Abstract
Background Racially and ethnically minoritized autistic individuals face intersectional disparities in access to services and research representation. Prior work tends to overlook individual preferences, lived expertise, and the of language skills in services. To address this gap, this community-based participatory study examines the presence and correlates of services in minoritized autistic adolescents and adults varying in language skills.
Method Participants (N = 73, ages 13-30, >50% with language scores ≤ −1.25 SD and ∼30% with NVIQ < 70 to 84) completed a behavioral assessment protocol. Participants and caregivers completed questionnaires. Descriptive and regression analyses evaluated patterns and predictors of number of services received, unmet needs, and barriers.
Results Participants received multiple services yet had multiple unmet service needs and barriers. Effects of individual differences and social-ecological variables on services received, unmet needs, and barriers varied in presence and magnitude when using a categorical versus continuous approach to individual differences.
Implications While the number of services received and unmet needs were similar to prior work, examination of individual services, unmet needs, and barriers suggested differences. Effects point to nuance in experiences during the transition to adulthood and motivate use of participatory approaches to build the evidence base informing practice and advocacy.
Learning outcomes After reading the article, the learner will be able to: 1) summarize knowledge gaps about access to services; 2) explain why evidence-based practice motivates use of a community-based participatory approach; and 3) describe implications of study findings for clinicians and autistic individuals.
In the United States, autism has a higher diagnosed prevalence in Black, Hispanic or Latine, and Asian or Pacific Islander youth than white youth (1 in 29 to 34 versus 1 in 41) (Maenner et al., 2023). While racial and ethnic disparities in diagnosis have been reduced (Durkin et al., 2017; Mandell et al., 2002, 2007; Shaw et al., 2022), disparities in services persist (Eilenberg et al., 2019; Liu et al., 2023; Rast et al., 2023; Smith et al., 2020). This inequity has cascading impacts. Autistic adults lose access to child- and education-based services in the transition to adulthood (“Every Student Succeeds Act [ESSA],” 2015; “Individuals with Disabilities Education Improvement Act [IDEIA] of 2004,” 2018), including speech-language services (Laxman et al., 2019), and report greater unmet service needs (Anderson et al., 2018; Turcotte et al., 2016). This loss not only puts adults at risk of poor outcomes (Howlin & Magiati, 2017), but also is higher for racially and ethnically minoritized adults (Schott et al., 2022) and adults with language impairment (Johnson et al., 2010). Given the prevalence of autism and of co-occurring structural language impairment (>50%) (Boucher, 2012), which includes challenges with phonology and/or morphosyntax (Schaeffer et al., 2023), reducing disparities is critical and motivates understanding access to services inclusive of race and language skills.
Challenges to understanding services access include systematic exclusion of minoritized autistic individuals from research (Maye et al., 2021) and of dimensional approaches to characterize individual differences, particularly when considering language performance across linguistic domains (Kover & Abbeduto, 2023). Both inclusion and nuanced research methods are needed for broadly replicable findings (National Institutes of Health, 2021) and to enhance access to services (National Institute on Deafness and Other Communication Disorders, 2022). Ways to reduce this bias are: 1) evidence-based practice, which integrates individual preferences, expert opinion, and the best-available evidence (Sackett et al., 1996), and 2) community-based participatory research, which provides a framework for partnering with individuals and experts (Wallerstein & Duran, 2006). This study uses these methods to examine services access, unmet needs, and barriers in minoritized autistic adolescents and adults.
Conceptualizing Expert Opinion and Partnership
There are two considerations in approaching services in minoritized autistic individuals. First, to identify who is an expert, services and diversity must be contextualized. Second, to integrate expert opinion, partnership must be responsive to the intersection of race and disability.
Contextualizing Services and Diversity
Multiple levels of environment shape individual experiences (Bronfenbrenner, 1977). These impacts are especially relevant for minoritized autistic individuals. For instance, U.S. law codified racial and disability-based segregation (Powell, 2012; Turnbull & Turnbull, 1998), leading to disparities regardless of individual differences (Crenshaw, 1989). Structural changes to end segregation, such as Brown v. Board of Education (1954), resulted from changes in social perceptions about disability and race; it was not that minoritized or autistic individuals changed (Powell, 2012). Thus, racial disparities are often due to social processes – and not, as prior work has suggested (Burke & Heller, 2017), due to race or ethnicity themselves. Dis/Ability Studies and Critical Race Theory (DisCrit) makes this relationship clear by framing race and dis/ability as mutually reinforcing social constructs that lead to nuanced social marginalization (Annamma et al., 2013). Further, Diversity Science underlines the importance of examining heterogeneity within minoritized individuals, as there is no one-to-one ratio between race, disability, or experiences (Plaut, 2010). In all, environment and heterogeneity are each important for understanding services and point to who is an expert: individuals with lived, personal, and professional experiences in autism, racial and ethnic minoritization, and services.
Nuances in Partnership
A next step to build the evidence base is to integrate expert opinion (Burns et al., 2011), which can be achieved through community-based participatory research (Wallerstein & Duran, 2010). Yet even with attention to participatory methods in autism research (Fletcher-Watson et al., 2019), high-level dialogue makes clear intersectional nuances are not a focus. Research trends to encourage partnership include advisory boards (Williams et al., 2023) and community involvement statements in journal articles (Autism, 2024), each of which involve disclosure. Some autistic adults prefer to keep diagnosis private and fear discrimination (Romualdez et al., 2021; Thompson-Hodgetts et al., 2020), a grave concern for minoritized individuals (Annamma, 2018). For example, if a community member co-authors a manuscript, they may feel pressured to disclose their diagnosis by a journal seeking to be transparent about the inclusion of autistic individuals in research (Oswald, 2024). This is problematic, because there is no guarantee of confidentiality or nondiscrimination, such as when an editorial board member who reviews a manuscript violates journal policies and professional ethics by sharing confidential information about minoritized autistic authors with a healthcare provider network. Without action, these practices work against partnership and amount to diversity dishonesty (Wilton et al., 2020). To support power-sharing (Wallerstein et al., 2019), research must be responsive to nuances in partnership versus checking a box (Plaut, 2010).
Summary
Appreciating services in minoritized autistic adolescents and adults involves social processes underlying individual experiences and guides identification of experts to partner with in research. In turn, partnering with experts means attending to nuance versus broad guidelines.
Evidence on Services for Minoritized Autistic Adolescents and Adults
Presence and Nature of Services
The third part of building the evidence base involves determining the best-available evidence (Sackett et al., 1996). Studies on services in autistic adolescents and adults primarily use caregiver report to assess services received, unmet needs, and barriers; see Supplementary Tables 1 to 3. Samples were primarily non-Hispanic white (77% to 93.6%) and male (71.8% to 85.8%), with varying rates of intellectual disability (17% to 58.8%). When reported, the number of services received differed: 0 of 4 in 39.1% of post-secondary adults (ages 19 to 23) (Shattuck et al., 2011), 3.05 of 16 in high school seniors (ages 17 to 22) (Taylor & Henninger, 2015), and 6.13 of 15 in adolescents and adults (ages 16 to 30) (Ishler et al., 2022). Commonly received services were mental health (35% to 62.6%) and disability-related medical services (18% to 65.5%). Unmet service needs varied, ranging from 1.85 of 16 (Taylor & Henninger, 2015) to 3.18 of 15 (Ishler et al., 2023), and pertained to entering adulthood: occupation or life skills therapy (14% to 33.3%) and vocational skills or training (33.3% to 35.9%). Individual barriers centered on access: location (15.5% to 33.3%), availability (15.5% to 49.3%), and cost (15.3 to 38.5%). Overall, there was diversity in both measurement and services outcomes.
Factors in Services Received, Unmet Needs, and Barriers
To explain services, studies examined sociodemographics, individual differences, and social-ecological factors. While important, approaches did not center heterogeneity.
Sociodemographics
Effects of race, ethnicity, sex, and gender on services were inconsistent. Findings included: no effects of race or ethnicity on services received (Ishler et al., 2022) or barriers (Ishler et al., 2023); higher likelihood of receiving no services in Black youth (OR = 3.33) and “other or mixed” youth (OR = 1.30) than white youth (Shattuck et al., 2011); and three more unmet service needs for “other” youth or Black youth versus white non-Hispanic youth (Taylor & Henninger, 2015). These patterns are based on comparisons of white versus minoritized (Ishler et al., 2022, 2023; Laxman et al., 2019; Song et al., 2022; Taylor & Henninger, 2015; Turcotte et al., 2016), with varying definitions of “minoritized.” One report defined “underrepresented minority race/ethnicity” as Black, Hispanic, or multiracial (Ishler et al., 2022), while another report on the same sample additionally included Native American or Alaska Native under “historically marginalized minority race/ethnicity” (Ishler et al., 2023). Effects of sex at birth or gender were also limited by incomplete reporting, such as reporting only males or not specifying use of gender or sex (Shattuck et al., 2011; Taylor & Henninger, 2015). Beyond running counter to federal guidelines for health disparity populations (Office of Management and Budget [OMB], 2024) and current guidelines (American Psychological Association [APA], 2019; Flanagin et al., 2021), these practices hinder the interpretation of sociodemographic patterns and do not provide a sufficient basis for understanding heterogeneity.
Individual Differences
There was no one effect of adaptive behavior, autism traits, or co-occurring diagnoses on services. In some cases, adaptive behavior was tied to services received: (a) scores < −2 SD predicted 1.5 more services received (Taylor & Henninger, 2015); (b) each one-unit SD higher on an adaptive behavior measure predicted 0.77 fewer services received (Ishler et al., 2022); and (c) higher functional mental skills predicted lower likelihood of receiving case management (OR = 0.22 to 0.25) or medical services (OR = 0.43 to 0.66) (Shattuck et al., 2011). Elsewhere, adaptive behavior did not predict unmet service needs (Taylor & Henninger, 2015) or barriers (Ishler et al., 2023). Of note is that for minoritized autistic individuals, racism and discrimination shape individual decision-making processes (e.g., going out with friends as a Black man – let alone an autistic man with language impairment – may not be safe regardless of individual ability) (Girolamo et al., under review). Therefore, adaptive behavior may not always be a meaningful correlate. Effects of co-occurring diagnoses were also mixed. On one hand, psychiatric disorders vary widely. Yet even when considering just intellectual disability, there was: increased likelihood (OR = 1.75) (Laxman et al., 2019) or no effect on services received (Ishler et al., 2022; Taylor & Henninger, 2015), as well as no effect on barriers (Ishler et al., 2023) or unmet needs (Laxman et al., 2019). One study found no effect of language on services received but defined “severe language impairment” as caregivers reporting adult “has a lot of trouble speaking clearly” or “doesn’t speak at all” and equated “severe language impairment” with “nonverbal” (Shattuck et al., 2011). Examination of autism traits was limited, finding no effects on services received or unmet needs (Taylor & Henninger, 2015). In all, these mixed patterns point to the importance of heterogeneity.
Social-Ecological Factors
Educational enrollment and socioeconomic status (SES) on services have varied findings. High school enrollment predicts services received (sr2= 0.12) (Ishler et al., 2022). Yet, services received decrease and unmet service needs increase in adolescence (Turcotte et al., 2016), such that exiting high school alone may not impact services received (Laxman et al., 2019) and barriers may be constant (Ishler et al., 2023). At the same time, high school exit predicts more unmet service needs (OR = 1.37) (Laxman et al., 2019), including speech-language and other services often provided by public education (Song et al., 2022). Thus, even as services decline into adolescence, educational enrollment may be important. Effects of SES, as assessed by income, are inconsistent. Lower income has been tied to more unmet service needs (Laxman et al., 2019) and fewer services received (Shattuck et al., 2011), but elsewhere, has had no effects on services received, barriers (Taylor & Henninger, 2015), or extent of barriers (Ishler et al., 2023). One note is income does not account for individual needs. In contrast, family financial burden predicted access barriers to services (sr2 = .04) (Ishler et al., 2023). Such barriers limit opportunities for community participation and point to the relevance of sense of community as a social-ecological factors (Cameron et al., 2022). Contextualizing the environment is important to understanding services.
Summary
Studies to date on services received, unmet needs, and barriers document broad patterns yet are insufficient for understanding nuance. Given relevant theory (Annamma et al., 2013; Plaut, 2010), incorporating individual differences in language, social-ecological factors, and expert opinion are needed to provide best-available evidence on services in minoritized autistic adolescents and adults.
The Current Study
This study took a community-based participatory approach to examine services in minoritized autistic adolescents and adults varying in language skills. We asked:
What are patterns in services received, unmet service needs, and barriers to having service need met?
To what extent are services received, unmet service needs, and barriers to services, predicted by individual differences (language skills, autism traits, NVIQ) and social-ecological variables (educational enrollment, sense of community), and how do effects compare when using a categorical versus continuous approach to individual differences?
Method
This study received institutional board approval and followed all ethical guidelines.
Procedures
Participatory Approach
Individuals who were not participants came together to discuss intersectional disparities in research and practice based on their lived, personal, and professional experiences. The initial aim of these discussions was to address community needs (e.g., conducting language assessment to provide access to services). These discussions led to community members suggesting a research project on services would support advocacy for their communities and expressing an interest in becoming partners. By nature of discussions, the team agreed upon use of DisCrit (Annamma et al., 2013) and Diversity Science (Plaut, 2010).
To support power sharing (Wallerstein et al., 2019), the team co-developed norms. One was autonomy in how members represented themselves, versus prescribing identities to them (Plaut, 2010; Romualdez et al., 2021). For instance, in a separate situation, a white investigator emailed a team member and invited them to be a community advisory board member on the basis of wanting autistic people “with one other identity,” which was tokenizing (Maye et al., 2021). For other members, intersectional discrimination led to coming into contact with the criminal justice system and professional harm, impacting how they publicly shared information about their diagnoses (Annamma, 2018). Thus, the team did not use public statements and honored individual preferences for public team affiliation.
A second norm was mutual benefit to partners beyond this project, as several partners recounted multiple unethical situations where senior researchers were eager to partner with community members for the researchers’ sole gain. For example, two members shared incidents when senior researchers approached each of them to participate on generously-funded grants, but the following occurred: 1) the senior researcher demanded a letter of support from the community member, which entailed the member completing a substantial amount of work with no offer of compensation, and 2) the senior researcher paid advisory board consultants $200 for five years of work, while failing to notify them they paid faculty consultants over $200 per hour. Thus, the first author met with community partners to develop a plan for supporting them in their goals (e.g., identifying project skills and tasks to support career pathways), leverage resources to support partners in their work (e.g., organizational and writing skills), and write proposals with transparency in budgeting and compensation of partners at or above the same rate as researchers.
The third norm was accessibility. Based on individual profiles, activities took place in multiple modalities and leveraged strengths. For example, if a team member indicated written text was inaccessible, while verbal conceptual insight was a strength, they met with another member to review information and provide insight verbally. Accessibility also involved flexible scheduling and transparency on the research process, including what to expect and team roles.
Selection Criteria
The partners above were not participants in this study. Selection criteria of participants were: (a) racially minoritized and/or ethnically minoritized per U.S. Census guidelines; (b) formal clinical diagnosis of autism, per requirements for inclusion in recruitment sources, and independent confirmation using the Social Responsiveness Scale-2nd Ed. (SRS-2) (Constantino, 2012) and expert clinical judgment; (c) ages 13 to 30, coinciding with approximately when transition planning begins in many states within the U.S. and 10 years post-federal eligibility for special education services (ESSA, 2015); (d) proficiency in English per self-report during screening, as assessments were in English; (e) adequate hearing and vision thresholds for responding to audiovisual stimuli on a computer screen, and; (f) use of primarily spoken language to communicate, as study activities required listening to and producing language responses. Selection criteria for caregivers were: (a) active caregiving role for autistic participants; (b) proficiency in English, as interviews were completed in English; and (c) adequate hearing and vision thresholds for completing study activities.
Recruitment
Recruitment took place in a multi-step process: (a) sharing study flyers with community and advocacy organizations serving minoritized autistic adolescents and adults by email, (b) providing consultation to individuals and families about the study in their modality of choice (phone, Zoom, email), (c) obtaining informed consent using a dynamic process, and (d) collecting data. Recruitment and data collection took place from 2022 to 2023 remotely on HIPAA-compliant Zoom, using a stopping rule of N = 68. The first author administered an assessment protocol, including questionnaires, to participants and caregivers.
Measures
Sociodemographics
To characterize the sample, this study collected participant race, ethnicity, sex assigned at birth, and gender; see Table 1. Respondents selected one or more options for race per U.S. Census categories (OMB, 2024), and could write in options. Race, ethnicity, and gender were excluded from analysis, as they are complex constructs beyond the scope of this study (APA, 2019). Sex was also excluded, as studies tend to aggregate sex assigned at birth and gender (Halladay et al., 2015).
Individual Difference Measures
Language
Participants completed normed assessments across language domains (Girolamo et al., 2023b; Magiati et al., 2014), which align with clinical practice patterns for service eligibility (Burke et al., 2023; Selin et al., 2022); see Table 2. Overall receptive-expressive language was assessed by the Clinical Evaluation Language Fundamentals-5th Ed., normed up to age 21 (Wiig et al., 2013). Age 21 norms were used for participants over age 21, per prior work on adults ages 18 to 49 (Botting, 2020; Clegg et al., 2021; Fidler et al., 2011). Receptive and expressive vocabulary were assessed by the Peabody Picture Vocabulary Test-5th Ed. (Dunn, 2019) and Expressive Vocabulary Test-3rd Ed. (Williams, 2019). Nonword repetition was assessed by percent accuracy on the Syllable Repetition Task (Shriberg et al., 2009).
NVIQ
Nonverbal ability was assessed using the Raven’s Progressive Matrices-2nd Ed. (Raven et al., 2018). The Raven’s does not use language, which yields nearly 1 SD higher scores in autistic individuals with language impairment than measures of verbal IQ (Grondhuis et al., 2018). For each item, participants select one of five images to complete a visual stimulus.
Autism Traits
Autism traits were measured using the SRS-2, which has caregiver and self-report forms for students and adults (Constantino, 2012). Respondents indicate the frequency of 65 items on a four-point scale. Item-level scores provide restricted and repetitive behaviors, social communication impairment, and overall t-scores.
Social-Ecological Factors
Sense of Community
The Brief Sense of Community Scale is a validated scale for diverse youth and adults (Cardenas et al., 2021; Lardier Jr et al., 2022; Lardier Jr et al., 2018; Opara et al., 2021; Peterson et al., 2008). Respondents rate eight statements from strongly disagree to strongly agree on a five-point scale. Higher scores indicate higher sense of community. Item-level scores are averaged to provide an overall score.
Educational Enrollment Status
Respondents reported if participants were enrolled in educational programming, and if applicable, type of programming (e.g., junior high, high school, specialized high school program, college courses, post-secondary program, other).
Services Received, Unmet Service Needs, and Barriers to Services
Per Taylor and Henninger (2015), services were measured using items from the NLTS-2 (Newman et al., 2011) plus two additional items: insurance not accepted and physical accessibility. Respondents selected if they received each of 16 services; if services not received were unmet needs; and if each of 12 items were barriers to having service needs met; see Table 3 for services and unmet needs and Table 4 for barriers. Services received and unmet needs included “other,” with a write-in option. No response under “other” had more than one service.
Data Processing
Two research assistants independently scored and checked measures. All disagreements were discussed until consensus was reached. NVIQ was auto-scored within its testing platform. Data were checked for missingness and replaced using predictive mean matching with one imputation in SPSS 29 (IBM Corp., 2023; Little & Rubin, 2019): language and NVIQ scores (n = 1; did not complete assessment); SRS-2 scores (n = 2; n = 1 missing form, n = 1 did not complete the SRS-2); and educational programming (n = 1; did not answer question). For the categorical approach, data were coded as follows: a) language impairment as ≤ −1.25 SD on ≥2 measures: CELF-5 Receptive Language Index, CELF-5 Expressive Language Index, SRT percent accuracy, PPVT-5 standard score, EVT-3 standard score (Girolamo & Rice, 2022; Tomblin et al., 1997); b) intellectual disability as NVIQ < 70 (American Psychiatric Association, 2013); and high levels of autism traits as SRS-2 total t-scores > 76 (Constantino, 2012). For the continuous approach, language scores were checked for multicollinearity using variance inflation factors (VIF). As VIF ≥ 5, CELF-5 core language scores, which had the highest correlations with other language measures, were used. To facilitate interpretation, CELF-5 core language scores and NVIQ were centered on M = 100, and SRS-2 total t-scores were centered on the sub-clinical threshold of 59. Educational enrollment was coded as enrolled or not, as secondary and post-secondary distributions were similar and as Mann-Whitney U tests revealed nonsignificant differences in median counts for services received, U = 158, z = −.12, p = .916; unmet needs, U = 178, z = .49, p = .642; and barriers, U = 123.5, z = −1.18, p = .245.
Analysis
To address the first research question, descriptive analysis was used to obtain primary outcome measures (M, SD, range) for total number of services received, unmet service needs, and barriers to having service needs met, as well as percentage for each individual service, unmet service need, and barrier. For the second question, primary outcome measures were rate ratios and secondary outcomes were effect sizes. Analyses used an a priori significance level of .05. Given sample size, correlations identified significant effects to use in analysis. As data were non-normally distributed, Spearman (1904) correlations were used. Next, negative binomial regressions with a log link function and robust variance estimator separately estimated the extent to which number of services received, unmet service needs, and barriers to services were predicted from categorical individual differences (language impairment, intellectual disability, high level of autism traits) and social-ecological variables (sense of community, educational enrollment; Coxe et al., 2009). Regression analyses were repeated with continuous individual differences (centered CELF-5 core language scores, SRS-2 total t-scores, NVIQ). Model results were checked for linearity, normality, homogeneity of variance, and equal variance.
Results
Participant Flow
Participants were 73 autistic individuals; see Table 1. Caregivers (N = 52) included primarily mothers (n = 47), as well as grandparents (n = 2), fathers (n = 2), and a sibling (n = 1). The female-to-male ratio for sex assigned at birth and gender was approximately 1:2. Over half were enrolled in educational programming, 28 of whom were in secondary education. Most participants (54.79%) had language impairment; see Table 2. Six (8.22%) had NVIQ < 70, and 15 (20.55%) had NVIQ of 70 to 84. Levels of autism traits varied: high (n = 33, or 45.21%), moderate (n = 19, or 26.03%), low (n = 10, or 13.70%), and subclinical (n = 11, or 15.1%).
Patterns in Services Received, Unmet Service Needs, and Barriers to Services
On average, participants received more than three services (M = 3.67, SD = 2.76), with a range from 0 to 12; see Figure 1. Nearly all received at least one service (93.15%), over three-quarters received at least two services (76.71%), and more than one-half received at least three services (58.90%). Only psychological or mental health services (65.75%) and disability-related medical services (54.79%) were received at above-chance rates (i.e., 50%), while approximately one-third received speech-language or communication services (34.25%); see Table 3.
Participants reported over three unmet service needs (M = 3.32, SD = 3.20), ranging from 0 to 14. Over three-quarters reported at least one unmet service need (78.08%), and over two-thirds reported at least two service needs (67.12%). Of those not receiving a given service, the most common unmet service needs pertained to entering adulthood: career counseling or job skills training (64.71%) and occupational or life skills therapy or training (50%). The next highest unmet service need pertained to more global access, speech-language services and communication services (45.83%). All other services were identified at well below-chance rates. Overall, even when receiving multiple services, participants had multiple unmet service needs.
In parallel to having multiple unmet service needs, participants reported multiple barriers to having service needs met (M = 5.52, SD = 3.31, range = 0-11); see Table 4. Most of the 12 possible barriers were selected at above-chance rates and pertained to access versus quality: location (68.49%), services not available (67.12%), insurance not accepted (58.90%), ineligible (56.16%), cost (54.79%), getting information about services (54.79%), and scheduling conflicts (50.68%). Nearly all participants reported at least one barrier (87.67%), over three-quarters reported at least three barriers (78.08%), and over half reported at least six barriers (52.05%). In all, participants received multiple services, unmet needs, and barriers.
Predictors of Services Received, Unmet Needs, and Barriers
Categorical Approach to Individual Differences
Spearman correlations revealed no significant relationships between services received, unmet needs, and barriers; see Table 5. Thus, receiving more services was not necessarily tied to fewer unmet needs or barriers. There were small to moderate effects of individual difference and social-ecological variables on each services outcome. Receiving more services was associated with having language impairment, having intellectual disability, and being enrolled in education. Descriptively, effect sizes of language impairment and educational enrollment were nearly twice that of intellectual disability. High levels of autism traits and lower sense of community were associated with more unmet service needs, and to a lesser extent, more barriers. There were few patterns among predictors. Having language impairment was associated with having intellectual disability, and lower sense of community was associated with high levels of autism traits.
Separate negative binomial regression models significantly predicted the number of services received and unmet service needs but not barriers; see Table 6. In the first model, for an individual without intellectual disability, without language impairment, and not in educational programming, the number of services received was 1.97. Holding all other variables constant, the number of services received was 1.70 times greater if an individual had language impairment than not (4.42 versus 2.60 services) and 1.56 times greater if in educational programming versus not (4.24 versus 2.71 services). The effect of intellectual disability was not significant. Next, analysis examined unmet service needs. The number of unmet service needs was 5.52 for an individual without high levels of autism traits and a sense of community score of 3.06, which roughly corresponds with neither agree nor disagree. Accounting for all variables, the number of unmet needs in individuals with high levels of autism traits was 1.82 times greater than those without (i.e., lower than) high levels of autism traits (4.29 versus 2.36 needs). In sum, receiving more services was tied to language impairment and educational enrollment, while unmet service needs only involved high levels of autism traits.
Continuous Approach to Individual Differences
When using continuous measures of overall language, NVIQ, and autism traits, effects somewhat differed from the categorical approach; see Table 7. Services variables did not necessarily correspond with one another, but there was a significant, moderate positive association between barriers to services and unmet service needs. As in the categorical approach, there were small to moderate effects of language, NVIQ, and educational enrollment on services received. Lower language scores, lower NVIQ, and being in educational programming significantly associated with receiving more services, but the effect size of NVIQ was 1.75 times larger than that of intellectual disability (ρ = −.42 versus ρ = .24). Effects of SRS-total t-scores and sense of community on number of unmet service needs and barriers were similar to the categorical approach. Higher SRS-2 total t-scores associated with more unmet needs and barriers, and sense of community showed the opposite pattern. Other differences pertained to predictor variables. Lower sense of community associated with higher SRS-2 total t-scores, and higher CELF-5 core language scores associated with NVIQ. However, the effect size of NVIQ with CELF-5 scores was over two times larger than that of intellectual disability (ρ = .59 versus ρ = .27). Last, higher CELF-5 core language scores were associated with not being in educational programming. In all, patterns varied when using a continuous approach to individual differences: 1) significant associations between unmet needs and barriers, as well as CELF-5 core language scores and educational enrollment, and 2) stronger associations between NVIQ than intellectual disability and CELF-5 core language scores, as well as educational enrollment status.
As in the categorical approach, separate negative binomial regression models were statistically significant for number of services received and unmet service needs but not barriers; see Table 8. However, effects differed. Number of services received was predicted by educational programming, but not CELF-5 core language scores or NVIQ. For a participant with NVIQ of 100, CELF-5 core language score of 100, and not in educational programming, the average expected number of services received was 2.11. Holding all other variables constant, the number of services received was 1.48 times greater if in educational programming versus not (4.07 versus 2.75 services); in comparison, the rate ratio was 1.56 in the categorical approach. Turning to unmet service needs, for an individual with a sense of community score of 3.06 (roughly indicating neither agree nor disagree) and an SRS total t-score of 59 (corresponding to the sub-clinical threshold for autism traits), the expected number of unmet service needs was 4.50. When accounting for all other variables, the number of unmet needs was 1.03 times greater for every one-unit increase in SRS-2 total t-scores above 59; in contrast, the rate ratio of high level of autism traits was 1.82. In all, receiving more services was tied only to educational enrollment, while unmet service needs appeared to be tied to SRS-2 total t-scores.
Summary
Participants received over three services yet reported over three unmet service needs and five barriers. Effects showed nuanced differences when characterizing individual differences as categorical or continuous, and some correlational effects became nonsignificant in regression. In particular, language impairment, but not language scores, predicted services received.
Discussion
In evaluating services access in minoritized autistic adolescents and adults varying in language skills, this study implemented a community-based participatory approach tailored to the intersection of race and disability. Team members agreed upon norms and a framework for diversity and services (Annamma et al., 2013; Plaut, 2010). This approach informs interpretation.
Clinical Implications
This study was based on the premise that representation in research is crucial for evidence-based practice (Office of Management and Budget, 2024; Sackett et al., 1996). Here, the number of services received and unmet needs (n = 3.34 and n = 3.32, respectively) were somewhat similar to high school seniors (n = 3.05 and n = 1.85) (Taylor & Henninger, 2015) but lower than adolescents and adults (n = 6.13 and n = 3.18) (Ishler et al., 2023; Ishler et al., 2022). Differences could be due to sampling, such as educational enrollment, co-occurring intellectual disability (<10% in this study versus 27%), or other diagnoses (language skills in this study versus psychiatric disorders in Ishler et al. (2022). It may also be that this sample had different experiences in nuanced ways. Participants were approximately two-thirds to over two times less likely to receive transportation or social work services (Ishler et al., 2023); three to eight times more likely to identify speech-language, transportation, and personal assistant or aide services as unmet service needs (Taylor & Henninger, 2015); and more likely to endorse each individual barrier to services except for language (Koffer Miller et al., 2022; Newman et al., 2011; Taylor & Henninger, 2015). These nuances motivate attention to the literature versus making assumptions.
Understanding clinically significant differences in services is important for practice. Effects on services differed when using categorical labels versus continuous scores to characterize individual differences. While primary disability labels often use cutoffs (ESSA, 2015; IDEIA of 2004, 2018), continuous measures yielded an association between overall language scores and educational enrollment that was not significant in the categorical approach and a stronger association between language scores and NVIQ. These patterns raise the question of clinical significance. For instance, Johnson et al. (1999) found a research definition of language impairment in autistic and nonautistic adults yielded a higher estimate than clinical judgment (11.7% versus 5.5%). Here, >55% of participants qualified for language impairment, which predicted the number of services received. The proportion receiving speech-language services (>33.3%) or having unmet needs (30.14%) surpassed those with language impairment. It is unknown to what extent the cutoff used here – or domains of language assessed – align with clinical judgment or if services were tied to other areas of language, such as pragmatics (Schaeffer et al., 2023). For clinicians, this motivates assessing all areas where autistic individuals may experience differences and want support (Musgrove, 2015).
Implications for Autistic Individuals
Findings also have implications for autistic individuals themselves. Models for barriers to having services needs met were not significant. However, associations between barriers and autism traits and sense of community point to the relevance of understanding disability within minoritized autistic individuals. Higher autism traits and lower sense of community were each associated with more barriers, as well as with one another. Per DisCrit (Annamma et al., 2013), these patterns might be explained by a model in which barriers to services, sense of community, and autism traits reinforce one another. Minoritized autistic individuals and their families face systemic barriers to accessing services (Weitlauf et al., 2024), including inadequate cultural responsivity in service delivery that can exacerbate lack of trust and rapport with professionals (Pham & Charles, 2023). Consequently, these interactions might have a dual effect of both failing to provide access to services to support individuals in areas where they may want support and negatively impacting sense of community. Amid a broader knowledge gap on sociocultural variation in autism traits (Golson et al., 2022), understanding how to reduce barriers through representative research and integration of individual preferences is key.
This study also leveraged a participatory approach that integrated components of the evidence base with partnership. To foster power-sharing (Wallerstein et al., 2019), this study took the same approach to diversity and representation as in recruitment of participants: attending to heterogeneity in social processes relevant to this study versus checking a box (Plaut, 2010). This approach provided a pathway to partnering with community members who are systematically excluded from the evidence base as both researchers and participants (Burns et al., 2011). The team considers these partnerships as the natural next step to advancing evidence-based practice (Sackett et al., 1996), inclusive of clinical services and research.
Limitations
Though this study provided new information on services in minoritized autistic adolescents and adults varying in language skills, it had several limitations. First, this study did not examine socioeconomic status, which is associated with disparities in the transition to adulthood in autistic adolescents and adults (Eilenberg et al., 2019). However, socioeconomic status is a complex construct (Girolamo et al., 2023a). For instance, some measures use marital status, retirement or employment status, educational attainment, and occupational prestige (Hollingshead, 1975), but that does not account for cost of living or individual need. In turn, family financial burden may not apply to an adult who does not have living family members (Ishler et al., 2023). A second limitation was not reporting detailed information on participants, including exact multiracial categories, educational programming and “other” services received, and frequency, extent of need, and duration of services (Burke et al., 2023). Amid concerns about potentially making participants re-identifiable in this small sample study (O’Keefe & Rubin, 2015), this report opted to provide less information. A third limitation is a need for better services measures that attend to nuances in services (Burke et al., 2023), given qualitative patterns that provide rich information (Anderson et al., 2018), balancing precision with privacy and feasibility for routine care settings (Douglas et al., 2022).
Future Directions
Findings and limitations chart pathways forward to better understand services in the transition to adulthood in minoritized autistic adolescents and adults. First, heterogeneity within a 100% minoritized sample in the presence and nature of services justified use of DisCrit and Diversity Science (Annamma et al., 2013; Plaut, 2010). What is unknown, however, is how these theories inform larger-scale modeling of the transition to adulthood. Prior work has examined broad patterns in large datasets (Song et al., 2022; Turcotte et al., 2016) and longitudinal change in individual studies (Laxman et al., 2019), but future work is needed to merge theory-founded approaches to social processes and heterogeneity with statistical methods. A second area for future research is improved measures for not just services (Burke et al., 2023), but also for social drivers of health (Centers for Disease Control and Prevention, 2021). Here, sense of community was associated with services outcomes, but its clinical significance is unknown. This charge might be addressed through implementation science approaches (Waltz et al., 2019), using a purposeful process to engage multi-level partners and move knowledge into action (Field et al., 2014). Though not comprehensive, these directions for research serve to advance knowledge and advocacy for access to services for minoritized autistic adolescents and adults.
Conclusion
In studying services access, unmet needs, and barriers in minoritized autistic adolescents and adults, this community-based participatory study began with lived experiences as expert opinion to guide investigation. Using theory-driven approaches to investigation, findings showed heterogeneity in the presence and nature of services. Correlates of services differed when using continuous or categorical approaches to individual differences, including language, further supporting the relevance of heterogeneity in characterizing the transition to adulthood. This approach and findings may support iterative development of evidence-based practice.
Data Availability
All data produced in the present study are unavailable due to participants opting not to share their de-identified data in a dynamic consent process.
Acknowledgements
We thank participants, families and community partners. This work was supported by the American Speech-Language-Hearing Foundation (PI: Girolamo).
Footnotes
Authors: Alicia Escobedo, aescobedo4259{at}sdsu.edu Kyle Greene-Pendelton, kyle{at}sankofaspeech.com Samantha Ghali: samantha.ghali{at}ku.edu, Iván Campos, icamposslp{at}gmail.com Poornima Ram-Kiran, prk14{at}uw.edu
Disclosures: The authors have no conflicts of interest.
Acknowledgements: TG was supported by an American Speech-Language-Hearing Foundation New Investigators Research Grant (PI: Girolamo).
References
