Measuring autistic burnout: A psychometric validation of the AASPIRE Autistic Burnout Measure in autistic adults

Abstract

Autistic burnout is characterised by extreme exhaustion, loss of functioning, and reduced tolerance to stimulus, resulting from the cumulative stress associated with navigating a predominantly non-autistic world. To date, in mostly qualitative studies, autistic burnout has been associated with poorer mental health, well-being and life outcomes in autistic adults. To comprehensively investigate autistic burnout, identify affected individuals and evaluate supports, a valid and reliable measure is required. The current study explored the psychometric properties of the AASPIRE Autistic Burnout Measure. The Autistic Burnout Measure and other related measures (camouflaging, mental health) were completed online by 379 autistic adults. The Autistic Burnout Measure demonstrated a predominantly unidimensional structure, with high loadings across all 27 items, excellent internal consistency (ω = 0.98), and reasonable consistency over 12 months (r = 0.59). It also showed sound construct validity, with medium-to-large positive correlations with autistic traits, camouflaging, occupational burnout, depression and anxiety. The Autistic Burnout Measure also effectively differentiated between autistic participants who reported currently experiencing autistic burnout and those who were not (area under the curve = 0.92; 95% confidence interval = [0.86, 0.97]). Our findings indicate that the Autistic Burnout Measure has promising psychometric properties and may be a useful measure in future autism research and practice. However, further validation is necessary to determine whether the unidimensional structure holds across diverse samples.

Lay abstract

Autistic people have described autistic burnout as an intense experience of physical, emotional, mental and social exhaustion impacting their ability to complete everyday tasks and contributing to poorer well-being. To identify and measure autistic burnout in practice and research, we need a self-report measure that gives accurate and consistent results. In this study, 379 autistic adults completed a recently developed measure of autistic burnout online, the AASPIRE Autistic Burnout Measure. We analysed their ratings to determine whether the measure is reliable (i.e. ratings are consistent), valid (i.e. the tool measures what it says it measures), correctly identifies those currently experiencing burnout, and is associated with other relevant experiences, such as camouflaging, anxiety and depression. The Autistic Burnout Measure was found to be reliable and valid. Autistic adults reporting greater autistic burnout also reported more camouflaging, autistic traits and greater general burnout, depression, and anxiety. The Autistic Burnout Measure was accurate in identifying individuals who reported currently experiencing autistic burnout and those who did not. Overall, our findings suggest that the Autistic Burnout Measure may be suitable for use in research and practice to identify and better understand experiences of autistic burnout.

Initially emerging as a term discussed within online autistic communities, autistic burnout refers to a chronic state of incapacity, exhaustion, and distress stemming from the daily challenges autistic people face in navigating a predominantly non-autistic world (Raymaker et al., 2020). In the first study on autistic burnout, Raymaker et al. (2020) and the AASPIRE (Academic Autism Spectrum Partnership in Research and Education) team conducted an exploratory qualitative study with 19 autistic adults and narratives from 19 online repositories, blogs, and forums to develop a definition of autistic burnout. Through thematic analysis, autistic burnout was identified as a distinct phenomenon characterised by prolonged exhaustion (typically exceeding 3 months), loss of function (including difficulties in social interaction, communication, executive/cognitive functioning, and activities of daily living), and reduced tolerance to stimuli (such as intolerance to sensory input; Raymaker et al., 2020).

Subsequently, Higgins et al. (2021) used a grounded Delphi method to co-produce a definition of autistic burnout with autistic adults as experts by lived experience. Both Raymaker et al.’s (2020) and Higgins et al.’s (2021) definitions identified key characteristics of autistic burnout – such as exhaustion, loss of function, executive functioning difficulties, and heightened sensory sensitivity (Arnold et al., 2023b). Higgins et al. (2021) explicitly included interpersonal withdrawal as a defining characteristic of autistic burnout, which was not initially emphasised in Raymaker et al. (2020). With regard to the duration or frequency of autistic burnout, Higgins et al. (2021) noted that autistic adults in their study provided mixed reports regarding the duration of autistic burnout, varying from hours to days to weeks, months or years, with chronic phases lasting up to ‘5 years or more’ (p. 2362). Raymaker et al. (2020) described burnout as a chronic, pervasive, long-term experience of ‘typically 3+ months’ (p. 140). Both conceptualisations described burnout as distinct from depression and non-autistic burnout, particularly in its onset, symptoms and the specific causes that are related to the challenges autistic individuals face navigating an often hostile and unaccommodating neurotypical world.

Drawing from the perspectives and experiences of autistic adults, subsequent qualitative research explored potential risk and protective factors associated with autistic burnout. For example, Mantzalas et al. (2021) analysed 1127 online posts from a Twitter forum community for autistic people to investigate risk and protective factors and consequences of autistic burnout. Based on their findings, the authors endorsed Raymaker et al.’s (2021) definition of autistic burnout and highlighted camouflaging or ‘masking’ ¹ (i.e. the use of social strategies to minimise the visibility of autistic traits to blend in and avoid victimisation in a largely neurotypical society; Hull et al., 2019) as a key risk factor of autistic burnout (see also Zhuang et al., 2023 for a critical review of psychosocial correlates and consequences of camouflaging in autistic adults, which identifies burnout as a consequence based on camouflaging studies to date). Building on their earlier work, Mantzalas et al. (2022) proposed a conceptual model of autistic burnout that encompassed various individual, social and environmental risk factors, such as the social stressors experienced by autistic people (e.g. stigmatisation and lack of support) as predisposing risk factors, and consequences (e.g. depression and anxiety).

These qualitative studies were the first to explore the construct, characteristics and experiences of autistic burnout in the research literature. To further explore the experiences of autistic burnout, its risk and protective factors and relationships to mental health outcomes, quantitative research with valid and reliable measures of autistic burnout is needed.

Measuring autistic burnout: psychometric evidence to date

To date, two measures of autistic burnout have been developed. The first measure, the 27-item Autistic Burnout Measure (ABM), was developed by the Academic Autism Spectrum Partnership in Research and Education (AASPIRE), ² a team including autistic people, academic researchers, family members, disability professionals and clinicians (Raymaker et al., personal communication, September 19, 2020). Then, in 2023, Arnold and colleagues published the initial development and validation data of a second measure, the Autistic Burnout Severity Items (ABSI; Arnold et al., 2023b).

ABM

Drawing from the knowledge and expertise of Raymaker et al.’s (2020) team members and their earlier qualitative work, the ABM was developed and designed to be unidimensional (i.e. reflecting a single underlying factor), with all items summed to produce a total autistic burnout score. In their pilot study (Raymaker et al., personal communication), an initial draft of items was created based on the categories and characteristics defined in their qualitative research (Raymaker et al., 2020). A total of 80 autistic adults were then invited to rate these items by reflecting on their current experiences (within the past 3 months) compared with their ‘usual’ experiences (what they consider ‘normal’ or most typical for themselves). In this initial pilot study, the ABM demonstrated promising face and content validity and excellent internal consistency (α = 0.95). It also showed reasonable construct validity, evidencing large positive associations with depression (r = 0.56), as measured by the Centre for Epidemiological Studies Depression Scale (Andresen et al., 1994), and stress (r = 0.68), as measured with the Perceived Stress Scale (Cohen et al., 1983), although there was no significant association between the ABM and camouflaging (r = 0.04), measured by the Camouflaging Autistic Traits Questionnaire (CAT-Q; Hull et al., 2019). Furthermore, based on participants’ responses to a single item asking whether they were currently experiencing autistic burnout (with their responses used as the criterion), the ABM was very good at accurately classifying people who reported experiencing versus not currently experiencing autistic burnout (AUC = 0.85, 95% CI = [0.75, 0.97]). However, the measure’s structural validity was not assessed due to the small sample size. In addition, no general or occupational burnout measures were included to establish convergent validity with other measures of burnout.

ABSI

To develop the ABSI, Arnold et al. (2023a, 2023b) invited autistic adults to provide open-ended responses about their experiences of autistic burnout. They explored aspects such as the onset, duration, frequency and consequences of burnout. Based on the insights gathered, they developed a survey comprising 48 items that captured key characteristics of autistic burnout identified in Raymaker et al. (2020) and Higgins et al. (2021). As their participants indicated considerable variability in duration and frequency of autistic burnout experiences, the ABSI asked respondents to reflect on their most recent or current episode of autistic burnout without specifying its duration. One hundred and forty-one autistic adults rated the 48 items, with the ratings then subjected to exploratory factor analysis (EFA) to identify groups of related symptoms. Participants also completed other self-report measures of camouflaging and mental health outcomes to assess construct validity. The resulting 20-item ABSI was reported to have four factors: exhaustion, cognitive disruption, heightened autistic self-awareness, and overwhelm and withdrawal (Arnold et al., 2023b). However, the authors conducted a bifactor analysis that did not converge, possibly due to the small sample size relative to the large number of items in the measure, leaving unresolved questions about the ABSI’s dimensionality. The ABSI demonstrated sound overall internal consistency (α = 0.88) and acceptable internal reliability across its subscales, with Cronbach’s alpha values ranging from 0.73 to 0.86.

Arnold et al. (2023b) tested the ABM alongside their ABSI to assess how both measures perform in measuring autistic burnout. Total scores for the ABM and ABSI were only moderately correlated (r = 0.55), suggesting that while both measures overlap, they likely capture different aspects of autistic burnout. This moderate correlation may also be influenced by variations in the duration of experiences evaluated by each measure. Both the ABM and ABSI showed significant positive associations with depression (ABM r = 0.67; ABSI r = 0.48). Furthermore, the ABSI exhibited a positive correlation with the CAT-Q at a medium effect (r = 0.34), whereas the ABM did not (r = 0.07).

More recently, Mantzalas et al. (2024) assessed the psychometric properties of the ABM in a sample of 238 autistic adults 18 to 75 years old (71% female, 17% male, 10% non-binary). For the ABM, a series of factor models were explored, and a hierarchical factor model was found to be the most parsimonious, with a higher-order ‘Autistic Burnout’ factor influencing four lower-order factors (‘Cognitive and Functioning Difficulty’, ‘Emotional and Sensory Dysregulation, ‘Avoidance and Exhaustion’ and ‘Social and Communication Difficulty’). The ABM was essentially unidimensional, with the higher-order ‘Autistic Burnout’ factor accounting for 77% of the total variance in ABM scores (Cronbach’s α = 0.95 for the full scale), supporting the interpretability of the total score.

Mantzalas and colleagues (2024) assessed the ABM alongside a well-validated measure of general burnout, the Copenhagen Burnout Inventory (CBI) (Kristensen et al., 2005). When completed by autistic adults, factor analytic evidence supported a general ‘Personal’ subscale (CBI-P), made up of two lower-order factors called Emotional Exhaustion (CBI-P-E) and Physical Exhaustion (CBI-P-P). The CBI-P-E subscale, in particular, showed reasonable specificity (area under the curve (AUC) = 0.767), performing similarly to the ABM (AUC = 0.789) in differentiating between those who reported currently experiencing autistic burnout from those who reported they did not.

The ABM demonstrated sound construct validity in their study, as it was highly positively correlated with depression, anxiety, stress, and fatigue self-report measures (with r = 040–0.59, suggesting distinct but interrelated constructs) and moderately positively correlated with camouflaging (r = 0.36). However, the associations between the ABM and the general personal burnout subscales (r = 0.48–0.49) were somewhat weaker than its association with depression (r = 0.59), suggesting that the burnout measures converged no better than they did with depression in this study, which could potentially be problematic for convergent validity.

Since its original development, the ABM has been used to measure autistic burnout in a Polish sample of autistic adults (Pyszkowska, 2024) and autistic women using the Dutch translation (Schoondermark et al., 2024). Both studies reported excellent reliability, with Cronbach’s alpha values of 0.95 and 0.98, respectively. The ABM correlated 0.3 with the CAT-Q (Pyszkowska, 2024) and 0.68 to 0.7 with depression and anxiety (Schoondermark et al., 2024). However, neither study examined the structural validity of the measure.

The present study

Emerging literature has highlighted autistic burnout as a shared experience among many autistic adults. Since 2020, efforts have been made to understand these experiences better and to develop and validate two measures of autistic burnout: the Autistic Burnout Screening Inventory (ABSI) and the ABM. So far, there have been two psychometric evaluations of the ABM’s properties (Mantzalas et al., 2024; Schoondermark et al., 2024), with only Mantzalas et al. (2024) investigating the measure’s factor structure. Overall, the ABM seems to be a promising tool for measuring burnout in autistic adults. However, confirmatory factor analysis (CFA) has not yet confirmed its factor structure, and test–retest consistency/ reliability has not been established. Further independent psychometric evaluation and replication of the measure’s psychometric properties is desirable, especially using a larger sample of autistic adults.

The current study therefore had four aims in investigating the psychometric properties of the ABM. First, we aimed to examine the structural validity of the ABM. Initially, when this study was conducted, we explored the measure’s factor structure through EFA, as no other study had investigated the ABM’s factor structure at that time. However, while the present study was under peer review, Mantzalas et al. (2024) published their work on the ABM’s factor structure, providing initial evidence for a unidimensional structure. Subsequently, and in light of these findings, we then revised this analytical aim to test the unidimensional, correlated four-factor and bifactor models proposed by Mantzalas et al. (2024), followed by exploring alternative models to identify the most parsimonious solution in our sample. The second aim was to investigate the internal consistency and test re-test reliability of the ABM, based on the most parsimonious factor model derived from the factor analyses. Third, we tested the relationship between autistic burnout and self-report measures of related constructs (autistic traits, general burnout, camouflaging, depression and anxiety) to assess construct validity. The fourth aim was to test the ABM’s ability to differentiate between individuals who currently reported experiencing autistic burnout and those who did not.

Method

Participants and recruitment

Participants were autistic adults who were (1) 18 years or older, (2) proficient in reading and writing English, and (3) either formally diagnosed or self-identifying as autistic. Participants were recruited internationally via social media (i.e. Twitter, Facebook and Instagram), email distributions or website announcements of autism organisations, and the online platform for crowd-sourcing participants, Prolific (https://www.prolific.com/). Participants were informed that they were being invited to participate in a research project exploring psychosocial experiences and correlates of camouflaging (including mental health and well-being, as well as autistic burnout in autistic adults). Recruitment occurred across multiple time points (see Figure 1). Data from 230 autistic adults were collected as part of a larger camouflaging research project in May-August 2021. Approximately 12 months later (May–September 2022), 133 of the initial 230 participants (58%) completed a follow-up survey, which allowed us to assess the ABM’s test–retest reliability. Independently from the larger research project, an additional 149 autistic adults were recruited in February 2023 via Prolific to increase the sample size from 230 to 379 for factor analysis. ³

Figure 1.

Participant recruitment and key demographic characteristics of the different data collection groups.

Participants were removed if the following conditions were observed: (1) did not meet inclusion criteria; (2) responded ‘no’ to a question about their data being valid for research use; (3) failed any attention check and ReCAPTCHA questions; (4) insufficient questionnaire completion ⁴ ; and (5) indicated ‘prefer not to say’ for > 20% of responses (see Figure 1 for the breakdown of exclusions and reasons).

The final combined sample of 379 autistic adults were aged 18 to 77 years (M = 33.47, SD = 10.72), with the majority being employed, university-educated and living in Australia (see Table 1 for participant characteristics). Around 60% reported at least one co-occurring mental health diagnosis (primarily depression and anxiety). Most reported a professional autism diagnosis (n = 277; 73.1%), with an age of diagnosis ranging from 2 to 61 years (M = 24.73, SD = 13.36), while 102 (26.9%) self-identified as autistic. Participants completed the Broad Autism Phenotype Questionnaire (BAPQ; Hurley et al., 2007) and BAPQ total scores did not differ significantly between the professionally diagnosed (Mdn = 4.28) and self-identified (Mdn = 4.3) participants (U = 14511.50, z = 0.41, p = 0.68). ⁵

Table 1.

Participant characteristics (N = 379).

Characteristics	N	%
Gender
Male	176	46.4%
Female	167	44.1%
Other Gender Expressions a	33	8.7%
Preferred not to say	3	0.8%
Ethnicity b
White	194	84.3%
Asian	17	7.4%
Multiethnic	12	5.2%
Black/African/Caribbean	5	2.2%
Hispanic/Latino	2	0.9%
Country of Residence
Australia	135	35.6%
United Kingdom	56	14.8%
United States of America	60	15.8%
Central America	23	6.1%
South America	12	3.2%
New Zealand	10	2.6%
Canada	16	4.2%
South Africa	23	6.1%
Asian Countries (Singapore, Hong Kong)	2	0.5%
Other European Countries (e.g. Sweden, Hungary, Poland)	42	11.1%
Highest Education Level
Primary School	11	2.9%
Secondary School	86	22.7%
Trade/ technical certificate	47	12.4%
Undergraduate degree	132	34.8%
Postgraduate degree	84	22.2%
Other qualifications (e.g. diploma, A-level certificate)	17	4.5%
Prefer not to say	2	0.5%
Employment Status
Employed full-time	128	33.8%
Employed part-time	83	21.9%
Unemployed	68	17.9%
Student	62	16.4%
Homemaker/Caregiver	21	5.5%
Unpaid (e.g. volunteering, or unpaid internship)	8	2.1%
Retired	6	1.6%
Preferred not to say	3	0.8%
Other Neurodevelopmental Diagnoses
No	274	72.3%
Yes	96	25.3%
Attention Deficit/ Hyperactivity Disorder	76 c	20.1%
Specific Learning Disorder	20	5.3%
Intellectual Disability	4	1.1%
Other (e.g. cerebral palsy, developmental coordination disorder)	16	4.2%
Preferred not to say	9	2.4%
Diagnosed with a Mental Health Condition (in the last 5 years)
No	145	38.3%
Yes	225	59.4%
Anxiety	172 c	45.4%
Depression	148	39.1%
Post-traumatic stress disorder	31	8.2%
Bipolar disorders	7	1.9%
Eating disorders	4	1.1%
Personality disorders	6	1.6%
Obsessive-compulsive disorders	13	3.4%
Other (e.g. psychotic disorders, dissociative disorders).	13	3.4%
Preferred not to say	9	2.4%
Diagnosed with a physical or medical condition
No	229	60.4%
Yes (e.g. autoimmune disease, diabetes etc.)	132	34.8%
Preferred not to say	18	4.7%

^aOther genders include non-binary (n = 22), Agender (n = 2), Demi-female (n = 1), Questioning (n = 2), Neutral (n = 1), Gender-fluid (n = 3) and Genderqueer (n = 2).

^bOnly individuals in the initial data collection were asked this question (n = 230).

^cN adds up to more than the total N of the ‘yes’ participants, as some reported more than one neurodevelopmental and/or mental health conditions.

Procedure

Ethics approval was obtained from the University of Western Australia Human Research Ethics Office (Reference: RA/4/2021/ET000065).

The study was hosted on the Qualtrics survey platform, where participants answered demographic questions and completed self-report measures (see the ‘Measures’ section). All participants provided informed consent and could choose ‘prefer not to say’ for any question. The initial group (n = 230) and the follow-up participants (n = 130) answered longer surveys for the larger project and received AUD20 reimbursement. The additional participants recruited for the factor analysis (n = 149) filled in a shorter survey focusing on burnout and mental health measures only and received AUD10 reimbursement.

To identify invalid responses and enhance data quality, reCAPTCHA and attention check questions were included. Participants had to achieve a reCAPTCHA score of >0.70 ⁶ and pass attention check questions ⁷ to ensure they were not bots. Finally, upon survey completion, participants were asked to indicate ‘yes or ‘no’ to whether they thought their data were valid for research (if they indicated no, their data were excluded from the analyses, as their response likely indicated invalid data, but they were informed that indicating ‘no’ would not affect their eligibility to be reimbursed for their participation). Accordingly, 379 participants were included in the data analysis (with 130 providing test–retest data).

Measures

AASPIRE Autistic Burnout Measure

The ABM is a 27-item self-report measure of autistic burnout. Respondents are asked to ‘compare what you are currently experiencing in your life with what you would consider to be your usual experience’. The instructions outline that ‘currently’ means within the past 3 months, and ‘usual’ means whatever is considered normal or most typical for each person. Items are prefaced with ‘In the past three months . . .’. and a summary of the items can be seen in Table 2. Individuals are not provided with a definition of autistic burnout or asked about what they understand by the term autistic burnout before completing the questionnaire. Respondents are asked to rate their level of agreement on a five-point Likert-type scale, with ratings ranging from strongly agree (4) to strongly disagree (0). The item ratings are summed to a total score (0 to 108; higher scores = more burnout).

Table 2.

Model fit of the examined ABM models from Mantzalas et al. (2024).

	Model fit indices
Models	χ2(df)	RMSEA [90% CI]	CFI	TLI	SRMR
Unidimensional	2858.77 (324)*	0.147 [0.141, 0.152]	0.748	0.727	0.075
Correlated Four-factor	1566.55 (318)*	0.122 [0.116, 0.128]	0.828	0.810	0.056
Bifactor	1247.61 (297)*	0.110 [0.103, 0.116]	0.871	0.847	0.047

df = degrees of freedom, RMSEA = root mean square error approximation, CI = confidence intervals, CFI = comparative fit index, TLI = Tucker–Lewis index, SRMR = standardised root mean square residual. RMSEA, CFI and TLI are robust fit indices.

^*Indicates p < 0.05.

Current autistic burnout experience

To investigate the interpretability of the ABM, we included an item in the follow-up data collection survey only (n = 133) asking participants, ‘Do you think that you are experiencing autistic burnout now?’. Response options included ‘Yes’, ‘No’, and ‘I do not know’.

Sydney Burnout Measure

To assess the ABM’s construct validity, the 34-item self-report SBM was administered to the additional study sample participants (n = 149) (Parker et al., 2023). ⁸ The Sydney Burnout Measure (SBM) was selected because although the items refer to the concept of ‘work’, this work is not restricted to an occupational context (Tavella & Parker, 2020b) and may include a current or previous job, home or care duties, or studying. Participants rate their current experience of burnout using a four-point Likert-type scale ranging from 0 (not true) to 3 (distinctly), with higher total scores (range = 0–102) indicating more significant burnout. Example items include ‘I feel emotionally drained and exhausted’ and ‘I find it hard to concentrate on the task at hand’. The SBM has demonstrated robust convergent validity, exhibiting a correlation of r = 0.73 with the well-established Maslach’s Burnout Inventory (Parker et al., 2023) and had excellent internal consistency in the current study (α = 0.98).

Patient Health Questionnaire

The Patient Health Questionnaire (PHQ-9) is a 9-item self-report measure of current depression symptom severity aligned with the Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM-5; American Psychiatric Association (APA), 2013) diagnostic criteria (Kroenke et al., 2001). Participants rate how often they have been bothered by problems such as ‘feeling down, depressed or hopeless’ over the past 2 weeks on a scale from 0 (not at all) to 3 (nearly every day; total score range 0–27). The PHQ-9 exhibits good sensitivity, specificity, and internal consistency (α = 0.89) in non-autistic and autistic samples (Arnold et al., 2020; Hull et al., 2019; Kroenke et al., 2001). In the current study, 254 (67%) of the participants scored above the screening cut-off score of 10, suggesting clinically elevated depressive symptoms suggesting possible depression.

Generalised Anxiety Disorder Questionnaire

For the 7–item Generalised Anxiety Disorder Questionnaire (GAD-7), participants rate the frequency of generalised anxiety symptoms over the last 2 weeks on a scale from 0 (not at all) to 3 (nearly every day; total score range 0–21) (Spitzer et al., 2006). The scale has good sensitivity and specificity, internal consistency (α = 0.92), and test–retest reliability of 0.83 in autistic and non-autistic samples (Hull et al., 2019; Spitzer et al., 2006).

Camouflaging of Autistic Traits Questionnaire

The Camouflaging of Autistic Traits Questionnaire (CAT-Q) is a 25-item self-report measure of camouflaging (Hull et al., 2019). Each item is rated on a 7-point Likert-type scale from strongly disagree (1) to strongly agree (7), with total scores ranging from 25 to 175 (higher scores = more camouflaging). The CAT-Q has excellent internal consistency (α = 0.94) and good test–retest reliability (r = 0.77) in autistic samples (Hull et al., 2019).

Broader Autism Phenotype Questionnaire

Rated on a 6-point Likert-type scale ranging from 1 (very rarely) to 6 (very often), the 36-item Broader Autism Phenotype Questionnaire (BAPQ) is a self-report measure of autistic social, communication and behavioural characteristics (Hurley et al., 2007). Originally developed as a measure of autistic-related traits in parents of autistic children, it has been used in some studies to measure autistic traits in autistic people, relatives of autistic people and in non-autistic adults (e.g. Hull et al., 2019; Nishiyama et al., 2014), despite some critique (see Piven & Sasson, 2014). Scores are averaged across the 36 items, so the average total score ranges from 1 to 6 (higher scores reflecting more autistic traits). The BAPQ has good sensitivity, specificity, and excellent internal consistency (α = 0.95–0.96) in autistic and non-autistic samples (Hull et al., 2019; Hurley et al., 2007), correlates 0.76 and 0.87 with the Autism Quotient (AQ) and the Social Responsiveness Scale (SRS-2) respectively (Nishiyama et al., 2014) and has evidenced large effect size group differences between autistic and non-autistic participants (Hull et al., 2019; Nishiyama et al., 2014).

Community involvement

The lead author is neurodivergent, although not autistic. The psychometric study is part of a broader research project on camouflaging and associated psychosocial correlates, which has engaged three autistic advisors who provided input on the study conceptualisation, relevance of psychosocial factors, and appropriateness of the survey, and who were reimbursed for their time.

Statistical analyses

Confirmatory factor analysis

A series of CFAs were performed in R [version 4.2.2] and RStudio [2022.07.02] (Macintosh Version 3.4.3; R Core Team, 2017) using Laavan (Rosseel, 2012) on the combined sample (N = 379) to assess the factor structure of the ABM. Based on the findings by Mantzalas et al. (2024), three models were investigated: (1) a unidimensional model whereby all ABM items load onto an autistic burnout factor; (2) a correlated four-factor model matching the four factors (i.e. Cognitive & functioning, Emotional & Sensory, Avoidance and Exhaustion and Social & Communication) identified in Mantzalas et al. (2024); and (3) a bifactorial four-factor model, which includes a general autistic burnout factor in addition to the four factors.

Given the ordinal item ratings, a polychoric correlation matrix was used, and factors were estimated using the weighted least squares (WLSMV) estimator (Fabrigar et al., 1999; Lloret et al., 2017). The goodness-of-fit of the models was judged based on the robust versions following fit indices: Comparative fit index (CFI), Tucker–Lewis index (TLI), root mean square error of approximation (RMSEA), and standardised root mean square residual (SRMR). For model evaluation, CFI and TLI values >0.95 indicate good fit, while RMSEA and SRMR values <0.06 indicate good fit (Hu & Bentler, 1999; Marsh et al., 2005; Savalei, 2021). To be preferred, the bifactor solution had to achieve a TLI value >0.01 over the hierarchical model (Gignac, 2016).

Exploratory factor analysis

Following CFAs, EFAs on the ABM were performed to investigate other potentially viable factor structures. An oblimin rotation method was used as it was reasonable to assume that factors would be inter-related (Fabrigar et al., 1999; Lloret et al., 2017). Factor extraction was based on several criteria, including Cattell’s scree test, parallel analysis, the Kaiser-Guttman criterion (Kaiser & Caffrey, 1965), Velicer’s MAP test (Velicer, 1976), Very Simple Structure (VSS; Revelle & Rocklin, 1979), and the Hull method (Auerswald & Moshagen, 2019). The interpretability of the extracted factors, based on the factor loadings, was also considered.

Subsequently, an exploratory bifactor analysis (BEFA) was conducted on the ABM in Mplus [version 8.5] using an orthogonal bi-Geomin rotation with the WLSMV estimator. The purpose of the BEFA was twofold: (1) to explore the feasibility of estimating both a general factor that influences all observed variables and subscales that are unique to certain subsets of items; and (2) to evaluate the replicability of these subscales when accounting for a general autistic burnout construct, as compared with factors extracted through an EFA (Reise et al., 2007; Revelle & Wilt, 2013).

Reliability and unidimensionality

McDonald’s coefficient omega (ω) provided the overall reliability of the total and subscale scores, with scores above 0.70 indicating good reliability (McDonald, 2013). McDonald’s coefficient omega hierarchical (ωH) provided the reliability of the ABM attributed to the general autistic burnout factor, while McDonald’s coefficient subscale omega hierarchical (ωHS) provided the reliability of the ABM’s subscales attributed to the specific factors. Explained common variances for the general factor (ECV_G) were also calculated. High ωH (e.g. >0.80) and ECV_G (e.g. >0.70) support the interpretation of unidimensionality (Reise, Scheines, et al., 2013; Stucky & Edelen, 2015). In addition, ωHS > 0.20 is sufficient to suggest that the subscales provide added value over a total score (see Dueber & Toland, 2023). Subscale coefficients were calculated based on the items meaningfully loading onto the specific factors (i.e. >0.20).

Finally, a two-way consistency intraclass correlation coefficient was calculated across two time points to assess the ABM’s test–retest reliability. Values of 0.50 to 0.75 and >0.75 indicate moderate and good reliability, respectively (Koo & Li, 2016).

Validity

Following the assessment of dimensionality and reliability, the construct and interpretability of the ABM were assessed. Construct validity was evaluated using Pearson r correlation coefficients between the total ABM score and total scores on measures of autistic-related characteristics (BAPQ), camouflaging (CAT-Q), anxiety (GAD-7), depression (PHQ-9), and general burnout (SBM). Effect sizes were interpreted according to Cohen (1988), where 0.20 to 0.49 indicates a small effect, 0.50 to 0.79 signifies a medium effect and ⩾0.80 is considered a large effect. Recent quantitative investigations (Arnold et al., 2023b; Mantzalas et al., 2024) led us to expect a positive (but not above 0.70, which is the threshold of collinearity; Carlson & Herdman, 2012) correlation between the ABM and measures of depression and anxiety.

Finally, Receiver Operating Characteristic (ROC) curve analyses were conducted to evaluate the interpretability of the ABM, specifically, the scale’s ability to discriminate between autistic people who indicated they were currently experiencing autistic burnout and those who were not. The ROC was plotted using the ABM total score (as the predictor) and participants’ responses to the single item that asked the respondents whether they were currently experiencing autistic burnout (as the criterion). Only the follow-up sample (Time 2) completed the single item (n = 133). Thirty-seven of the 133 participants responded with ‘I don’t know’ and were excluded from the ROC analysis, which was conducted with the remaining 96 participants (n = 57, 59%, reported currently experiencing autistic burnout; n = 39, 41% reported not experiencing burnout).

Following the ROC curve analysis for the ABM, we conducted additional ROC analyses for the SBM and PHQ-9. Pairwise comparisons were then performed to determine whether the ABM was equally, less, or more effective at predicting whether individuals are currently burnt out compared with the SBM (a general burnout measure) and the PHQ-9 (a depression measure). A significant difference (AUC – AU2 > 0) would support the conclusion that the ABM is better able to discriminate between those who are currently experiencing autistic burnout.

Results

Data screening and missing data

Participants who did not meet inclusion criteria (n = 6), failed reCAPTCHA and attention checks (n = 3) or did not respond to 20% or more of all the survey items (n = 2) were excluded from the analyses. There was <15% of missing data across each of the data sets, and Little’s MCAR test confirmed that missingness was completely at random (Little, 1988). Missing data were imputed by conducting Multiple Imputation in R using the Multivariate Imputation by Chained Equations (MICE) package (van Buuren & Groothuis-Oudshoorn, 2011) and pooling estimates across 20 datasets.

Confirmatory factor analysis

Table 2 provides the fit indices for the CFAs conducted to evaluate the three models outlined in Mantzalas et al. (2024): the unidimensional, correlated four-factor and bifactor models. According to all examined fit indices, none of the examined models produced a good fit to the data. While providing poor model fit, the bi-factor model did show better fit across all indices than the correlated four-factor model, suggesting a general autistic burnout factor improved model fit.

Exploratory factor analysis

As the CFAs did not produce a good fit for the data, an EFA was also conducted to explore other possible structures. Factor extraction methods suggested different numbers of factors: Cattell’s scree test, the VSS, and the Hull method supported the extraction of a single factor (see Figure 2 for scree plot), Velicer’s MAP test suggested three factors, and parallel analysis and the Kaiser-Guttman criterion suggested the extraction of four factors, therefore solutions ranging from one to four factors were considered. Among these, the single-factor and the three-factor solutions provided the most interpretable results (see Table 3).

Figure 2.

Scree for exploratory factor analysis of the autistic burnout measure (N = 379).

Table 3.

Factor loadings for single- and three-factor EFA models of the ABM (N = 379).

			Three-factor solution (oblimin rotation)
Item	Shortened Item Wording	Single-factor solution	F1	F2	F3
1	Trouble thinking clearly	0.80	0.85	0.04	–0.01
2	Harder time making decisions for myself	0.77	0.80	0.01	0.05
3	Harder time solving challenging problems	0.79	0.87	–0.01	0.02
4	Harder time holding information in my mind for short periods of time	0.76	0.78	0.06	0.01
5	Harder time recalling things I know	0.68	0.81	0.05	–0.12
6	Harder time controlling my impulses	0.61	0.43	0.38	–0.15
7	More moody	0.71	0.20	0.68	–0.11
8	Feeling more irritable	0.72	0.05	0.85	–0.12
9	Harder time tolerating sensory input	0.81	0.06	0.68	0.17
10	Harder time preventing sensory overstimulation	0.75	–0.01	0.75	0.18
11	Had more, or more severe, meltdowns	0.74	0.04	0.82	–0.05
12	Had more, or more severe, shutdowns	0.76	–0.01	0.63	0.24
13	Harder time ignoring unimportant sensory input	0.75	–0.03	0.69	0.18
14	Harder time deciding what is and is not important to pay attention to	0.79	0.53	0.13	0.22
15	Harder time getting along with people I know well	0.73	0.08	0.25	0.50
16	Harder time getting along with people at work, school, or in other community settings	0.70	0.08	0.25	0.46
17	Harder time communicating my point to others	0.79	0.33	0.18	0.38
18	Harder time finding the right words to communicate what I mean	0.81	0.32	0.19	0.40
19	Harder time doing basic day-to-day activities	0.76	0.50	0.01	0.34
20	Harder time managing work or school	0.75	0.49	0.0	0.36
21	Harder time managing the steps I need to take to complete tasks	0.85	0.53	0.03	0.41
22	Avoiding social situations	0.72	0.02	0.05	0.77
23	Isolate myself from others	0.74	0.12	0.03	0.71
24	Avoiding stimulating environments	0.74	–0.04	0.19	0.71
25	Avoiding activities that require effort	0.78	0.25	0.06	0.59
26	Felt more mentally exhausted	0.86	0.30	0.30	0.36
27	Felt more physically exhausted	0.78	0.33	0.26	0.29
	Sum of squared loadings	15.55	6.66	5.75	5.18
	Proportion Variance Explained	58%	25%	21%	19%

F1 = Cognitive Dysfunction, F2 = Affective and Sensory Dysregulation and F3 = Social Interaction and Functional Impairment.

Note. Factor loadings > 0.30 are displayed in bold. Cross-loadings > 0.30 are indicated using italics.

Single-factor solution

The single-factor solution accounted for 58% of the total variance, with all 27 items exhibiting robust loadings (range = 0.61–0.86), indicating that they shared a strong association with a general underlying unidimensional construct.

Three-factor solution

Compared with the single-factor solution, the 3-factor solution accounted for only an additional 7% of the total variance in ABM scores (65%), with the factors capturing 25%, 21%, and 19% of the total variance, respectively. Each of the three factors exhibited distinct loadings for at least four items, indicating some level of uniqueness and distinction between the factors.

The first factor (F1), labelled ‘Cognitive Dysfunction’, largely captured items relating to challenges with executive functioning, such as planning and decision-making, and was similar to Mantzalas et al.’s (2024) Cognitive and Functioning Difficulty factor. The second factor (F2) labelled ‘Emotional and Sensory Dysregulation’, encompassed items associated with emotional dysregulation and sensory overload, and was similar to Mantzalas et al.’s (2024) Emotional & Sensory factor. Finally, the third factor (F3), labelled ‘Social Interaction and Functional Impairment’, encompassed items related to daily functioning, social communication, avoidance, and withdrawal, and was similar to a combination of Mantzalas et al.’s (2024) Avoidance and Exhaustion factor and Social and Communication factor. However, there were several cross-loadings, as shown in Table 2, specifically F1 and F3 exhibited significant overlap, wherein items relating to daily functioning (Items 19–21) and communication (17 and 18) had comparable loadings on both factors.

The three factors had high inter-correlations (r = 0.64–0.71), indicating a substantial degree of shared variance (Morin et al., 2016). Overall, inspection of the pattern coefficients suggested that a single-factor solution, with strong factor loadings and accounting for 58% of the variance alone, was the favoured factor solution.

Bi-factor exploratory factor analysis

Morin et al. (2016) noted that high correlations and cross-loadings among factors may indicate hierarchical structures. Therefore, the three-factor model was submitted to a BEFA to explore the ABM’s dimensionality further and determine the validity of interpreting the ABM as essentially unidimensional. The BEFA showed that all items meaningfully loaded onto the general factor (λ = 0.58–0.87; see S1 in Supplementary Materials). All items had higher loadings on the general autistic burnout factor than on the specific factors, indicating that the general factor predominantly explained the variances.

Regarding dimensionality, the ωH was 0.97, and the general factor explained 81% of the common variance (ECV_G). The specific factors, on the other hand, did not appear to provide additional value above the general factor, ωS = 0.00–0.14. These findings suggest that while the specific factors indicate some multidimensionality, the measure primarily reflects a strong general factor of autistic burnout. As such, an aggregate total score on the ABM is interpretable (Reise, Bonifay, et al., 2013).

Reliability

Internal consistency

The 27-item ABM total score had excellent internal reliability, ω = 0.98.

Test–retest reliability

The ABM total score demonstrated moderate stability over time (12 months), with an intraclass correlation coefficient of 0.59 (95% CI = [0.48, 0.70]).

Construct validity

Correlations between the ABM total score and the other measures are presented in Table 4. Higher ABM total scores were significantly correlated with higher depression and anxiety, with medium effect sizes, and with general burnout assessed using the SBM, with a large effect. Furthermore, higher ABM scores were associated with more autistic traits and higher use of camouflaging behaviours, with small to moderate effect sizes.

Table 4.

Internal consistency, descriptive statistics and correlations between the ABM and other theoretically related construct measures (N = 379).

Variables	Cronbach’s alpha (α)	M (SD)	Pearson’s r correlations with ABM	95% confidence interval [upper, lower]
Autistic Burnout (ABM)	0.97	88.83 (25.34)	–
Autistic Traits (BAPQ)	0.89	4.25 (.60)	0.33**	[0.42, 0.23]
Camouflaging (CAT-Q)	0.90	127.18 (22.26)	0.36**	[0.45, 0.27]
Anxiety (GAD-7)	0.89	11.59 (5.86)	0.54**	[0.61, 0.47]
Depression (PHQ-9)	0.88	13.96 (7.20)	0.52**	[0.60, 0.45]
General Burnout (SBM)	0.98	55.78 (28.10)	0.78**	[0.83, 0.71]

Descriptive statistics and correlation analyses were calculated for 379 autistic adults, except for the SBM which was completed by only a subsample of 149 of the participants.

^**p < 0.001.

Interpretability

ROC analyses evaluated the discriminative ability of the ABM, SBM and PHQ-9 in distinguishing individuals who reported currently experiencing autistic burnout from those who did not (see Figure 3). The ABM had the highest AUC with an observed coefficient of 0.92 (95% CI = [0.86, 0.97]), indicating excellent discriminative ability. The SBM also showed strong performance (AUC = 0.87, 95% CI = [0.80, 0.94]), while the PHQ-9 exhibited moderate predictive accuracy (AUC = 0.78, 95% CI = [0.68, 0.88]). The optimal threshold for each measure was determined using Youden’s index, with the ABM yielding the highest sensitivity (0.92) and specificity (0.77) (see Supplemental Table S2).

Figure 3.

Receiver operating curve comparison.

ABM = Autistic Burnout Measure. SBM = Sydney Burnout Measure. PHQ-9 = Patient Health Questionnaire.

To determine whether the ABM exhibited significantly greater discriminative accuracy than the other measures, pairwise comparisons of AUC values were conducted using DeLong’s test (DeLong et al., 1988) (see Supplemental Table S3). The results indicated that the ABM had a significantly greater AUC than the PHQ-9 (p = 0.00, 95% CI = [0.06, 0.22]), suggesting superior predictive ability. The two burnout measures (ABM and SBM) did not differ significantly in AUC (p = 0.13, 95% CI = [−0.01, 0.11]), indicating comparable performance. The difference in AUC between the SBM and PHQ-9 was also statistically significant (p = 0.02, 95% CI = [0.01, 0.16]), suggesting discriminative accuracy. These analyses were also repeated by merging participants who responded ‘I don’t know’ with the ‘no’ respondents, as per Mantzalas et al. (2024). While this added uncertainty and, unsurprisingly, reduced discriminative accuracy, the ABM exhibited significantly greater discriminative accuracy over the SBM. All other findings provided the same pattern of results as when removing ‘I do not know’ responses (see Supplementary Material). Overall, these findings suggest that the burnout measures are likely better than the PHQ-9 at identifying individuals reporting experiencing autistic burnout. In addition, while the ABM and SBM provided similar predictive accuracy, there was some evidence that the autistic-specific burnout measure was better at identifying individuals experiencing autistic burnout.

Discussion

This study extended investigations of the psychometric properties of the ABM. First, we examined the structural validity of the ABM. Building on the work of Mantzalas et al. (2024), CFA evaluated a series of factor models, including their exploratory four-factor model (Cognitive and Functioning Difficulty, Emotional and Sensory Dysregulation, Avoidance and Exhaustion, and Social and Communication Difficulties). Although we could not verify this factor structure through CFA, we observed a similar pattern of results through EFA, in which a three-factor model was found, comprising cognitive dysfunction, similar to the Cognitive and Functioning Difficulty factor in Mantzalas et al.’s (2024), emotional and sensory dysregulation (similar to the Emotional and Sensory Dysregulation factor) and social and functional impairment subscales (somewhat related to the Social and Communications factor from Mantzalas et al., 2024).

Moreover, consistent with Mantzalas et al. (2024), a detailed EFA and bifactor modelling analysis indicated that a single hierarchical model with one ‘Autistic Burnout’ factor is likely preferred. The higher-order factor influenced all 27 items and accounted for 58% of the variance in total scores, suggesting that the ABM is essentially unidimensional. The unidimensional nature of the ABM supports the calculation of a total score and interpretation of the measure’s internal consistency (Ziegler & Hagemann, 2015), which was found to be excellent in this sample (ω = 0.98; ωΗ = 0.97). However, further validation should explore whether the unidimensional structure holds or whether revisions are needed to capture multidimensionality better.

Regarding test–retest stability, although autistic burnout ratings are expected to fluctuate as individuals develop strategies to manage burnout and as stressors and protective factors evolve over time, the relative ranking of individual burnout levels remained moderately stable over the 12 months. This moderate stability suggests that individuals who experience higher levels of burnout tend to report higher levels relative to others consistently, pointing to the potentially chronic nature and experiences of burnout for some autistic people rather than a transient state. This finding, if replicated in other studies, could suggest that longer-term, individualised approaches to supports and recovery may be preferred over short-term alleviation.

When evaluating the construct validity of the ABM, higher levels of autistic traits were associated with elevated scores of autistic burnout. This correlation corroborates qualitative reports of autistic adults describing experiencing a loss of previously acquired functioning (sometimes referred to by autistic adults as ‘autistic regression ⁹ ’) as a consequence of autistic burnout, whereby individuals experience an apparent exacerbation of social, behavioural and sensory autistic traits (Higgins et al., 2021; Raymaker et al., 2020). In addition, the association between autistic traits and burnout raises the possibility that individuals with higher levels of autistic traits may be more vulnerable to burnout, potentially due to the compounding challenges of sensory overwhelm, social stressors, and inadequate accommodations. Alternatively, and in turn, autistic burnout might reduce the ability to camouflage or mask autistic traits, which could lead to an increased outward expression of autistic traits, making them appear more pronounced to both the individual and others.

Indeed, camouflaging has been consistently described as a significant contributing factor to autistic burnout in qualitative studies (Higgins et al., 2021; Mantzalas et al., 2021; Raymaker et al., 2020). In our sample, camouflaging was moderately positively associated with autistic burnout. This observation aligns with Mantzalas et al. (2024), who also reported a moderate positive correlation (r = 0.36) between the ABM and CAT-Q, as well as Pyszkowska (2024), who noted a similar moderate correlation with the CAT-Q (r = 0.30) using the Polish adaptation of the ABM. The moderate associations observed between camouflaging and autistic burnout may stem from variability in how autistic individuals engage in camouflaging behaviours. For instance, individuals may have supportive networks that enable them to unmask and be their authentic selves. In addition, some people employ coping strategies, such as taking masking breaks, which might help prevent them from becoming fatigued. Furthermore, the cross-sectional design of these studies does not account for the cumulative effects of burnout over time, potentially overlooking stronger associations that would emerge in a longitudinal analysis.

In line with qualitative accounts (Higgins et al., 2021; Mantzalas et al., 2022; Raymaker et al., 2020), higher autistic burnout was also significantly associated with higher anxiety, depression and general burnout. The correlation between the ABM and depression (r = 0.52; r = 0.56, after adjusting for attenuation) fell below the 0.70 threshold for measure exchangeability, suggesting the constructs are related but separable (Carlson & Herdman, 2012). Yet, the disambiguation between the two conditions remains unclear. Both Arnold et al. (2023b) and Mantzalas et al. (2024) found strong associations between ABM and depression (r ranging between 0.59 and 0.67), leading to questions about whether autistic burnout could potentially be conceptualised as ‘autistic depression’. Indeed, evidence of a conceptual overlap between depression and burnout is also evident in the non-autistic burnout literature (Bianchi et al., 2013, 2015a, 2015b; Bianchi & Sowden, 2022; Parker & Tavella, 2021). However, non-autistic people assert the two are related, but distinct, experiences: ‘(burnout) doesn’t feel as suffocating as depression, it just is a state of pure exhaustion’ (p. 4) (Tavella & Parker, 2020a); similarly, autistic people have stressed that despite the shared characteristics between the conditions (e.g. fatigue, anhedonia, and social withdrawal), depression and autistic burnout are also differentiated and that failure to distinguish between the two contributes to misidentification, misdiagnosis, and mismanagement of both conditions. To illustrate, a participant in Higgins et al.’s (2021) study stated that ‘symptoms such as sensory sensitivity and the need to isolate to recover [in autistic burnout] is different to typical depression’ (p. 2362). While behavioural activation may be beneficial for treating depression, it could potentially worsen autistic burnout, whereas reduced activity and withdrawal may be helpful in recovery from burnout (Raymaker et al., 2020). Further research is necessary to clarify the unique and overlapping causes, characteristics, and recovery strategies associated with depression and autistic burnout in autistic people.

Furthermore, according to Mantzalas et al.’s (2022) conceptual model of burnout, anxiety and depression were both conceptualised as antecedents to autistic burnout, whereby co-occurring anxiety and depression contribute to the cumulative stress associated with being an autistic person and thus increase individual risk for autistic burnout. In their quantitative cross-sectional investigation, Arnold et al. (2023b) found that depression was the strongest predictor of autistic burnout as measured by the ABM. However, other qualitative reports by autistic people suggest that autistic burnout may precede anxiety and depression (Raymaker et al., 2020). There may also be a reciprocal bidirectional relationship between depression and autistic burnout, with each potentially exacerbating the chronicity or severity of the other. In the current study, 39% of the participants had a co-occurring diagnosis of depression. Qualitative research has already indicated that co-occurring depression and autistic burnout can contribute to a higher risk of suicidality and self-harm (Mantzalas et al., 2022; Raymaker et al., 2020). Therefore, irrespective of whether autistic burnout is considered a separate condition, or a form of ‘autistic depression’, understanding the directionality and causal connections between autistic burnout and mental health, as well as how people manage and overcome autistic burnout alongside other concurrent conditions, should guide the creation of more impactful interventions and treatment approaches.

We also observed a significant large association between autistic burnout and occupational burnout, as measured by the SBM (r = 0.80). While the association between autistic burnout and occupational burnout supports the construct validity of the ABM, it also raises questions as to whether more contemporary general/ occupational burnout measures such as the SBM, which encompass a broader range of experiences (e.g. interpersonal impact, withdrawal, unsettled mood), do not exclusively measure occupational burnout but consider non-work-related stressors (e.g. family stress, caregiving roles or studying), may also be well-suited to capturing burnout experiences in autistic people. Indeed, Mantzalas et al. (2024) also found the emotional exhaustion subscale of the Copenhagen Burnout Inventory to be as valid a measure of autistic burnout as the ABM.

The ABM showed significantly greater accuracy than the PHQ-9 in distinguishing individuals experiencing autistic burnout, aligning with Mantzalas et al. (2024). However, it performed similarly to the SBM, suggesting general burnout measures may still capture aspects of autistic burnout. Both studies found general burnout measures correlated more strongly with depression than the ABM, indicating they may reflect overall emotional distress rather than autistic burnout specifically. The ABM, with its focus on autism-specific factors like sensory sensitivity and masking, may better differentiate autistic burnout from depression and could be more effective than general burnout measures like the CBI and SBM.

Overall, the ABM most effectively discriminated between individuals experiencing burnout and those not. Schoondermark and colleagues (2024) also noted that the Dutch version of the ABM demonstrated reasonable specificity and sensitivity in correctly classifying the autistic women who answered their single question on whether they recognised themselves at that moment as experiencing autistic burnout or not. Although these findings provide preliminary evidence of concurrent criterion validity in two studies, further work is needed to establish the measure’s criterion validity more firmly.

Limitations

Our findings must be interpreted considering the study’s limitations. Recruitment for this study primarily occurred through social media, meaning that those interested in the topic and late-diagnosed autistic adults were more likely to participate (Rødgaard et al., 2022). Indeed, the mean age at which autism was diagnosed in this study was 25 years. Consequently, the experiences of burnout and the measurement properties of the ABM when applied to early diagnosed autistic adults remain unclear. In addition, most of the participants were Caucasian, cisgender, tertiary educated autistic adults. As a result, the experiences of autistic people with multiple marginalised identities (such as those who are ethnically and gender diverse as well as autistic) or those with higher support needs are not represented in this study, which limits its generalisability findings. While our study included both individuals with a formal autism diagnosis and those who self-identified as autistic, our use of the BAPQ, which was not developed as a screening tool for autism, limited our ability to determine whether participants met specific clinical thresholds. Furthermore, there was a 12-month test–retest interval in the current study. Given the wide temporal variability in people’s reported experiences of autistic burnout, which can range from days to years (Arnold et al., 2023a), future research examining the test–retest reliability of the ABM might consider implementing shorter test–retest intervals to capture these potential variations.

At the time of conducting this study, we only had access to the 27-item version of the ABM and unpublished pilot data on the measure. The pre-publication indicated that a 14-item version of the ABM might be preferred based on feedback suggesting that the 27-item measure is too lengthy and some instructions may be too abstract for certain individuals. Future research would benefit from testing the 14-item version of the ABM in a larger and more demographically diverse sample. Finally, we could only evaluate ABM scores based on a single item asking whether participants believed they were experiencing burnout. This criterion may not have been ideal, especially since we did not provide participants with a definition or operationalisation of autistic burnout during the study (unlike Schoondermark et al., 2024, who did offer their participants a definition of autistic burnout before requesting them to assess the degree to which they identified as experiencing burnout).

Conclusion and future directions

The current study contributes to recent efforts to develop and validate a measure of autistic burnout for autistic adults, providing evidence that, along with findings from Mantzalas et al. (2024), suggests that the ABM is likely a psychometrically reliable and valid measure of autistic burnout. However, the ABM requires further validation in larger and more diverse samples, particularly ethnically and gender-diverse autistic adults with a range of support needs, as well as with non-autistic participants. Despite the association between autistic burnout and depression being below the threshold for exchangeability in the current study, findings from Arnold et al. (2023a, 2023b), and Mantzalas et al. (2024) suggested some overlap between the conditions. Therefore, it is yet to be determined whether autistic burnout might be considered an ‘autistic depression’ (Mantzalas et al., 2024) or a precursor to depression. Understanding the extent and nature of the overlap and distinctive features between the two conditions, as well as how current or future measures can evaluate these, is essential for tailoring assessments and supports to address both autism-related issues (e.g. sensory overstimulation and the need for withdrawal and isolation) and general depressive or burnout symptoms (e.g. anhedonia and low mood). Future research should continue to explore the usefulness and utility of existing measures originally developed for the general population (e.g. Copenhagen Burnout Inventory or Sydney Burnout Measure) in capturing experiences of autistic burnout. Furthermore, when the updated 14-item version of the ABM is published, its psychometric properties should be assessed and compared with those of the 27-item version. Overall, the findings from the current study bolster the case for using and interpreting the 27-item ABM as a unitary measure of autistic burnout among autistic adults.