

Long

J. S.

(Eds.), Testing structural equation models (pp. 136–162). Sage.

13.

Brunsting

N. C.

Bettini

Rock

Common

E. A.

Royer

D. J.

Lane

K. L.

Xie

Chen

Zeng

(2023). Working conditions and burnout of special educators of students with EBD: Longitudinal outcomes. Teacher Education and Special Education, 46(1), 44–64. https://doi.org/10.1177/08884064221076159

14.

Brunsting

N. C.

Bettini

Rock

M. L.

Royer

D. J.

Common

E. A.

Lane

K. L.

Xie

Chen

Zeng

(2022). Burnout of special educators serving students with emotional-behavioral disorders: A longitudinal study. Remedial and Special Education, 43(3), 160–171. https://doi.org/10.1177/07419325211030562

15.

Brunsting

N. C.

Sreckovic

M. A.

Lane

K. L.

(2014). Special education teacher burnout: A synthesis of research from 1979 to 2013. Education and Treatment of Children, 37(4), 681–711. https://doi.org/10.1353/etc.2014.0032

16.

Burke

R. J.

Greenglass

(1995). A longitudinal study of psychological burnout in teachers. Human Relations, 48(2), 187–202. https://doi.org/10.1177%2F001872679504800205

17.

Carver-Thomas

Darling-Hammond

(2019). The trouble with teacher turnover: How teacher attrition affects students and schools. Education Policy Analysis Archives, 27(36), 1–32. http://doi.org/10.14507/epaa.27.3699

18.

Cattell

R. B.

(1966). The scree test for the number of factors. Multivariate Behavioral Research, 1(2), 245–276. https://doi.org/10.1207/s15327906mbr0102_10

19.

Cece

Guillet-Descas

Lentillon-Kaestner

(2021). The longitudinal trajectories of teacher burnout and vigour across the scholar year: The predictive role of emotional intelligence. Psychology in the Schools, 59(3), 589–606. https://doi.org/10.1002/pits.22633

20.

Chen

F. F.

(2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504. https://doi.org/10.1080/10705510701301834

21.

Cheung

G. W.

Rensvold

R. B.

(2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 9(2), 233–255. https://doi.org/10.1207/S15328007SEM0902_5

22.

Cole

D. A.

Preacher

K. J.

(2014). Manifest variable path analysis: Potentially serious and misleading consequences due to uncorrected measurement error. Psychological Methods, 19(2), 300–315. https://doi.org/10.1037/a0033805

23.

Cormier

C. J.

McGrew

Ruble

Fischer

(2022). Socially distanced teaching: The mental health impact of the COVID-19 pandemic on special education teachers. Journal of Community Psychology, 50(3), 1768–1772. https://doi.org/10.1002/jcop.22736

24.

Cormier

C. J.

Scott

L. A.

Cornelius

K. E.

Rosenberg

M. S.

(2023). Achieving a more diverse special education teacher workforce: Guiding questions for researchers and policymakers. Multicultural Learning and Teaching, Advance online publication. https://doi.org/10.1515/mlt-2022-0007

25.

Cormier

C. J.

Scott

L. A.

Powell

Hall

(2022). Locked in glass classrooms: Black male special education teachers socialized as everything but educators. Teacher Education and Special Education, 45(1), 77–94. https://doi.org/10.1177/08884064211061038

26.

Cormier

C. J.

Wong

McGrew

J. H.

Ruble

L. A.

Worrell

F. C.

(2021). Stress, burnout, and mental health among teachers of color. The Learning Professional, 42(1), 54–62.

27.

Covell

McNeil

J. K.

Howe

R. B.

(2009). Reducing teacher burnout by increasing student engagement. School Psychology International, 30(3), 282–290. https://doi.org/10.1177/0143034309106496

28.

Demerouti

Bakker

A. B.

(2008). The Oldenburg Burnout Inventory: A good alternative to measure burnout and engagement. In Halbesleben

(Ed.), Handbook of stress and burnout in health care (pp. 65–78). Nova Science Publishers.

29.

Demerouti

Bakker

A. B.

Vardakou

Kantas

(2003). The convergent validity of two burnout instruments: A multitrait-multimethod analysis. European Journal of Psychological Assessment, 19(1), Article 12. https://doi.org/10.1027/1015-5759.19.1.12

30.

Demerouti

Mostert

Bakker

A. B.

(2010). Burnout and work engagement: A thorough investigation of the independency of both constructs. Journal of Occupational Health Psychology, 15(3), Article 209. https://doi.org/10.1037/a0019408

31.

Drame

E. R.

Pierce

N. P.

Cairo

(2022). Black special education teacher eductors’ practice of resistance. Teacher Education and Special Education, 45(1), 27–42. https://doi.org/10.1177/08884064211070570

32.

Dykstra

J. R.

Sabatos-DeVito

M. G.

Irvin

D. W.

Boyd

B. A.

Hume

K. A.

Odom

S. L.

(2013). Using the Language Environment Analysis (LENA) system in preschool classrooms with children with autism spectrum disorders. Autism, 17(5), 582–594. https://doi.org/10.1177/1362361312446206

33.

Fraiman

Y. S.

Cheston

C. C.

Cabral

H. J.

Allen

Asnes

A. G.

Barrett

J. T.

Batra

Bernstein

Bleeker

Dietz

P. M.

Lewis

S.-T. T.

Marsh Ma

Mahan

J. D.

Michelson

C. D.

Poynter

S. E.

Vining

M. A.

Watson

Sox

C. M.

(2022). Effect of a novel mindfulness curriculum on burnout during pediatric internship: A cluster randomized clinical trial. Journal of American Medical Association Pediatrics, 176(4), 365–372. https://doi.org/10.1001/jamapediatrics.2021.5740

34.

Frank

A. R.

McKenzie

(1993). The development of burnout among special educators. Teacher Education and Special Education, 16(2), 161–170. https://doi.org/10.1177/088840649301600208

35.

Garwood

J. D.

Werts

M. G.

Varghese

Gosey

(2018). Mixed-methods analysis of rural special educators’ role stressors, behavior management, and burnout. Rural Special Education Quarterly, 37(1), 30–43. http://doi.org/10.1177/8756870517745270

36.

Gustavson

von Soest

Karevold

Røysamb

(2012). Attrition and generalizability in longitudinal studies: Findings from a 15-year population-based study and a Monte Carlo simulation study. BMC Public Health, 12, Article 918. https://doi.org/10.1186/1471-2458-12-918

37.

Hakanen

J. J.

Schaufeli

W. B.

Ahola

(2008). The Job Demands-Resources Model: A three-year cross-lagged study of burnout, depression, commitment, and work engagement. Work & Stress, 22(3), 224–241. https://doi.org/10.1080/02678370802379432

38.

Halbesleben

J. R. B.

Demerouti

(2005). The construct validity of an alternative measure of burnout: Investigating the English translation of the Oldenburg Burnout Inventory. Work & Stress, 19(3), 208–220. https://doi.org/10.1080/02678370500340728

39.

Hawrot

Koniewski

(2018). Factor structure of the Maslach Burnout Inventory–Educators Survey in a Polish-speaking sample. Journal of Career Assessment, 26(3), 515–530. https://doi.org/10.1177/1069072717714545

40.

Hopman

J. A.

Tick

N. T.

van der Ende

Wubbels

Verhulst

F. C.

Maras

Breeman

L. D.

van Lier

P. A.

(2018). Special education teachers’ relationships with students and self-efficacy moderate associations between classroom-level disruptive behaviors and emotional exhaustion. Teaching and Teacher Education, 75, 21–30. https://doi.org/10.1016/j.tate.2018.06.004

41.

L. T.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6(1), 1–55. https://doi.org/10.1080/10705519909540118

42.

Hultell

Melin

Gustavsson

J. P.

(2013). Getting personal with teacher burnout: A longitudinal study on the development of burnout using a person-based approach. Teaching and Teacher Education, 32, 75–86. https://doi.org/10.1016/j.tate.2013.01.007

43.

Jennings

P. A.

Greenberg

M. T.

(2009). The prosocial classroom: Teacher social and emotional competence in relation to student and classroom outcomes. Review of Educational Research, 79(1), 491–525. https://doi.org/10.3102/0034654308325693

44.

Jorgensen

T. D.

(2017). Applying permutation tests and multivariate modification indices to configurally invariant models that need respecification. Frontiers in Psychology, 8, Article 1455. https://doi.org/10.3389/fpsyg.2017.01455

45.

Karavasilis

(2019). Work satisfaction or burnout and their impact on innovative work behavior of Greek teachers. Journal of Contemporary Education Theory & Research, 3(2), 3–10. https://doi.org/10.5281/zenodo.3635027

46.

Kestian

J. G.

(2020). The efficacy of interventions for reducing teacher burnout: A meta-analysis [Doctoral dissertation, Oaktrust Texas A&M Repository, Texas A&M University].

47.

Kristensen

T. S.

Borritz

Villadsen

Christensen

K. B.

(2005). The Copenhagen Burnout Inventory: A new tool for the assessment of burnout. Work & Stress, 19(3), 192–207. https://doi.org/10.1080/02678370500297720

48.

Kulkarni

S. S.

Bland

Gaeta

J. M.

(2022). From support to action: A critical affinity group of special education teachers of color. Teacher Education and Special Education, 45(1), 43–60. https://doi.org/10.1177%2F08884064211061189

49.

Langher

Caputo

Ricci

M. E.

(2017). The potential role of perceived support for reduction of special education teachers’ burnout. International Journal of Educational Psychology, 6(2), 120–147. https://doi.org/10.17583/ijep.2017.2126

50.

Lilienfeld

S. O.

Strother

A. N.

(2020). Psychological measurement and the replication crisis: Four sacred cows. Canadian Psychology/Psychologie Canadienne, 61(4), Article 281. https://doi.org/10.1037/cap0000236

51.

Madigan

D. J.

Kim

L. E.

(2021). Does teacher burnout affect students? A systematic review of its association with academic achievement and student-reported outcomes. International Journal of Educational Research, 105, Article 101714. https://doi.org/10.1016/j.ijer.2020.101714

52.

Maricuţoiu

L. P.

Pap

Ștefancu

Mladenovici

Valache

D. G.

Popescu

B. D.

Ilie

M. D.

Vîrgă

(2023). Is teachers’ well-being associated with students’ school experience? A meta-analysis of cross-sectional evidence. Educational Psychology Review, 35, Article 1. https://doi.org/10.1007/s10648-023-09721-9

53.

Maslach

Jackson

S. E.

Leiter

M. P.

Schaufeli

W. B.

Schwab

R. L.

(1986). Maslach Burnout Inventory. Consulting Psychologists Press.

54.

McGrew

Ruble

Cormier

C. J.

Dueber

(2023). Special educators’ mental health and burnout: A comparison of general and teacher specific risk factors. Teaching and Teacher Education, 132, Article 104209. https://doi.org/10.1016/j.tate.2023.104209

55.

McLeskey

Tyler

N. C.

Saunders Flippin

(2004). The supply of and demand for special education teachers: A review of research regarding the chronic shortage of special education teachers. Journal of Special Education, 38(1), 5–21. https://doi.org/10.1177/00224669040380010201

56.

Montero-Marin

Taylor

Crane

Greenberg

M. T.

Ford

T. J.

Williams

J. M. G.

García-Campayo

Sonley

Lord

Dalgleish

Blakemore

S. J.

, MYRIAD Team, & Kuyken

(2021). Teachers “finding peace in a frantic world”: An experimental study of self-taught and instructor-led mindfulness program formats on acceptability, effectiveness, and mechanisms. Journal of Educational Psychology, 113(8), 1689–1708. https://doi.org/10.1037/edu0000542

57.

Muthén

L. K.

Muthén

B. O.

(2023). Mplus user’s guide (8th ed.). Muthén & Muthén.

58.

Nichols

A. S.

Sosnowsky

F. L.

(2002). Burnout among special education teachers in self-contained cross-categorical classrooms. Teacher Education and Special Education, 25(1), 71–86. https://doi.org/10.1177/088840640202500108

59.

Park

E. Y.

Shin

(2020). A meta-analysis of special education teachers’ burnout. SAGE Open, 10(2), 1–18. http://doi.org/10.1177/2158244020918297

60.

Pressley

(2021). Factors contributing to teacher burnout during COVID-19. Educational Researcher, 50(5), 325–327. https://doi.org/10.3102/0013189x211004138

61.

R Core Team. (2023). R: A language and environment for statistical computing. R Foundation for Statistical Computing. https://www.R-project.org/

62.

Robinson

O. P.

Bridges

S. A.

Rollins

L. H.

Schumacker

R. E.

(2019). A study of the relation between special education burnout and job satisfaction. Journal of Research in Special Educational Needs, 19(4), 295–303. https://doi.org/10.1111/1471-3802.12448

63.

Schaufeli

W. B.

Bakker

A. B.

Salanova

(2006). The measurement of work engagement with a short questionnaire: A cross-national study. Educational and Psychological Measurement, 66(4), 701–716. https://doi.org/10.1177%2F001316440528247

64.

Schaufeli

W. B.

Enzmann

(1998). The burnout companion to study and practice: A critical analysis. Taylor & Francis.

65.

Schwarzer

Schmitz

G. S.

Tang

C. S.-K.

(2000). Teacher burnout in Hong Kong and Germany: A cross-cultural validation of the Maslach Burnout Inventory. Anxiety, Stress, & Coping, 13(3), 309–326. https://doi.org/10.1080/10615800008549268

66.

Scott

L. A.

Cormier

C. J.

Boveda

(2022). Critical issues for the preparation and workforce development of racialized special educators. Teacher Education and Special Education, 45(1), 5–7. https://doi.org/10.1177%2F08884064211070571

67.

Scott

L. A.

Powell

Oyefuga

Cormier

C. J.

Padhye

(2021). Complementary review of the literature on attrition and retention patterns of special education teachers of color: What we know and how we move forward. Multiple Voices, Disability, Race and Language Intersections in Special Education, 21(1), 3–39. https://doi.org/10.56829/2158-396X-21.1.3

68.

Seppälä

Mauno

Feldt

Hakanen

Kinnunen

Tolvanen

Schaufeli

(2009). The construct validity of the Utrecht Work Engagement Scale: Multisample and longitudinal evidence. Journal of Happiness Studies, 10, 459–481. https://doi.org/10.1007/s10902-008-9100-y

69.

Simbula

Guglielmi

Schaufeli

Depolo

(2013). The Italian validation of the Utrecht Work Engagement Scale: Characterization of engaged groups in a sample of school teachers. Bollettino di Psicologia Applicata, 268, 43–54.

70.

Singer

J. D.

(1993). Once is not enough: Former special educators who return to teaching. Exceptional Children, 60(1), 58–72. https://doi.org/10.1177/001440299306000106

71.

Siuty

M. B.

Atwood

(2022). Intersectional disruptor: A special educator of color living and teaching in the intersections. Teacher Education and Special Education, 45(1), 61–76. https://doi.org/10.1177/08884064211062872

72.

Squillaci

(2020). Parents-teachers’ relationships: Effects on the burnout of special education teachers. La Revue Internationale de Leducation Familiale, 47(1), 151–168. https://doi.org/10.3917/rief.047.0151

73.

Squillaci

(2021). Analysis of the burnout levels of special education teachers in Switzerland in link with a reform implementation. European Journal of Special Needs Education, 36(5), 844–853. https://doi.org/10.1080/08856257.2020.1809802

74.

Sutcliffe

J. P.

(1980). On the relationship of reliability to statistical power. Psychological Bulletin, 88(2), 509–515. https://doi.org/10.1037/0033-2909.88.2.509

75.

Szigeti

Balázs

Bikfalvi

Urbán

(2017). Burnout and depressive symptoms in teachers: Factor structure and construct validity of the Maslach Burnout Inventory-Educators Survey among elementary and secondary school teachers in Hungary. Stress and Health, 33(5), 530–539. https://doi.org/10.1002/smi.2737

76.

Taris

T. W.

Le Blanc

P. M.

Schaufeli

W. B.

Schreurs

P. J.

(2005). Are there causal relationships between the dimensions of the Maslach Burnout Inventory? A review and two longitudinal tests. Work & Stress, 19(3), 238–255. https://doi.org/10.1080/02678370500270453

77.

Trainor

A. A.

Bettini

Scott

L. A.

(2019). Introduction to the special issue—A necessary step in pursuit of equity: Developing a racially/ethnically diverse special education teaching force. Remedial and Special Education, 40(4), 195–198. https://doi.org/10.1177%2F0741932519843176

78.

U.S. Department of Education. (2020). Teacher shortage areas. https://tsa.ed.gov/#/reports

79.

Wheeler

D. L.

Vassar

Worley

J. A.

Barnes

L. L. B.

(2011). A meta-analytic reliability generalization meta-analysis of coefficient alpha for the Maslach Burnout Inventory. Educational and Psychological Measurement, 71(1), 231–244. https://doi.org/10.1177%2F0013164410391579

80.

Worley

J. A.

Vassar

Wheeler

D. L.

Barnes

L. L.

(2008). Factor structure of scores from the Maslach Burnout Inventory: A review and meta-analysis of 45 exploratory and confirmatory factor-analytic studies. Educational and Psychological Measurement, 68(5), 797–823. https://doi.org/10.1177/0013164408315268

81.

Zabel

R. H.

Zabel

M. K.

(2002). Burnout among special education teachers and perceptions of support. Journal of Special Education Leadership, 15(2), 67–73. https://doi.org/10.1177/088840648300600405