With a diagnostic prevalence of one in 36 children in the United States (Shaw et al., 2023), autism is thought to occur at similar rates across racial, ethnic, and demographic groups; however, access to services is not equitably distributed across these groups, with well-documented health disparities in receipt of autism-related diagnostic and support services throughout the years (e.g., Benevides et al., 2017; Smith et al., 2020; Suhrheinrich et al., 2021). In this study, we aim to identify how the intersection of racial, ethnic, and other demographic factors relates to receipt of early intervention services among young autistic children enrolled in Part C Early Intervention. To address well-documented health disparities in autism-related services, a recent multi-stage screening program in Part C Early Intervention (EI) settings was found to improve access to diagnostic services, closing the gap in diagnostic age between children by race, language, and socioeconomic resources (Eisenhower et al., 2021; Sheldrick, Carter, et al., 2022). In this paper, we further examine the experiences of EI-enrolled families who participated in this multi-stage screening protocol by examining whether, following autism diagnosis, there are demographic disparities in the quantity of service hours they go on to receive. We employ the following methods to capture the complexity and intersectionality of health disparities: (1) using open-ended questions to assess race/ethnicity using categories most meaningful to participants, and (2) using latent class analysis to examine the intersectionality of caregivers’ demographic characteristics in predicting the service receipt of young autistic children.
The Part C Early Intervention Context
Under Part C of the Individuals with Disabilities Education Act (IDEA, 2004), children ages 0–36 months who have delays in one or more areas of development, including delays in socialization or communication often associated with autism, are eligible to receive EI services. In the Massachusetts (MA) EI context, the weekly services received by children with autism diagnoses encompass both general EI and autism-specific EI services. In MA, 20.1% of all children under age 3 receive EI services, while other states range from 2.1 to 21.9% (average across states: 7.2%; Keating & Heinemeier, 2022). Part C of IDEA mandates that these services are provided at little to no cost to families, and in MA, there are no out-of-pocket costs to receive EI services through designated, general EI agencies. Additionally, those with autism diagnoses are legally eligible to receive intensive, autism-specific Part C EI services through specialty EI agencies.
In this paper we examine whether disparities are present in the number of EI service hours received by children diagnosed with autism. Part C EI does not offer fixed recommendations for the dosage, or number of hours, to be given. Rather, the quantity of services is supposed to be determined by children’s developmental and functional characteristics, such that children with greater developmental needs or lower adaptive skills receive more services (Johnson et al., 2007; National Research Council, 2001). Thus, given that EI services are free and most often delivered in the home, one might hypothesize that the quantity of services is predicted by children’s developmental characteristics rather than only by factors such as socioeconomic resources (Wei et al., 2014; Zuckerman et al., 2017). At the same time, research also suggests that demographics may determine service receipt above and beyond a child’s functional and developmental skills (Bishop-Fitzpatrick & Kind, 2017; Dallman et al., 2021; Suhrheinrich et al., 2021).
Health Disparities in Autism-Related Services
Recent national research shows that eventual rates of autism diagnosis by age eight may now be similar across racial groups in the U.S., although race-based disparities in diagnostic rates are present in some states (Shaw et al., 2023). Nonetheless, the timing of autism diagnoses continues to show disparities by race, with children of color receiving diagnoses at later average ages than White children (Constantino et al., 2020; Parikh et al., 2018), which minimizes opportunities for tailored services in early childhood. Further, once diagnosed, children from marginalized groups -- including children of color, children from low-income households, and children with non-English-speaking parents -- are less likely to receive autism-related services than peers from households with greater resources (Bilaver et al., 2021; Čolić et al., 2022; St. Amant et al., 2018; Yingling & Bell, 2019). These disparities exist not only in both general EI and autism-specific EI services (Khetani et al., 2017; Mendez et al., 2024; Shenouda et al., 2022), but also in general health care (Dallman et al., 2021; Karpur et al., 2018) and, for children older than 36 months, in school-based services (Bilaver & Havlicek, 2019; St. Amant et al., 2018). However, while the presence of disparities is well-documented, there is mixed evidence as to whether disparities are in fact primarily due to disparities in the initial autism diagnostic process (i.e., the disparate rates and ages of diagnosis), and whether addressing disparities in diagnosis would resolve or mitigate the disparities in access to subsequent services. Thus, this study assesses whether disparities in service receipt are observed, even in a context in which timely diagnosis was equitably available to children across diverse backgrounds (Eisenhower et al., 2021; Sheldrick & Carter, et al., 2022).
An Intersectional Approach to Health Disparities
Health disparities research has often failed to address the ways that multiple social identities may interact with one another to predict outcomes. An intersectionality framework acknowledges the ways in which one’s social identities interact, resulting in distinct experiences of discrimination and privilege (Crenshaw, 1989; Guidroz & Berger, 2009). Children’s and families’ intersecting identities may have non-linear impacts on experiences of discrimination and privilege that are missed by additive models. Indeed, demographic patterns in the U.S., such as the strong links between immigration, English proficiency, and education (U.S. Census Bureau, 2022), or between education and income (Semega et al., 2016), suggest that demographic factors are not effectively examined in isolation or in additive models, in which each category of sex, race/ethnicity, and language is included as a binary variable in a regression model, which requires that effects of sex be added to those of race/ethnicity and to those for language to predict outcomes for any particular group. In the EI service context, these multiple, overlapping identities -- such as race, ethnicity, national origin, language, income, and education -- may interact to predict distinct patterns of experiences with EI services.
Using latent class analysis (LCA) to group families based on intersecting combinations of marginalized and privileged identities can enable us to consider demographic factors through a lens of intersectionality, going beyond additive models. While linear models require exponential increases in sample size for each factor of social identities, LCA enables a consideration of multiple demographic factors relative to one another, without the heightened need for statistical power. LCA thereby allows one to examine intersectionality among social identities in one’s sample. In our study, we will then examine whether these identity groups, or latent classes, in our sample are linked to distinct outcomes in terms of receipt of EI services (Goodwin et al., 2017).
Measuring Race and Ethnicity
Race and ethnicity are crucial aspects of identity that relate to health outcomes. Assessing race and ethnicity through open-ended self-identification has advantages over the close-ended, multiple-choice items relied on in most health disparities research, such as that mandated in National Institutes of Health (NIH)-funded research and outlined by the U.S. Office of Management & Budget’s Directive Number 15 (Office of Management and Budget, 1997). These close-ended approaches, often with five race options and two ethnicity options, have been shown to misclassify as many as 56–78% of respondents and to result in high missingness, especially for people of color (Eisenhower et al., 2014; Gore et al., 2021; Viano & Baker, 2020). An open-ended format, when aggregated, may provide more accurate groups than forced-choice formats and may better represent people’s racial and ethnic identities (Suyemoto et al., 2020; Woolverton & Marks, 2023).
The Study Context
Our sample was drawn from a larger, university-community partnership in which we implemented a multi-stage autism screening and diagnostic assessment program with three Part C EI agencies in MA. Our partner EI agencies served high numbers of children from racially and ethnically marginalized backgrounds (69.2–84.3%), Spanish speakers and other linguistic minority families (28.3–60.6%), and low-income households (61.0-76.3% publicly insured or uninsured; MA Dept. of Public Health, 2016). This multi-stage screening and diagnostic program was aimed at reducing racial, income, and linguistic disparities in the rates and ages of autism diagnosis. The program was largely successful in increasing rates of autism detection and reducing racial, linguistic, and economic disparities in autism diagnosis (Eisenhower et al., 2021; Sheldrick & Carter et al., 2022). The current study looks at what happens after these children were diagnosed with autism in terms of the intensity of EI services they went on to receive.
The Current Study
We examined whether health disparities were present in the intensity of services received among young children with diagnoses of autism spectrum disorder (ASD) in the Part C EI context in Massachusetts (MA). Given that our sample had received a screening and diagnostic program that showed evidence of equitable access to early autism diagnosis, we were interested to examine whether health disparities in post-diagnostic services emerged even in the context of equitable diagnostic rates. We examined services prior to age three years, when children can benefit from no-cost general EI and autism-specific EI services in MA.
Methodologically, we employed an intersectional, rather than an additive, approach to examining identity-related experiences of disparities in service access. First, using LCA, we considered whether children could be meaningfully categorized into latent classes based on demographic identity factors. Second, we examined whether these identity-based latent classes predicted differences in the intensity, or quantity, of EI services children received following their autism diagnosis. We used open-ended, self-identification items for race and ethnicity, to improve upon the misclassification that is common in race and ethnicity checkbox items. Finally, we simultaneously considered whether measured child developmental characteristics predicted EI service receipt. We expected that, not only would demographics-based groupings predict the quantity of EI services received -- evidence of disparities -- but that these demographic groupings would predict service receipt above and beyond developmental characteristics.
Methods
Participants and Procedure
Participating children were receiving Part C Early Intervention (EI) services and had received a diagnosis of autism through a larger, multi-stage, EI-based, autism screening and evaluation program. The study consists of two samples. The full sample of 508 children who received an autism diagnosis through the larger study (August 2014 – May 2019) were included in latent class analysis. A follow-up subsample of 225 children, comprised of parents who were interviewed about their number of hours of EI service receipt post-autism diagnosis, was used to examine differences in service receipt across latent classes. The full sample provided demographics and child developmental measures during diagnostic assessments held as part of the multi-stage screening and assessment program. The follow-up subsample provided parent-reported EI services during quarterly phone interviews at child ages of 30, 33, and 36 months.
Full Sample Participants
Participants included in the full sample were those who had received an autism diagnosis through the larger study. All were enrolled in EI, had a parent1 who spoke sufficient English or Spanish for screening completion, and had no medical conditions that might interfere with administration or interpretation of diagnostic measures. Of the 508 children, 80.4% were boys. On average, children had been diagnosed at 27.7 months (range: 14–39 months); although Part C EI services end at 36 months, four of these children received their diagnostic evaluation after turning 3 years old due to cancellations or other delays. Parents completing the demographic surveys were mostly female (85.1%; 84.3% biological or adoptive mothers), were 34.0 years old on average (range: 18–75), and 64.2% were partnered (i.e., married or cohabitating). As shown in Table 1, 56.0% of families earned less than 185% of the MA poverty level. Just over half of parents were born outside the U.S., having resided in the U.S. for 12.6 years on average (range: 0–37); 41.6% reported English as the primary language spoken to the child. Most (83.3%) families provided demographic data for two parents, while the remaining 16.7% indicated that the child did not have a second caregiver involved in child-rearing.
Table 1
Demographics of the full sample and the follow-up sample participants
Follow Sample (n = 508) | Follow-Up Sample (n = 225) | ||||
|---|---|---|---|---|---|
Child Demographics | |||||
Average age at diagnosisa | 27.7 mos. (range 14–39; SD: 4.8) | 26.7 mos. (range 16–35; SD: 4.0) | |||
Time elapsed from diagnosis to 1st service interviewa | N.A. | 5.6 months (range: 1.1–19.5; SD: 3.4) | |||
Sexa (% male, female) | 80.4%, 19.6% | 81.3%, 18.7% | |||
Family Demographics | |||||
Parent 1 | Parent 2 | Parent 1 | Parent 2 | ||
Gender (% female, male, other gender) a | 99.3%, 0.2%, 0.4% | 1.2%, 98.3%, 0.5% | 99.0%, 0.5%, 0.5% | 1.6%, 98.4%, 0.0% | |
Not partnered | 35.8% | 17.4% | 33.2% | 10.7% | |
Employment: working a paid job | 55.2% | 84.7% | 51.6% | 83.6% | |
Earning < 185% of poverty level | 56.0% | 62.3% | |||
Language(s) spokena English Spanish Other Bilingual (Eng/other) | 41.6% 37.5% 17.9% 3.0% | 40.9% 40.2% 16.0% 2.9% | 44.3% 35.7% 18.1% 1.9% | 39.9% 38.8% 18.0% 3.3% | |
English Proficiency | |||||
Poor, fair, good: | 37.6% | 36.4% | 38.9% | 35.0% | |
Very good, excellent, or native speaker: | 62.4% | 63.6% | 61.1% | 65.0% | |
Born outside the U.S. | 53.1% | 55.2% | 52.4% | 54.4% | |
Years in the U.S. (for those born outside U.S.)a | 11.6 yrs | 13.8 yrs | 11.4 yrs | 13.9 yrs | |
Educ: High school or less | 15.6% | 20.8% | 20.4% | 22.0% | |
Race / Ethnicity: Asian Black Latiné Othera White | 4.7% 20.1% 44.7% 9.2% 16.1% | 4.1% 17.9% 39.0% 10.3% 14.6% | 4.4% 18.7% 45.8% 9.9% 16.4% | 3.1% 15.1% 39.1% 11.4% 15.6% | |
Follow-up Sample Participants
A smaller subset of 225 children diagnosed with autism, drawn from the full sample, were followed through quarterly parent interviews to track their receipt of post-diagnostic EI services. Families were excluded from follow-up for the following reasons: family was enrolled prior to the initiation of quarterly interviews (n = 105), child’s age at diagnosis was > 35.5 months, precluding inclusion in the 30-, 33-, or 36-month service interviews, which were conducted six or more weeks post-diagnosis (n = 44), or the family could not be reached (n = 134). Follow-up demographics are in Table 1. Relative to families not included in the follow-up (n = 283), follow-up families were more likely to earn < 185% of the poverty level [χ2(1) = 5.85, p < .05] and Parent 1 was more likely to have a high school diploma or less [χ2(1) = 6.47, p < .05], but did not differ on other Table 1 demographics. Most (82.2%) follow-up sample respondents provided demographics for two parents, while 17.8% indicated that the child did not have a second parent or caregiver who was involved in child-rearing.
Measures
Family Demographics Included in the Latent Class Analysis
Demographics were provided via parent questionnaire during the child evaluation, in English or Spanish and in writing or orally, based on parent preference. To assess parent place of birth, parents were asked, “Where were you born? Please provide city, state, country,” and, when relevant, “Where was the second parent born? Please provide city, state, country.” Responses were grouped into U.S.-born (including Puerto Rico) and non-U.S.-born. To assess parent English proficiency (ELP), parents were asked (1) “What was the first language you learned to speak?” and (2) “If English is NOT your first language, how good is your ability to speak and understand English?” (rated as poor, fair, good, very good, or excellent). Parents who indicated English was their first language or rated their English ability as very good or excellent were categorized as High ELP; all others were categorized as Low ELP.
To assess poverty status, parents reported household income from all sources before paying taxes; annual, monthly, and weekly income values were provided across 10 response options ranging from $0–15,000 to $125,001+. Parents also reported how many adults and children relied on this income (household size). Based on income and household size, families were grouped as above vs. below 185% of the state poverty level ($46,424 for a family of four), as 185% is the eligibility cut-off for multiple sources of public assistance (ASPE, 2019); when income was missing, we relied on self-reported use of public assistance. To assess parent education, parents reported the highest grade in school they completed, with eight options ranging from ≤ 8th grade to professional or graduate degree. Education was dichotomized as high school diploma or less vs. at least some post-high school education.
For parent race and ethnicity, we used two open-ended items (Suyemoto et al., 2016):
-
Race is based on how you look (often skin tone or facial features) and how you think of yourself (e.g., Black, Asian, White, etc.). In your own words, to which race or racial group(s) do you belong?
-
Ethnicity typically emphasizes the common history, nationality, geography, language, food, or dress of groups of people (such as Haitian, African-American, European-American, Dominican, Irish, Cantonese, etc.). In your own words, to which ethnic group(s) to you belong?
We later aggregated parents’ responses into categories based on the groups present in the responses. Because parents frequently reported race and ethnicity across both items (e.g., Latino for race and Latino for ethnicity), the resulting categories reflect both race and ethnicity.
To assess parent relationship status, parents were asked their current relationship status (multiple responses were allowed). Parents who endorsed: (a) married, living together, domestic partnership were classified as partnered; (b) single/never married, separated, divorced, widowed, or engaged-but-not-living-together were classified as unpartnered. Regarding employment status, parents indicated whether they were in paid full-time jobs, paid part-time jobs, or not working a paid job. Responses were dichotomized into employed and not employed.
Child Characteristics
These measures were collected during diagnostic evaluations and are included here to understand how child characteristics contributed to children’s service receipt within EI.
Developmental Characteristics
The Mullen Scales of Early Learning (MSEL; Mullen, 1995) includes four scales to assess developmental functioning in children 0–68 months: Visual Reception, Fine Motor, Receptive Language, and Expressive Language. We utilized raw scale scores rather than T scores, while covarying for child age in months, to maximize variability, as many children’s raw scores corresponded to T scores < 20. Lower scores indicate fewer demonstrated skills in a given area. The MSEL has shown good internal reliability, test-retest and interscorer reliability, and validity in autistic and non-autistic samples (Swineford et al., 2015).
Autism Characteristics
The Autism-Diagnostic Observation Schedule, Second Edition (ADOS-2; Lord et al., 2012), a diagnostic observational tool for autism, was conducted by a research-reliable clinician. The ADOS-2 has strong inter-rater and test-retest reliability, internal consistency, and content and construct validity (McCrimmon & Rostad, 2013; Zander et al., 2016). We use the Social Affect, Restricted & Repetitive Behaviors, and total algorithm scores.
Adaptive Behaviors
The Vineland Adaptive Behavior Scales, 3rd Edition (VABS-III; Sparrow et al., 2016), a parent interview measure, provides an Adaptive Behavior Composite (ABC) standard score based on four domains: Communication, Daily Living Skills, Socialization, and Motor Skills. Lower scores indicate the routine performance of fewer adaptive skills. The VABS-III has shown good internal consistency, test-retest reliability, interrater and inter-interviewer reliability, and validity (Pepperdine & McCrimmon, 2017).
Follow-Up Sample Post-Diagnostic Interviews
Receipt of Intervention Services
Within the follow-up sample (n = 225), parent phone interviews were conducted every 3 months after the child’s autism diagnosis to track children’s receipt of Part C EI services. Interviews were conducted in English or Spanish (64.6% English). To determine the number of hours of Part C EI services children were receiving per week, interviewers asked parents to report on each type of EI therapy [e.g., speech therapy, physical therapy, occupational therapy, applied behavior analysis (ABA)] received, including services provided by their general EI agency, autism specialty EI agencies, and any out-of-pocket or insurance-covered services. Parents reported the frequency of sessions/week and session length (i.e., number of minutes per session); responses were then combined to determine, for each child, the mean number of weekly service hours received. Weekly service hours were calculated for families with at least one quarterly interview across child ages 30, 33, and 36 months; when multiple interviews were completed, weekly service hours were averaged across interviews.
Time Elapsed between Autism Diagnosis and Interview
We expected that children for whom more time had elapsed between receiving an autism diagnosis and the quarterly interview may have had more “ramp-up” time in which to begin more EI services. To account for this, we calculated time elapsed between date of autism diagnosis (date of diagnostic evaluation) and date(s) of first phone interview(s) in days, and we covaried this when predicting service hours.
Analytic Approach
Prior to our main analyses, we used independent-samples t-tests and chi-square tests to examine differences between the 225 follow-up participants and the 283 participants without follow-up data, and we examined missingness and distributional properties of all key variables.
Next, we conducted latent class analysis (LCA), a model-based clustering method, using R, in the full sample to examine how participants’ demographic characteristics grouped together. LCA characterizes classes, or groups of participants based on the intersection of observed characteristics (Lanza & Rhoades, 2013); as such, LCA allowed us to explore demographic intersectionality or intersectionality of marginalized statuses. Factors included in the LCA were race/ethnicity, poverty, education, immigration status, English proficiency, employment status for each parent, and presence of a cohabitating second parent, as the latter is commonly linked to socioeconomic resources. Our model selection criteria included Akaike’s Information Criterion (AIC), the Bayesian Information Criterion (BIC), and adjusted BIC (ABIC), with smaller values indicating improved fit. The p-values for the bootstrap likelihood ratio test (BLRT) and entropy > 0.8 were prioritized when deciding the number of classes. We also considered theoretical and conceptual interpretability of the solution (Kim, 2014; Nylund et al., 2007). Each family was assigned to the class for which their estimated probability of membership was highest. Next, we used linear regressions and ANCOVAs in SPSS v. 26 in the follow-up sample to test whether latent class was associated with weekly hours of services received and/or child characteristics.
Results
Preliminary Analyses and Missing Data
Of the demographic variables used in the LCA (Table 1), no data were missing. Of the developmental measures shown in Table 4 (MSEL, ADOS-2), 0.2-1.8% of data were missing, while adaptive behavior data was missing at 32.5%, as the VABS-3 was administered in only the last four out of five years of data collection; during these four years, VABS-3 was missing at 0.9%. Little’s MCAR test on the follow-up sample indicated that data cannot be assumed to be missing completely at random [χ2 (25) = 3.68, p > .05]. To address missing follow-up sample data we used multiple imputation with five imputed datasets (Graham et al., 2007) and predictive mean matching under the missing at random assumption.
Identifying Latent Classes of Demographic Factors
We conducted latent class analysis with our full sample (n = 508) to identify how participants’ demographic characteristics clustered together. We formed typologies of intersecting demographic characteristics by running latent class models with 4–10 classes; fit indices for the 4–8 class models are shown in Table 2. Solutions with 8–10 classes were ruled out based on a notable increase in the BIC relative to the 7-class solution.
Table 2
Fit Criteria for Latent Class Analysis models
Model | Log Likelihood | AIC | BIC | ABIC | VLMR LRT | LMR ALRT | BLRT | Entropy |
|---|---|---|---|---|---|---|---|---|
4 Class | -3541.581 | 7233.161 | 7550.448 | 7312.388 | 0.0000 | 0.0000 | 0.0000 | 0.903 |
5 Class | -3422.832 | 7033.664 | 7431.329 | 7132.962 | 0.0430 | 0.0440 | 0.0000 | 0.904 |
6 Class | -3385.88 | 6997.761 | 7475.805 | 7117.129 | 0.2553 | 0.2578 | 0.0000 | 0.891 |
7 Class | -3345.596 | 6955.193 | 7513.616 | 7094.632 | 0.1948 | 0.1968 | 0.0000 | 0.91 |
8 Class | -3305.381 | 6912.762 | 7551.565 | 7072.272 | 0.0843 | 0.0857 | 0.0000 | 0.937 |
The BLRT maintained significance (p < .0001) and entropy was 0.891–0.937 for all displayed models; there was a notable decrease in log likelihood, AIC, and adjusted ABIC as the number of classes increased from four to seven. While the six- and seven-class solutions contained lower ABIC values compared to the smaller classes, the scree plot “elbow” point in the BIC occurred at the five-class solution. The five-class solution has the lowest BIC, and the LMR ALRT is significant at < 0.05 for the five-class but not the six-class solution, suggesting that five is significantly better than four classes, but six is not better than five classes. Thus, as the most parsimonious and theoretically consistent, the five-class solution was selected as the best-fitting model. Table 3 shows estimated conditional probabilities for each demographic factor for the five-class model. Classes were described as follows:
Table 3
Estimated item probabilities for the 5-Class LCA model, showing all demographic variables entered into the LCA
Class 1 (26.2%) | Class 2 (21.7%) | Class 3 (20.1%) | Class 4 (16.1%) | Class 5 (15.9%) | |
|---|---|---|---|---|---|
Unpartnered parents | 0.28 | 0.43 | 0.18 | 0.64 | 0.16 |
Unemployed parent 1 | 0.53 | 0.41 | 0.49 | 0.36 | 0.40 |
Unemployed parent 2 | 0.10 | 0.19 | 0.12 | 0.27 | 0.13 |
Family falls below 185% of the poverty line | 0.79 | 0.53 | 0.49 | 0.66 | 0.19 |
Parent 1 born outside the U.S. | 0.97 | 0.26 | 0.97 | 0.14 | 0.16 |
Parent 2 born outside the U.S. | 0.98 | 0.30 | 0.96 | 0.25 | 0.09 |
Parent 1 has low ELP | 0.88 | 0.15 | 0.60 | 0.01 | 0.0 |
Parent 2 has low ELP | 0.87 | 0.18 | 0.43 | 0.02 | 0.0 |
Parent 1 has high school or less | 0.33 | 0.12 | 0.12 | 0.06 | 0.02 |
Parent 2 has high school or less | 0.51 | 0.16 | 0.05 | 0.06 | 0.04 |
Parent 1 is Asian | 0.0 | 0.01 | 0.24 | 0.0 | 0.0 |
Parent 2 is Asian | 0.0 | 0.0 | 0.20 | 0.0 | 0.03 |
Parent 1 is Black or African American | 0.0 | 0.0 | 0.26 | 0.89 | 0.02 |
Parent 2 is Black or African American | 0.0 | 0.13 | 0.19 | 0.69 | 0.0 |
Parent 1 is Hispanic or Latiné | 0.98 | 0.79 | 0.0 | 0.0 | 0.07 |
Parent 2 is Hispanic or Latiné | 0.87 | 0.65 | 0.03 | 0.05 | 0.0 |
Parent 1 is White | 0.0 | 0.08 | 0.04 | 0.01 | 0.82 |
Parent 2 is White | 0.0 | 0.06 | 0.01 | 0.0 | 0.88 |
Class 1: Latiné immigrant parents living in poverty (n = 133; 26.2%): The largest class, this group included predominantly two-parent Latiné immigrant families, born outside the U.S., with low English language proficiency. They were mixed on employment status and notably less likely than other groups to have post-high school education. This class contained the highest number of characteristics associated with marginalization in the U.S.
Class 2: U.S. born Latiné parents living in poverty (n = 110; 21.7%): Members were predominantly Latiné, U.S. born, two-parent families, with high English language proficiency and high probability of post-high school education; just over half were employed and just over half had family incomes below 185% of the poverty line.
Class 3: Other immigrant parents (n = 102; 20.1%): Relative to other classes, Class 3 was not clearly defined by a single race/ethnicity; Black/African American (19.5–26.3%) and Asian (20.3–24.1%) were most commonly endorsed. Members were almost exclusively born outside of the U.S. and lived with two parents; roughly half earned below 185% of the poverty line (49%). The majority had some post-high school education. Probabilities were mixed for employment and English proficiency (with 60% having another native language besides English).
Class 4: U.S. born Black parents living in poverty (n = 82; 16.1%): This class primarily included Black and African American, U.S.-born, employed, unpartnered parents. Families had a high probability of experiencing poverty, greater Parent 2 unemployment than other groups, high English proficiency or English as a first language, and most had post-high school education.
Class 5: U.S.-born White parents (n = 81; 15.9%): Members were predominantly White, U.S.-born, two-parent families with high English proficiency (100% highly proficient or first language English). They had primarily above-poverty incomes and post-high school education with mixed employment. This class reflected the greatest combination of demographic factors associated with privilege in the U.S. context.
Developmental and Contextual Differences across Latent Classes
Prior to analyzing whether EI service receipt differed across latent classes, we examined whether classes differed in age of diagnosis, age of EI entry, or child characteristics. As shown in Table 4, ANOVAs with post hoc tests indicated that these demographics-based classes differed in age of diagnosis (Class 1 > Class 5 only) but not in age of EI entry. As for child characteristics, most differences between classes were between Class 1 and Class 5, with Class 1 showing lower expressive language, receptive language, adaptive communication, and adaptive motor skills. Class 4 also showed lower receptive language than Class 5. Adaptive motor skills were also lower for Class 1 vs. Classes 2 and 3, and Class 4 vs. 5. There were no differences in visual reception, fine motor, daily living skills. socialization, or total adaptive skills.
Table 4
Comparisons of latent classes on developmental and diagnostic measures (N = 508)
Latent Classes | ||||||||
|---|---|---|---|---|---|---|---|---|
Class 1 (n = 133) | Class 2 (n = 110) | Class 3 (n = 102) | Class 4 (n = 82) | Class 5 (n = 81) | Total sample | |||
Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) | Overall Mean (SD) | R2 | F | |
Age at Diagnosis (months) | 28.5a (4.7) | 27.7 ab (4.8) | 27.8 ab (5.3) | 27.6 ab (4.4) | 26.5b (5.2) | 27.7 (4.9) | -- | 2.06 |
Age at Early Interv. Entry | 22.2 (6.6) | 20.4 (7.6) | 21.7 (6.3) | 22.0 (6.1) | 21.0 (6.2) | 21.5 (6.6) | --- | 1.06 |
Mullen (MSEL) Scores | ||||||||
Receptive Language | 12.7a (4.2) | 13.5ab (5.1) | 13.5ab (5.5) | 12.3a (4.8) | 13.7b (7.3) | 13.1 (5.3) | 0.106 | 11.73*** |
Expressive Language | 13.4a (4.1) | 14.8ab (5.0) | 15.3ab (5.1) | 15.2ab (5.3) | 15.3b (7.0) | 14.7 (5.3) | 0.206 | 25.63*** |
Visual Reception | 22.3 (4.6) | 22.5 (3.8) | 21.7 (5.4) | 21.0 (4.3) | 21.9 (5.2) | 21.9 (4.7) | 0.144 | 16.68*** |
Fine Motor | 21.7 (3.8) | 21.5 (3.5) | 20.8 (4.1) | 20.6 (3.4) | 20.9 (3.9) | 21.2 (3.8) | 0.246 | 32.17*** |
Early Learning Composite Score | 57.5 (7.5) | 60.1 (10.5) | 59.9 (13.2) | 57.3 (8.4) | 63.8 (17.1) | 59.5 (11.6) | -- | 4.65** |
ADOS-2 Algorithm Scores | ||||||||
SA Total Score | 14.6 (4.3) | 15.1 (4.2) | 15.3 (3.6) | 15.4 (3.7) | 14.8 (3.6) | 15.0 (4.0) | -- | 0.72 |
RRBs Total Score | 5.4 (2.0) | 5.5 (1.8) | 5.4 (1.5) | 5.6 (1.8) | 5.5 (1.6) | 5.5 (1.8) | -- | 0.31 |
Overall Score | 20.0 (5.1) | 20.6 (4.7) | 20.7 (4.2) | 21.0 (4.2) | 20.3 (4.0) | 20.5 (4.5) | -- | 0.69 |
Vineland (VABS-III) | ||||||||
Adaptive Behavior Composite standard score | 61.1 (11.7) | 63.4 (9.4) | 62.5 (10.8) | 62.2 (10.9) | 64.6 (10.6) | 62.6 (10.8) | -- | 1.06 |
Communication | 50.2a (17.0) | 56.6ab (16.1) | 55.6ab (19.1) | 54.1ab (19.7) | 62.0b (19.1) | 55.1 (18.4) | -- | 3.91** |
Daily Living Skills | 66.7 (14.9) | 67.3 (13.7) | 65.7 (15.4) | 65.4 (14.7) | 63.8 (14.5) | 66.0 (14.6) | -- | 0.60 |
Socialization | 65.2 (12.3) | 65.8 (9.6) | 64.9 (8.8) | 66.1 (9.2) | 67.6 (8.8) | 65.9 (10.1) | -- | 0.64 |
Motor Skills | 69.5a (17.0) | 79.9bc (14.0) | 78.6bc (15.0) | 73.8ab (15.1) | 81.4c (12.5) | 76.0 (15.6) | -- | 8.07*** |
Latent Class as a Predictor of Children’s Service Receipt
Within the follow-up sample (n = 225), we examined whether latent classes differed in their receipt of Part C EI services. Overall, children were receiving an average of 8.8 h/week of Part C EI services at ages 30–36 months (SD = 6.8; range: 0–29.5 h). As Fig. 1 shows, children in Classes 1 through 4 received an average of 6.9–9.1 h/week, with Class 1 (Latiné immigrants living in poverty) receiving the fewest hours at 6.9 h/week on average, while children in Class 5 (U.S. born White parents) received 12.0 h/week on average. An ANOVA showed that the five classes significantly differed in hours [F(4, 220) = 3.58, p < .01]; namely, on LSD post hoc tests, hours were significantly lower for children in Classes 1–4 versus Class 5 (U.S. born White parents). No other classes differed significantly from one another.
Fig. 1
Average hours of weekly Part C EI services by class membership
Note. These are unadjusted number of hours of EI services. Each error bar reflects a 95% confidence interval. The ranges of intervention hours received across classes are: Class 1: 0–25 h; Class 2: 0–29.5 h; Class 3: 0–23 h; Class 4: 0–29.5 h; Class 5: 1–28 h. LSD post hoc tests show that classes 1–4 differed from class 5, reflected by asterisks at *p < .05, ***p < .001. No other differences among the classes were found
×
![]()
Latent Class as a Predictor of Service Receipt, Above and Beyond Child Characteristics
Next, in a multiple regression model, we examined the association between latent class membership and Part C EI weekly service hours after covarying for child characteristics (Mullen, VABS-III, ADOS-2 scores) as well as age at diagnosis and contextual information (time elapsed between date of diagnosis and interview data about service receipt). As shown in Table 5, after child characteristics were included in the model in Step 2, latent class differences were still significant: Classes 1–4 all differed from the reference group, Class 5, in weekly service hours received. When contextual variables (age of diagnosis, time elapsed) were added to the model in Step 3, Classes 1 and 4 continued to show significantly fewer EI weekly service hours than Class 5 in this final model. Children’s expressive language significantly predicted service hours in the first two steps of the model, like in the final model, ADOS-Social Affect scores were associated with service hours. In addition, age of diagnosis predictive of weekly service hours in the final model, with older-diagnosed children receiving fewer weekly service hours.
Table 5
Regression model of demographics-based latent class and child characteristics predicting average hours of Part C EI weekly intervention services in the follow-up sample (N = 225)
Step 1 | Step 2 | Step 3 | |||||||
|---|---|---|---|---|---|---|---|---|---|
Variable | B | SE | p- value | B | SE | p-value | B | SE | p-value |
Mullen Receptive Language | − 0.058 | 0.129 | 0.654 | − 0.077 | 0.124 | 0.535 | − 0.058 | 0.118 | 0.627 |
Mullen Expressive Language | − 0.257 | 0.128 | 0.044* | − 0.321 | 0.127 | 0.012* | − 0.138 | 0.122 | 0.261 |
Mullen Visual Reception | − 0.002 | 0.162 | 0.992 | 0.017 | 0.157 | 0.916 | − 0.011 | 0.147 | 0.939 |
Mullen Fine Motor | − 0.137 | 0.192 | 0.475 | − 0.067 | 0.188 | 0.720 | 0.221 | 0.178 | 0.215 |
Vineland Scaled Score | 0.083 | 0.057 | 0.148 | 0.072 | 0.055 | 0.199 | − 0.006 | 0.055 | 0.911 |
ADOS SA Score | − 0.182 | 0.130 | 0.160 | − 0.208 | 0.126 | 0.099 | − 0.269 | 0.123 | 0.028* |
ADOS RRB Score | − 0.353 | 0.266 | 0.184 | − 0.393 | 0.258 | 0.127 | − 0.196 | 0.242 | 0.417 |
Class 1 | -5.938 | 1.421 | < 0.001*** | -4.929 | 1.345 | < 0.001*** | |||
Class 2 | -3.144 | 1.453 | 0.031* | -2.458 | 1.352 | 0.069 | |||
Class 3 | -3.376 | 1.477 | 0.022* | -2.573 | 1.375 | 0.061 | |||
Class 4 | -3.846 | 1.556 | 0.013* | -3.147 | 1.447 | 0.030* | |||
Age at diagnosis | − 0.502 | 0.167 | 0.003** | ||||||
Time elapsed | 0.045 | 0.049 | 0.361 | ||||||
R2 change: mean (range) | 0.064, (0.054–082) | 0.073, (0.072-0.074) | 0.129, (0.115-0.137) | ||||||
p mean (range) | 0.056 (0.010-0.099) | 0.002 (0.002-0.002) | < 0.001 (< 0.001-<0.001) | ||||||
F change mean (range) | 2.11, (1.75–2.72) | 4.46 (4.39–4.52) | 18.27 (16.38–19.40) | ||||||
Age at Diagnosis as a Mediator of Latent Class Differences in Service Receipt
We next examined whether any child or contextual characteristics acted as mediators of the association between latent class and weekly service hours. Age at diagnosis was the only potential mediator, as it was the only variable that was associated with both latent class (for Classes 1 vs. 5 only; see Table 4) and weekly service hours (see Table 5). To determine whether age at diagnosis explained the association between latent class (Classes 1 vs. 5) and service hours, we conducted an indirect effect analysis with PROCESS Macro version 3.2 (Hayes, 2017) using bootstrap estimation with 5000 samples. The overall indirect effect model was significant [F(5, 219) = 11.80, R2 = 0.212, p = < 0.001]. Latent class (Classes 1 vs. 5) was associated with age at diagnosis [unstand. Beta = -2.31, 95% CI (-3.92, − 0.70), p = .005], and earlier age at diagnosis was associated with a higher number of weekly service hours [unstand. Beta = -0.67, 95% CI (-0.88, -0.47), p < .001]. Consistent with a partial indirect effect, latent class (Classes 1 vs. 5) was less strongly related to service hours after controlling for age at diagnosis, though still significant [unstand. Beta = 3.55, 95% CI (1.03, 6.06), p = .006]. The indirect coefficient was significant, and the CI range did not contain zero [unstand. Beta = 1.55, 95% CI (0.51, 2.77)], indicating a significant indirect effect at p < .05. In sum, age at diagnosis partially mediated the difference in service hours between children in Class 1 and those in Class 5. In other words, children with more privileged identities were able to obtain earlier autism diagnoses and, in turn, received more hours of EI services.
Discussion
Children with autism diagnoses are eligible to receive developmental services until age 3 through the Part C Early Intervention (EI) system; however, in this study of toddlers with autism diagnoses, we observed stark differences in the quantity of weekly Part C EI service hours received between groups of children with distinct patterns of demographic characteristics. Overall, children with autism diagnoses received an average of 8.8 h per week of Part C EI services; however, children from the demographic cluster with the most privileged identities received nearly twice as many weekly service hours as children with the most marginalized identities. These group differences in children’s service receipt were partially mediated by group differences in age of autism diagnosis, despite these differences in age of diagnosis being quite small (approximately one month). By using an intersectional approach relying on latent class analysis (LCA), we were able to highlight disparities that would be missed by examining demographic factors one at a time.
Intersectional Evidence of Disparities in Services
Our LCA identified demographic clusters among families of young children with autism diagnoses who were diagnosed through three, urban Part C EI agencies in Massachusetts (MA). Rather than looking at family demographic factors in isolation or as unique contributors, we took an intersectional approach to capture how demographics and identities -- immigration status, race and ethnicity, caregiver education, household poverty status, number of caregivers, English proficiency, employment -- combine and interact to predict access to developmental services.
The best-fitting model produced five latent classes, or clusters, in our diverse sample of 508 toddlers; four clusters reflected groups holding multiple marginalized identities (Classes 1–4) and one class held more privileged identities (Class 5: U.S.-born White parents; 16%). These clusters reflect broader trends in how demographic factors tend to intersect in the U.S.: for example, Latiné families were the group most likely to have immigrated to the U.S. (Class 1; Budiman et al., 2020), while the group of predominantly Black American families (Class 4) faced the lowest rates of paternal employment, consistent with the employment bias and high unemployment that disproportionately impact Black men over other groups (Bureau of Labor Statistics, 2022).
Disparities Across Demographic Clusters in Part C Early Intervention Services
These demographic clusters differed in the quantity of EI services received. Differences were stark: children in the most privileged group (Class 5: U.S.-born White parents) received an average of 12.0 h/week, reflecting significantly higher services than each of the other four demographic clusters. Meanwhile, the differences in service hours between the other four demographics were not significant. Children in the most marginalized group (Class 1: Latiné immigrant parents living in poverty with low English proficiency) received only 6.9 h/week, while children in the other latent classes received an average of 9.1 h/week (Class 2: U.S.-born, English-proficient Latiné parents experiencing poverty), 8.7 h/week (Class 3: immigrant parents), and 8.2 h/week (Class 4: U.S.-born Black parents experiencing poverty).
Considerations in Evaluating the Number of Weekly Service Hours
It may be helpful to put the numbers for average service receipt into the context of existing service recommendations for young children on the autism spectrum; however, universally accepted standards for weekly service hours do not exist. While historically, 25 + hours per week of autism services have been recommended as a standard (Johnson et al., 2007; National Research Council, 2001), these recommendations stem from a medicalized context in which curing autism was historically the goal, and the dominant model involved time-intensive interventions. Such high hours of adult-directed time, particularly for very young children who still require significant daily naps, may be detrimental to children’s development or leave them overscheduled (Wilkenfeld & McCarthy, 2020). Instead, client-centered approaches suggest that quantity should be tailored to child and family needs or goals, with the message that more is not necessarily better (Pellecchia et al., 2019). Thus, in our study, we cannot conclude that these groups of children ought to be receiving more weekly service hours than they currently receive, without knowing more about the types and quality of services that they are receiving.
What Can We Learn, and What Can’t We Learn, by Focusing on Hours of Services?
Our study is limited to assessing the quantity of services, and not the type, quality, or goals of the EI services received. Regarding type of services, while we tracked receipt of all EI services, we did not analyze separately by service type, partly because parents were often unaware of the specific service types being received. As such, our results do not distinguish between general or autism-specific EI services, between autism interventions and other services (e.g., speech therapy, occupational therapy, physical therapy, or general developmental services), or between types of autism interventions being received (e.g., ABA, Early Start Denver Model, Floortime). Notably, the dominant autism intervention at the time of this study in the Greater Boston region was ABA. Further, EI services show differential benefits across distinct child social-emotional profiles, suggesting a need for tailored service types by child needs (Chavez et al., 2024). Examining service type is a vital future direction given the differences in methods and targeted outcomes across types, as well as how these types fit with individual child needs.
Likewise, the quality or effectiveness of services were not tracked in this study. Given that ABA is explicitly named in insurance mandates in MA and other states (Choi et al., 2020), it is safe to assume that many of the hours being counted in the present study reflect ABA or similar interventions. However, recent meta-analyses indicate that ABA and related interventions have a relatively weak evidence base, in spite of wide use (Reichow et al., 2018; Sandbank et al., 2020) and flaws in their efficacy studies (Rodgers et al., 2020) such as undisclosed conflicts of interest (Bottema-Beutel & Crowley, 2021), and lack of adverse events monitoring (Bottema-Beutel et al., 2021). Further, ABA is experienced by many recipients as harmful, ableist, and marginalizing (Kupferstein, 2018; Shkedy et al., 2021), with many autistic advocates, professionals, and others calling for a paradigm shift toward services that are affirming, reflect autistic people’s input, and align with the neurodiversity paradigm (Pellicano et al., 2013; Pukki et al., 2022). Finally, future research on the effectiveness or quality of services received should also consider both benefits and harms, as parents and providers report that the autism diagnosis and EI process is not without harms (Petruccelli et al., 2021).
Finally, within a given service type, the goals and targets of services may differ dramatically. Better outcomes may depend on whether services align with families’ values and goals, regardless of their intensity (Shyman, 2016). Autistic adults report that their most-prioritized support goals for autistic children include improving child quality of life, increasing adult support for the child, and reducing harmful child behaviors, while their least-prioritized goals were reducing autism characteristics and developing play or academic skills (Waddington et al., 2023). In all, future research should go beyond the number of service hours to examine type, quality, and goals of services, including the degree to which services are effective, affirming, and aligned with child and family goals (Autistic Self-Advocacy Network, 2021).
In spite of the limitations, we argue that identifying disparities in the quantity of service hours is a valuable initial step, by shedding light on the degree of access children and families are being granted to systems of care. Future studies should continue to take an intersectional approach to examining disparities around effectiveness and quality of services. Even at this early age, there is room for services to foster valued, neurodiversity-affirming outcomes for autistic children, such as autonomy, self-determination, and advocacy (Brown et al., 2021; Leadbitter et al., 2021), in line with the desires of many autistic people (AWN, 2024; Hamsho et al., 2023).
Differences Across Demographic Clusters in Child Characteristics
These demographic groupings of children differed on some child characteristics. Those in Class 1 (the children of Latiné immigrant parents living in poverty), and to a lesser degree, Class 4, showed lower language and adaptive skills than Class 5. The results are consistent with decades of study negatively linking language development to socioeconomic hardship (Fernald et al., 2013; Hindman et al., 2016) as well as the evidence that existing language measures fail to fully capture the language skills of Black American children (e.g., Lee-James & Johnson, 2022).
Disparities in Services Above and Beyond Child Characteristics
Children with more clinically-elevated ADOS social affect scores received fewer, rather than more, weekly service hours, an unexpected direction of effect. However, this effect only emerged in the final model; on its own, ADOS social affect showed a bivariate correlation of only r = .035 with service hours, suggesting a suppression effect once age of diagnosis and time elapsed were added. Children with fewer expressive language skills received more service hours, although this effect disappeared in the final model. No other child characteristics – including visual reception skills, receptive language, fine motor skills, or adaptive skills – were associated with children’s service hours. This pattern, in which demographics were the determining factor in children’s service receipt, more so than children’s actual skills, serves as evidence of inequity. This finding is consistent with past research documenting disparities; for instance, in their study of 60 autistic children ages 2–7 years, McIntyre and Zemantic (2017) found that family income predicted number of service hours received, while most child characteristics, such as adaptive skills and autism characteristics, did not. Here too, access to therapies appeared to be impacted, not by child needs, but by one’s privileged or marginalized context.
Interestingly, our findings contrast markedly with the recent findings of Berg et al. (2024) who found that, among their primarily Massachusetts-based sample receiving intensive EI autism services, children received an average of 22 h/week of total EI services -- in contrast to our sample’s average of 8.8 h -- and that service hours were not predicted by any demographic factors. Like our study, Berg’s sample was recruited through an existing, clinic-based, research study, the Boston Outcomes of Autism in Toddlers study. However, the two studies differed notably in demographics: Berg et al.’s sample of primarily White, above-median income, college-educated parents mostly closely aligns with our Class 5, who received an average of 12 weekly service hours. The contrasting findings suggests that incremental disparities in services may be most dramatic between families who are already at the greatest levels of marginalization. Our contrasting results may also reflect the outcomes of additive versus intersectional approaches to analyzing demographic disparities.
Child Age at Diagnosis in Relation to Hours of Services
Older-diagnosed children received fewer weekly service hours on average. Further, age at diagnosis partially mediated the difference in weekly service hours between Class 1 and Class 5. In this case, even a small delay in age of diagnosis -- a difference of two months between Classes 1 and 5 -- impacted children’s subsequent receipt of services. This is consistent with our findings that, in the same larger study, Spanish-speaking families waited longer for an initial diagnostic appointment, due to lower appointment availability (Chavez et al., 2022). Thus, it is likely that the systemic, institutional factors that are present in the diagnostic process, and which result in delayed detection for children with marginalized identities, are also present in the services sector of the EI system, creating inequities in service receipt as well. In addition, age of diagnosis and service receipt may both be influenced by parents’ and providers’ degree of concern about autism (Sheldrick et al., 2019). In terms of improving equity, this finding suggests that, if we were to make age of autism detection more equitable across children, we might lessen at least some of the disparities in children’s service uptake.
Putting Intersectional Service Disparities into Context
A crucial part of an intersectional lens is considering the ways in which participants’ identities are shaped by systemic oppression and social context. As noted, in the U.S. context, disparities are pervasive across systems meant to provide developmental and educational therapies, as well as general health care, contributing to continued oppression. Singh and Bunyak (2019) in a systematic review, identified (1) economic hardship, (2) restrictive U.S. immigration policy, and (3) lack of professional competence among providers (including presence of bias and racism) as common barriers to provision of equitable care to autistic children. To the first point, we found support for the impact of economic hardship -- with Classes 1–4, marked by high poverty rates, receiving fewer EI services than children in Class 5 – in spite of the fact that EI services are free of charge to families in MA. Indeed, the economic barriers to accessing services go beyond the potential cost of therapy. To the second point, immigrant status appreciably shaped our LCA groups, with children of non-U.S.-born parents receiving fewer hours. Risk of deportation, restrictive U.S. immigration policies, fear, and social isolation may hinder immigrant families’ ability to seek services. To the third point, variability in EI professional competence, such as lack of culturally appropriate resources, multilingual providers, or culturally responsive interpreters, may discourage families from accessing or increasing care. Further, provider biases may impact care coordination, such as decisions around how many hours to recommend, when to increase services, and types of services to offer (Schnierle et al., 2019).
Limitations and Future Directions
We exercised caution when interpreting these results. The effects of privilege and marginalization operating in the Part C EI services, such as the availability of proximal EI agencies or the availability of bilingual, culturally responsive staff, may differ in other U.S. regions. Further, our demographics reflect that of our region; for instance, while our large proportion of Latiné families is aligned with the make-up of MA, where 12.4% of residents identify as Latiné (U.S. Census, 2019), this pattern may not be generalizable to other regions.
Additionally, our data reflect parents’ awareness of the number of services hours children received. Future studies should utilize EI billing records or cross-validate parent-reported hours with billing records to assess accuracy. Further, future research should also examine whether there were disparities in the number of service hours per week that are recommended by the diagnosing clinician at the time of autism diagnosis, which may contribute to subsequent service disparities; this information was not available in the context of our protocol. Our study focuses on children’s services in the EI context, prior to transitioning to the school special education system. An intersectional lens to examining disparities in the school context and other systems is also warranted.
Conclusion
Understanding the disparities in service access among families with toddlers with autism diagnoses is critical given the potential benefits of early services on child and family well-being. By taking an intersectional approach, we can consider how disparities in EI services relate to multiple, intersecting marginalized identities. These disparities may not have emerged had we examined each demographic characteristic separately. Our findings demonstrate that even when children receive relatively equitable, early access to an autism diagnosis, significant disparities still emerge in their access to post-diagnostic services. These differences in service receipt are not explained by child characteristics, but they are partially mediated by the differences in age of diagnosis across demographic groups. The study highlights the need to apply an intersectional lens to addressing barriers to equitable care for autistic children in the Part C EI system.
Acknowledgements
The authors would like to gratefully acknowledge the numerous people who have contributed to the ABCD Early Screening Project since it was first established in 2013. We thank Dr. Frances Martinez Pedraza, who initiated the multi-stage screening and detection process that served as the starting point for this larger study, through her efforts to improve screening and detection, especially for Spanish-speaking families. We extend our gratitude to the participating children and families as well as to our partner Early Intervention agencies – Thom Boston Metro Early Intervention Agency, Harbor Area Early Intervention Services, and Bay Cove Early Intervention Services – and providers. Thank you also to the members of the ABCD Project research team for their hard work, commitment, and dedication to serving our families and community agencies, with special recognition to those conducting interviews about service receipt including Victoria Acuna, Brianna Armstrong, Thomas Bell, Lucas Brandao, Sophie Brunt, Luisa Buitrago, Maxwell Cooney, Felicia Doiron, Naphtalie Dorcius, Dustin Ducharme, Frances Dumont Reyes, Mariana Lacolla, Araci Ferreira Legua, Elizabeth Frenette, Mayte Forte, Shannon Gregg, Noah Hoch, Alexa LaRoche, Nada Laroussi, Daniela Marchione, Jackeline Morales, Latisha Morant, Gillian Nolan, Sabrina Ozit, Carolina Paiz Martinez, Nora Portillo, Emily Restrepo, Angelica Rivera, Yerielis Rivera, Brenda Rodriguez, Michael Sacco, Looknoo Patcharapon Thammathorn, Yakira Valerio, Angela Venini, Juan Diego Vera, and Dio Zamora.
Declarations
Ethical Approval
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The current study was approved by the University of Massachusetts Boston Institutional Review Board.
Informed Consent
Informed consent was obtained from individual parent participants included in the study. Children did not provide assent due to their very young age.
Competing Interests
Dr. Sheldrick and Dr. Carter are two of the co-creators of the POSI, which is one of the two initial screeners used to identify the children who went on to receive autism diagnoses in the larger study. They conduct research related to this instrument but receive no royalties. Dr. Carter is also co-creator of the BITSEA, which is the second initial screener used in the larger study. Dr. Carter receives royalties on the sale of the BITSEA, which is distributed by MAPI Research Trust.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.