Introduction

Autism spectrum disorder (ASD) is a developmental disorder that is reliably diagnosed in early childhood. ASD is characterized by communication problems, social interaction impairments, and restricted, repetitive patterns of behavior and interests (American Psychiatric Association 2013). In the past 10 years, ASD has become a growing public health topic as the prevalence of children with ASD has increased from 1 in 150 in 2000 to 1 in 68 (Centers for Disease Control and Prevention 2014) just 10 years later. ASD has a heterogeneous etiology, and current research has not yet identified a specific biological treatment for the core symptoms of ASD. At present, the evidence-based treatments for core ASD symptoms in young children supported by research are early intensive behavioral intervention (EIBI) (Remington et al. 2007) programs that are based on the principles of Applied Behavior Analysis (ABA), such as Discrete Trial Teaching (Smith 2001), Pivotal Response Treatment (Smith and Iadarola 2015), the Early Start Denver Model (Dawson et al. 2010; Estes et al. 2015), and others (Schreibman and Stahmer 2014).

Children with ASD who are enrolled earlier in intensive early interventions have better outcomes, which include improved daily living and cognitive skills (Peters-Scheffer et al. 2011; Perry et al. 2011; Dawson et al. 2010). With such promising data to support the effectiveness of early intervention services, it is troubling that an estimated 9 in 10 children with ASD are not fully utilizing behavioral health services even with the existence of the Individuals with Disabilities Education Act (IDEA) (Ruble et al. 2005). IDEA mandates the provision of intervention services to children with disabilities, including both general interventions, such as speech and occupational therapy, as well as autism-specific therapies, such as EIBI and emerging interventions such as Naturalistic Developmental Behavioral Interventions (Schreibman et al. 2015) and developmental social-pragmatics (Smith and Iadarola 2015). IDEA does not endorse any particular type of intervention or specify the need to provide “optimal” intervention (National Center for Learning Disabilities 2014). These intervention services are governed by Part C of the IDEA for children under 3 years of age and Part B for preschool-aged (3–5 years) children. In order to address the accessibility of intervention services for children with ASD, the utilization patterns of the service system must first be identified. It is important to identify sociodemographic factors and potential barriers for families of children with ASD who are not fully utilizing intervention in order to develop effective strategies for optimizing service utilization.

This study used data collected in the Childhood Autism Risks from Genetics and the Environment (CHARGE) Study (Hertz-Picciotto et al. 2006) to determine if demographic factors are associated with the number of hours of intervention services children with ASD receive and with how early they are enrolled in these services. Our first aim was to evaluate differences in service utilization based on race/ethnicity. We hypothesized that (a) White families would receive the most hours per week of services compared to other races/ethnicities, and (b) White families would start intervention earlier than non-White families. Our second aim was to evaluate differences in services utilization based on the primary language spoken at home. We hypothesized that (a) English speaking families would receive more hours per week of services than non-English speaking families, and (b) English speaking families would start intervention earlier than non-English speaking families. Our third aim evaluated the differences in service utilization based on whether the family had private insurance or public insurance (e.g., Medicaid) to pay for the child’s birth, which served as a proxy for income and socioeconomic resources. We hypothesized that (a) families with private insurance would receive more hours per week of services than families with public insurance, and (b) families with private insurance would start intervention earlier than families with public insurance. Our last aim was to evaluate differences in services utilization based on maternal education. We hypothesized that (a) families with higher maternal education would receive more hours per week of services than families with lower maternal education, and (b) families with higher maternal education would start intervention earlier.

Methods

Participants

Participants were a subset of 696 children with ASD and their families from the CHARGE study, an ongoing population-based case-control study whose primary focus is identifying the genetic and environmental contributions to ASD risk. The CHARGE study inclusion criteria were children who: (a) were between the ages of 24 and 60 months, (b) lived with at least one biologic parent, (c) had a parent who speaks English or Spanish, (d) were born in California, and (e) lived in catchment areas of a priori selected Regional Centers in California, which are funded through the State’s Department of Developmental Services (DDS) to provide services for individuals with ASD and other developmental delays and cater to both a mix of small urban and rural populations. Although three groups were recruited for the CHARGE study (general population, ASD, other developmental delays), this study focuses on the subset of participating children whose diagnosis of ASD was confirmed by study clinicians. These children were initially identified primarily through lists received from the DDS, as well as through referrals from the Regional Centers, health and service providers, and other studies at the UC Davis MIND Institute. The families of children with ASD who were successfully recruited were more likely to be higher levels of education, more likely to be born in the US, and to have private insurance The CHARGE protocol was approved by the institutional review boards at the University of California in Davis and Los Angeles and the State of California Committee for the Protection of Human Subjects.

Procedures

For this paper, we specifically limited our analysis and description of methods to the subset of children with a confirmed diagnosis of ASD and their families. Participants in our subset were interviewed over the phone by study staff using the Environmental Exposure Questionnaire to collect medical and demographic information. Participants visited the MIND Institute for two study visits that included several diagnostic assessments, such as the Autism Diagnostic Interview-Revised (ADI-R) (Rutter et al. 2005) and the Autism Diagnostic Observation Schedule (ADOS) (Lord et al. 2012). ADI-R and ADOS were used to confirm the presence of ASD. ADOS comparison scores were used to measure the severity of ASD symptoms on a 10-point scale (Lord et al. 2012; Gotham et al. 2009). Also at the study visit, a physician obtained a medical and intervention services and treatment utilization history using the Services and Treatment Interview. Parents also completed the Multiple Language Questionnaire, which was used to quantify the amount of time different languages were spoken at home to determine the primary home language.

Measures

Demographic Data

Mother’s race/ethnicity and education were ascertained using the Environmental Exposure Questionnaire, a structured telephone interview with the mother. Information about type of insurance to pay for delivery came from birth files and served as a proxy for income and socioeconomic resources (income is typically reported by <50 % of study participants). We used maternal race/ethnicity because mothers are usually the primary caregivers and most likely the individual to seek services for their child. Race/ethnicity were divided into five groups: (1) White, (2) Hispanic (any race), (3) Asian, (4) Black, and (5) Mixed or other race. Educational attainment was categorized into three groups: (1) high school or less, (2) some college/vocational, and (3) Bachelor degree or higher. Insurance at time of delivery (i.e. insurance available to pay for the child’s birth) was categorized into public (government program) and private insurance.

Autism Diagnostic Observation Schedule

The Autism Diagnostic Observation Schedule (ADOS) is a semi-structured, standardized assessment of communication and social interaction skills. The ADOS includes four different modules and the assessment administrator choses the one that best fits the child’s expressive language level. It measures Communication, Social Interaction, Play, and Imaginative Use of Materials (Lord et al. 2012). Scoring is then scaled to ADOS comparison scores for symptom severity: Score 1–3 = Minimal symptoms, 4–6 = Mild to Moderate, or 7–10 = Severe (Gotham 2009).

Services and Treatment Interview

This survey was created for the purpose of assessing general usage of any services and treatments and questions were not specific to services and treatments solely for ASD, although our analyses are limited to the ASD subset. Services and treatments could have been attained from multiple sources including government-funded Regional Centers, public school systems, and privately sourced centers. Trained physicians collected information on the types and hours of services received and age of entry into services through any of the previously listed sources. In order to capture the various types of services provided by any of the previously listed sources, we asked about the number of hours received for both individual (i.e. one-on-one; see Table 3 for examples) and classroom-based (i.e. group) services. This distinction also allowed us to investigate if there were socioeconomic or ethnic disparities in service utilization related to service type/format (individual vs. group). We also collected the age of entry into classroom-based programs. Questions included: “What individual and/or group services, related to [CHILD’s] developmental or behavioral condition has [CHILD] received at school and/or outside of school?”, “When did he or she begin classroom programs?”, “How many days per week does child attend school program?”, and “How many hours/day does child attend school program?”

Multiple Language Questionnaire

Primary language spoken at home was collected from the Multiple Language Questionnaire, which is a brief questionnaire developed for the CHARGE study to quantify the proportion of time different languages are spoken in the home. For purposes of this analysis, we classified primary home language as English (if spoken ≥50 %) and otherwise, as non-English.

Statistical Analysis

The services received were categorized into either classroom-based programs or individual services (both private and in school) with the expectation that many of the children would have had both types of services. Our predictors of interest were maternal race/ethnicity and educational attainment, primary language spoken at home, and delivery payer. To determine whether these demographic characteristics were associated with the child’s age of entry for classroom-based programs, we fitted linear regression models adjusted for ASD symptom severity and child’s age at study enrollment. The predictors were modeled individually in separate models and together in one model. β-coefficients and standard errors (SE) were calculated to measure the association between each predictor and the age of entry for classroom-based programs. We conducted log-linear (Poisson) regression models with robust error variances (Zou 2004) to examine whether our predictors were associated with weekly hours of services, with analyses adjusted for ASD symptom severity and child’s age at study enrollment. These analyses were performed for both classroom-based and individual services to investigate if socioeconomic and ethnic disparities are present for either type of service. Hours of services per week were divided into three groups: <15, 15–24, and ≥25 (referent). Prevalence ratios (PR) and 95 % confidence intervals (CI) were calculated to measure the association between each group and the referent. All analyses were carried out using SAS version 9.4 (SAS Institute Inc., Cary, NC).

Results

Of the 696 children with ASD, 656 (94 %) children were receiving classroom-based programs and 678 (97 %) children were receiving individual services. Among families of children receiving classroom-based programs and/or individual services, 59 % of mothers were non-Hispanic White, 92 % were primarily English-speaking, 44 % had Bachelor degrees, and 83 % were privately insured at the time the child was born (Table 1). Sociodemographic characteristics were very similar for those receiving classroom-based services, those not receiving classroom-based services, and those receiving individual-based services; those not receiving individual services were less likely to have graduate education and more likely to have vocational training or college without a bachelor’s degree. Also, children of Hispanic families were more likely to receive no individual services (32.5 %) than to receive individual services (24.3 %), a trend not seen in any other race/ethnicity group. At the time of participation in the CHARGE study, approximately 74 % of the children receiving classroom-based and/or individual services were 2 to 3 years old (Table 2); 58 % of children in classroom-based programs and 61 % of children receiving individual services exhibited severe ASD symptoms (ADOS comparison score ≥7) (Table 2). Among the most utilized individual services reported were speech therapy, occupational therapy, and behavior modification (Table 3).

Table 1 Family characteristics of children receiving and not receiving intervention services, N = 696
Full size table
Table 2 Characteristics of children utilizing classroom-based and individual intervention services
Full size table
Table 3 Individual services ever and currently utilized, N = 678
Full size table

Utilization Predictors and Age of Classroom-based Program Entry

Maternal race/ethnicity was significantly associated with age of entry for classroom-based services independent of our other predictors (maternal education, primary language spoken at home, and delivery payer) and after controlling for child’s age at study enrollment and ASD symptom severity. Specifically, compared to children of White mothers, children of Black mothers were half a year younger, on average, when they entered services [2 years 7 months (1 month) vs. 2 years ½ month (2 months), p = 0.001]. We also found that children whose primary language spoken at home was not English were approximately 3 months older at entry of classroom-based services than children whose primary language was English, which is one-third of a school year behind [2 years 6½ months (½ month) vs. 2 years 4 months (¾ month), p = 0.06]. Mother’s educational attainment and delivery payer were not associated with age of entry for these services (Table 4).

Table 4 Associations between demographic characteristics and age of entry for classroom-based programs, N = 614
Full size table

Utilization Predictors and Hours of Classroom-based Program

Maternal educational attainment was significantly associated with hours of classroom-based intervention services after adjusting for ASD symptom severity and child’s age at study enrollment. Children of mothers with some college education or vocational training were 24 % less likely to receive <15 (vs. ≥25) hours of classroom-based services than children of mothers with a Bachelor degree (PR = 0.76, 95 % CI 0.61, 0.95). However, there were no significant differences in the proportion of children utilizing classroom-based services between 15 and 24 h/week based on maternal education. Maternal race/ethnicity, primary language spoken at home, and delivery payer were not associated with hours of services (Table 5).

Table 5 Associations between demographic characteristics and hours per week of classroom-based programs, N = 540
Full size table

Utilization Predictors and Age of Classroom-Based Program Entry among Children Receiving Individual Services

The CHARGE Study did not collect the age at which the child started individual services. For children who receive individual services, their age of entry for classroom-based services was used as age of program entry. Among children receiving both individual and classroom based services, maternal race/ethnicity was associated with the age of entry for classroom-based services after controlling covariates and independent of our other predictors. Compared to children of White mothers, the mean age of entry for classroom-based services was approximately half a year younger for children of Black mothers [2 years, 6 ½ months (1 month) vs. 2 years (2 months), p = 0.0004]. Mother’s educational attainment, primary language spoken at home, and delivery payer were not associated with age of entry for classroom-based services among children receiving individual services (Table 6).

Table 6 Associations between demographic characteristics and age of entry (in years) for classroom-based intervention services among children receiving individual services, N = 600
Full size table

Utilization Predictors and Hours of Individual Services

Delivery payer was significantly associated with hours of individual services after adjusting for ASD symptom severity and child’s age at study enrollment in addition to all predictors of interest. Children of mothers who had public insurance (e.g., Medi-Cal) to pay for delivery were 37 % more likely to receive <15 (vs. ≥25) hours of individual services per week than children of mothers who had private insurance (PR = 1.37, 95 % CI 1.14, 1.66). We also observed a marginal association between receiving fewer hours of individual services and primary language spoken at home. Children from families whose primary language was not English were more likely to receive <15 (vs. ≥25) hours of individual services compared to children from English-speaking households (PR = 1.33, 95 % CI 0.99, 1.80). Maternal race/ethnicity and educational attainment were not associated with hours of these services (Table 7).

Table 7 Associations between demographic characteristics and hours per week of individual intervention services, N = 616
Full size table

Post-Hoc Analysis

Our post-hoc analysis also revealed that there were no obvious differences in ASD symptom severity across maternal race/ethnicity. In addition, we performed analyses with sampling weights to counterbalance underrepresented groups based on demographic characteristics such as education and race/ethnicity. The results of these analyses did not show any noteworthy differences from the results without the weights. We also performed analyses to determine demographic discrepancies in groups with missing vs. non-missing outcome data. We found that missing age of entry into classroom-based services was associated with Hispanic ethnicity and some college or vocational training for both individual services and classroom-based services analysis (11.5 and 6.4 % missing, respectively). Missing hours of services per week was associated with Hispanic ethnicity, non-English primary language households, and public insurance delivery payers for individual services only (17.7 % missing; Supplemental Tables 1, 2).

Discussion

This study aimed to identify sociodemographic characteristics that are associated with the utilization of intervention services, either classroom-based programs or individual services, in families of children with ASD. We examined four family characteristics (mother’s race/ethnicity, education, language spoken at home, and insurance type at time of delivery) to determine if these are related to number of hours of classroom-based programs and individual services received per week and the age of the child at the entry into intervention services. Our results suggest that maternal race/ethnicity, language spoken at home, and insurance type at time of delivery are related to disparities in the service utilization for families of children with ASD. Our post-hoc analysis also revealed that there were no obvious differences in ASD symptom severity across maternal race/ethnicity, suggesting that differences in intervention service utilization are not due to disorder severity.

We found that the age of entry for classroom-based intervention services was significantly younger for children of Black mothers than White mothers. A possible explanation for this observation is that because we did not capture age of entry into individual services, children of White mothers may have received individual services at an earlier age and resulted in delayed entry into center-based services. Conversely, this trend might be explained by Black mothers wanting and seeking opportunities to enter their children into classroom-based services sooner in order to allow them to reduce the impact on employment (Folk and Beller 1993). General public childcare trends also show that Black families are more likely to utilize center-based care in comparison to their Hispanic and White counterparts (Radey and Brewster 2007). These trends, although not specifically about families of children with ASD, may be the explanation for why we see children of Black mothers entering classroom-based intervention services earlier than children of White mothers. Another possible explanation of this trend is that children of Black mothers may have received a more general developmental delay or behavioral disorder diagnosis prior to an ASD diagnosis, which is consistent with previous research (Mandell et al. 2007; Fountain et al. 2011; Magana et al. 2013). Another diagnosis at an early age may have allowed them to enter services and treatment earlier.

We also observed that children from families whose primary language spoken at home is not English had a slightly higher mean age of entry for classroom-based services and were more likely to receive fewer hours/week (<15) of individual services compared to English-speaking families. In addition, children of Hispanic families were more likely to receive no individual services (32.5 %) than to receive individual services (24.3 %), a trend not seen in any other race/ethnicity group. This finding suggests that language may play a role in accessibility of services for families or in their choice of interventions services, which is consistent with other findings that families of non-white or ethnic minorities are less likely to see a case manager or developmental pediatrician (Thomas et al. 2007; Broder-Fingert et al. 2013; Magana et al. 2013). Public agencies must provide information and support in a broad range of languages and dialects that are culturally attuned. This would allow for more individuals to initially access services, as well as help families understand the nature and utility of early intervention options, especially for those families with cultural stigmas regarding developmental disabilities. Many non-English families lack providers fluent in their native language who can help guide them through the steps necessary for receiving a diagnosis of ASD and accessing intervention services.

Although we did not directly collect information on household income, we used insurance type at time of delivery as a proxy for socioeconomic status and examined whether receiving public assistance affected service utilization; we found that the number of hours of individual intervention services was significantly lower for families with public insurance in comparison to families with private insurance. During the years of the study, insurance status was likely not directly related to the provision of services since ASD-specific intervention services were accessible through government-funded Regional Centers in California for both privately and publicly insured families prior to the implementation of autism insurance legislation (SB946) in 2012. However, it has also been found that children with developmental disabilities from homes with lower income and lower education are less likely to access specialized health services, in part because their parents are less aware of the need for, and benefit from, these services (Porterfield and McBride 2007). Low-income families may also have difficulty accessing services because of the additional time often required to bring children to individual therapies that are not home-based, which may result in lost wages or additional transportation costs. In addition, individual intervention services may include private services that parents paid for in addition to the services provided through the Regional Centers or school districts. This result is consistent with past studies that, not surprisingly, found that families whose household income was over $50,000 per year were more likely to use private speech language therapy and private ABA in comparison to families with lower income levels (Thomas et al. 2007; Irvin et al. 2012). These measures of socioeconomic status are likely reflected in the disparities in access to individual services and may relate to factors such as maternal employment, program availability, and cultural preferences.

The differences in utilization of intervention services revealed in our study can be addressed in several ways. Increased education about ASD and its treatment should be targeted to families with lower household incomes or who have public insurance to support early diagnosis and intervention use. This may result in increased utilization of services and possibly offset the difference in hours of individual services received per week. There can also be increased culturally sensitive outreach targeted to families of diverse race and ethnicities about classroom-based or group-based intervention services that allow children to learn skills through peer interaction (Nahmias et al. 2014). In addition, it may allow parents periods of respite. Both of these suggestions can be carried out through increased ASD screening and better outreach to community groups such as cultural centers, churches, daycares, and public assistance agencies. Lastly, expanded case management would also benefit those unfamiliar with the system and the types of services that can be provided.

Although the study of disparities in services and treatments was not a primary aim of the CHARGE Study, it was designed with population-based recruitment in order to best capture the ethnic and socioeconomic distribution within our catchment areas. We also ensured a bilingual, bicultural (English–Spanish, Latino) study team for all contacts with participants (recruitment, informed consent, interviews, psychometric evaluations and medical histories). As a result, the CHARGE ASD group is one of the most diverse, relative to other US studies of ASD. Nevertheless, funding did not permit us to include other languages. As a case-control study, the population examined was children who had already received an ASD diagnosis and whose families were proficient in either English or Spanish. These families had already sought help early, and we did not capture families with perhaps the greatest barriers to services and/or whose child was not identified with a disability until he/she entered the school system (kindergarten or later). Other limitations include the possible biases that relate to data collection on the child’s age of intervention service entry, hours of services per week, and insurance status. The age of the child when she/he entered services and hours of services that the child received per week were reported by parents but was not further verified. This could have led to errors in recall given that a year or more may have passed between the child entering services and participation in the CHARGE study between the ages of 2 and 5. However, these errors in recall may be random and not specific to any outcome of interest. We collected the insurance status from the birth certificates, which may have differed from insurance at the time of entering intervention services if the families’ economic circumstances changed.

In regards to missing data, we found our primary outcomes, age of entry and hours of services per week, were disproportionately missing for families with Hispanic ethnicity, some college or vocational training, non-English primary language households, and public insurance. We acknowledge that these missing data were not random and may have introduced bias in our measures of associations. Although a quarter of the study sample were Hispanic and over 40 % were non-White or Hispanic, our ability to characterize in detail several minority groups was limited by small sample sizes for Pacific Islanders, Native Americans, and mixed race groups. Additionally, despite exceptional diversity relative to most autism studies, non-White and Hispanic groups were underrepresented relative to the State population. In addition, the families who enrolled into CHARGE were more likely to be higher educated, more likely to be born in the US, and to have private insurance in comparison to the families who receive public insurance. However, we conducted a sensitivity analysis using weights to account for differences in enrollment by sociodemographic factors, and our findings were unchanged.

There were also other variables that would have benefited our study but were not collected such as: (1) age of diagnosis of ASD, (2) any prior diagnoses of other related developmental delays that could have led to uses of services and therapies prior to entry to preschool, (3) specific services received in the home versus in the classroom which may include parent management training, (4) age of entry into individual services, (5) Part B versus Part C funding, (6) quality of the services received, (7) family income, and (8) cultural differences such as attitudes and stigma surrounding developmental disabilities. Our study population was also limited to preschool aged children. As a result, the degree to which our results can be generalized to older ages cannot be determined.

Despite these limitations, our participants had an ASD diagnosis confirmed using gold standard instruments administered by reliable clinicians, ensuring a well-defined sample population. Our sample size was large, and sufficiently diverse to identify racial and ethnic differences in service utilization, age at entry and hours per week. Our analysis also revealed no differences in ASD symptom severity across race and ethnicity. Lastly, we conducted a weighted analysis to account for demographically underrepresented groups in the study, which did not alter the results.

Underlying factors such as the ability to advocate for autism-related services for their child and awareness of the potential services may contribute to the barriers that can lead to the underutilization of services. Efforts should be made to help families understand the benefits of early intervention and to ensure that interventions offered are family-centered and culturally/linguistically accessible. In order to ensure that all families of children with ASD have equal access to intervention services, systematic efforts must be made to increase the availability of culturally and linguistically accessible information and improve community advocacy in order to further identify and reduce barriers to full intervention service utilization.