Patterns of Intervention Utilization Among School-Aged Children with Autism Spectrum Disorder: Findings from a Multi-Site Research Consortium

Abstract

When designing and interpreting results from clinical trials evaluating treatments for children on the autism spectrum, a complicating factor is that most children receive a range of concurrent treatments. Thus, it is important to better understand the types and hours of interventions that participants typically receive as part of standard of care, as well as to understand the child, family, and geographic factors that are associated with different patterns of service utilization. In this multi-site study, we interviewed 280 caregivers of 6-to-11-year-old school-aged children on the autism spectrum about the types and amounts of interventions their children received in the prior 6 weeks. Reported interventions were coded as “evidence-based practice” or “other interventions,” reflecting the level of empirical support. Results indicated that children received a variety of interventions with varying levels of empirical evidence and a wide range of hours (0 to 79.3 hours/week). Children with higher autism symptom levels, living in particular states, and who identified as non-Hispanic received more evidence-based intervention hours. Higher parental education level related to more hours of other interventions. Children who were younger, had lower cognitive ability, and with higher autism symptom levels received a greater variety of interventions overall. Thus, based on our findings, it would seem prudent when designing clinical trials to take into consideration a variety of factors including autism symptom levels, age, cognitive ability, ethnicity, parent education and geographic location. Future research should continue to investigate the ethnic, racial, and socioeconomic influences on school-aged intervention services.

According to current estimates, 1.8% of the United States’ population of 8-year-old children has autism spectrum disorder (ASD), a neurodevelopmental condition characterized by social communication differences, and restricted and/or repetitive patterns of interests or behaviors, including hyper- or hypo- reactivity to sensory information (Maenner et al., 2020; American Psychiatric Association 2013). When designing and interpreting results from clinical trials evaluating treatments for children on the autism spectrum, a complicating factor is that most children receive a range of concurrent treatments. Evidence-based interventions are increasingly available in the community, but numerous factors can significantly hinder or facilitate the implementation and fidelity with which these interventions are delivered (Brookman-Frazee et al., 2012). It is important to understand the various factors associated with service utilization, including the types and amount of interventions that school-aged participants typically receive within standard care and the level of empirical support of services rendered.

An intervention can be categorized as an evidence based intervention (EBI) if its efficacy and/or effectiveness has been demonstrated by at least: (a) 2 high quality experimental or quasi-experimental group design studies conducted by at least 2 different research groups, (b) 5 high quality single-case design studies with a total of at least 20 participants conducted by at least 3 different research groups, or (c) 1 high quality experimental or quasi-experimental study and 3 single-case design studies conducted by at least 2 research groups (Steinbrenner et al., 2020). Currently, 28 ASD-specific EBIs have been identified, including naturalistic developmental behavioral intervention, functional communication training, antecedent-based intervention based on applied behavior analysis (ABA), and social skills interventions (Steinbrenner et al., 2020). Research consistently illustrates that EBI use is associated with improved outcomes, including symptom reduction (Steinbrenner et al., 2020). Nonetheless, many interventions that do not meet EBI criteria are widely utilized to treat core ASD and co-occuring symptoms across settings, including in homes, schools, and community-based organizations (Locke et al., 2016; Pickard et al. 2018; Brookman-Frazee et al., 2012). Some of these may occur within broad therapies (e.g., speech language therapy, occupational therapy), which consist of various interventions that are selected and delivered based on unique patient contexts and the clinical judgment of trained clinicians from broader fields of practice. The extent to which the interventions delivered within these broad therapies are EBIs cannot be discerned based on a named field of practice alone. These broad therapies can be considered types of “evidence-based practice,” – with an unknown use of specific EBIs. Across disciplines, “evidence-based practice” refers to the three-pronged approach that integrates patient context and preferences, clinical expertise, and research evidence, that are ideally appropriately balanced across all three factors (APA Presidential Task force on Evidence-Based Practice, 2006; Satterfield, 2009). Other specific interventions that some children receive (e.g., acupuncture, special diets) have been individually researched and lack efficacy data, suggesting that they can neither be considered an EBI nor a field of evidence-based practice (Steinbrenner et al., 2020).

Research has identified multiple factors related to access to and use of autism-related interventions. Child-related factors, such as younger age, are associated with greater use of complementary and alternative medicines (e.g. vitamins, special diets; Owen-Smith et al., 2015; Shattuck et al., 2011), while older age and milder ASD symptoms are linked to less receipt of allied health services, which include a broad range of services (e.g. physical therapy; Dallman et al., 2020). Family-related factors (e.g. caregiver race, ethnicity, socioeconomic status) also predict access to autism-related interventions (Liptak et al. 2008; Irvin et al. 2012; Lin and Yu 2015). Higher SES has been associated with more use of ABA services and occupational therapy among preschoolers (Irvin et al., 2012). Higher caregiver education levels have been associated with wider variety in services utilized, and greater use of dietary and vitamin therapy and other complementary and alternative medicines (Patten et al., 2013; Owen-Smith et al., 2015). A recent systematic review found consensus across studies that racial and ethnic minorities are less likely to access and use autism-related services (Smith et al., 2020). Additionally, state-specific policies and service systems impact access to and use of autism-related interventions. For example, families living in states with autism insurance mandates report greater use of autism-related interventions (Barry et al., 2017; Saloner & Barry, 2017; Douglas et al., 2017). However, little is known about which of these factors are associated with use of various types of autism-related interventions, including access to and use of EBIs compared to other types of interventions that may not be as effective in alleviating symptoms and improving skills.

The present study aimed to characterize the types and number of hours of various interventions received by school-age children on the autism spectrum, who were participating in a large multi-site research study without an intervention component. The rationale for this lies in the need to better understand service utilization patterns within this particular group. Seminal autism treatment trials have ranged in their reporting of interventions that participants received in the community before and during the study timeframe. For example, Lovaas (1987) tracked the educational placements and number of research-based intervention hours received by participants in all conditions during his 2+-year behavioral intervention study, but did not track or consider the “variety of treatments from other sources in the community such as those provided by small special education classes.” In Dawson et al.’s (2010) Early Start Denver Model (ESDM) trial, control group parents reported the number of hours of “individual” and “group” therapies their children received in the community per week. Their reports yielded averages of 9.1 hours/week of individual therapies and 9.3 hours/week of group therapies. Among ESDM intervention group participants, parents were interviewed with an intervention history form every 6-months and reported the average hours per week (5.2 hours/week) that participants engaged in “other therapies.” Most recently in a multi-site, two-phase trial of the ESDM, Rogers et al. (2019) collected parent reported amount, type, and teacher/interventionist to child ratio of all child interventions received outside of the trial every 6-months, categorized as one of 13 predetermined intervention types (e.g., ABA/early intensive early intervention, special education, typical class, speech therapy). They analyzed these data by study arms in the average hours/week of each intervention type (categorized as one of the 13 intervention types with group intervention hours divided by the number of students per teacher, and categorized as individual or group intervention) and found few significant differences. On average, children in the ESDM intervention group received 17.4 hours of intervention/week and those in the community control group received 14.2 hours/week. Interestingly, Rogers et al. found site-based differences in ESDM treatment outcomes, which could not be explained by interventions received or other child or family factors. However, a more nuanced understanding of evidence-based interventions received (e.g., ABA, social skills training) versus other types of intervention (e.g., time spent in typical classroom, special education, occupational therapy) may have better explained these site differences in child responses to the ESDM treatment. Therefore, nuanced service utilization data that takes the level of evidence for specific interventions rendered into account are important to consider when designing clinical treatment trials and interpreting their results. Additionally, recent reviews stress the challenge of evaluating the impact of educational and clinical services on treatment outcomes, and the need for systematic assessment of such services in clinical trials (McCracken et al., 2021; Dawson, 2021).

The present study builds on previous research in several key areas First, caregiver-reported intervention histories were categorized and analyzed based on their level of empirical support. As outlined above, previous literature illustrates child, family, and geographic factors associated with receipt of various autism-related interventions. However, the present study provides a greater understanding of which factors relate to the use of evidence-based interventions specifically, compared to use of other interventions, including those representing general fields of evidence-based practice and those without empirical support. Second, in contrast to previous studies, caregivers of school-aged children in different regions of the United States were comprehensively surveyed about all interventions that their children were receiving. The current study extended prior work that examined only one type of service delivery, was limited to a single geographical region, or primarily focused on younger children on the spectrum. To this end, we collected information on both child (i.e., ASD symptom severity, cognitive level, age, race, and ethnicity) and family (i.e., caregiver education level and household income) factors, across study sites in five states. We hypothesized significant heterogeneity in use of evidence-based interventions based on these factors. We also hypothesized inequities in the total number of evidence-based intervention hours based on race, ethnicity, caregiver education level, and household income level. To test these hypotheses, we analyzed associations of child and family factors and study site with (a) the number of hours of EBI and (b) Other practices, and (c) the number of different types of intervention.

Method

We utilized data collected by the Autism Biomarkers Consortium for Clinical Trials (ABC-CT; McPartland et al., 2020), a multi-site study of school-aged children on the autism spectrum, designed to examine research methods used in ASD clinical trials.

Participants

Participants (N=280; 215 males) were recruited across five sites (Boston Children’s Hospital (Massachusetts); Duke University (North Carolina); University of California, Los Angeles (UCLA; California); University of Washington (Washington); and Yale University (Connecticut). Selection criteria included child age between 6.0 – 10.5 years old and diagnosis of ASD based on the Autism Diagnostic Observation Schedule-2 (ADOS-2; Lord et al., 2012), the Autism Diagnostic Interview-Revised (ADI-R; Rutter et al., 2003), and expert clinical judgment of Diagnostic and Statistical Manual, Fifth Edition (DSM-5) criteria (American Psychiatric Association, 2013). Additionally, children were required to have a full-scale IQ score between 60–150. The IQ cut-off score of 60 was selected to optimize inclusion of participants with intellectual disability (ID) who could successfully provide data in a longitudinal experimental battery. Twenty-two children had FSIQ of 70 or below, 51 fell between 71–84, 167 fell between 85–115, and 40 fell at 116 or above. All participant clinical characteristics are presented in Table 1. The participants in this sample included 67.9% who identified as white and 81.4% who identified as non-Hispanic. Household income and parental education were measured categorically via a caregiver report survey. All participant and family demographics are presented in Table 2. Informed consent and child assent were obtained for all participants.

Table 1.

Participant clinical characteristics, age, and gender

Characteristic	Mean (SD)	Range
Participant
Age	8.5 (1.6)	6.0 – 11.6
Gender (M:F)	215:65
ADOS-2 Total Comparison Score	7.65 (1.8)	4 – 10
DAS-II Full Scale IQ	96.6 (18.1)	60 – 150

Notes: ADOS-2: Autism Diagnostic Observation Schedule-2;

DAS-II: Differential Ability Scales, Second Edition

Table 2.

Participant and caregiver demographics

Characteristic	n	Percent
Participant Race
White	190	67.9
American Indian or Alaskan Native	2	0.7
Black/African American	22	7.9
Asian	15	5.4
Mixed race	45	16.1
Other	6	2.1
Hispanic	52	18.6
Biological Father Highest Education Level
Less than high school/General educational	47	16.8
development (GED)/High school diploma
Some college/Associate’s degree	63	22.5
Bachelor’s degree/Some graduate school	85	30.4
Graduate degree	77	27.5
Missing	8	2.9
Biological Mother Education Level
Less than high school/General educational	16	5.7
development (GED)/High school diploma
Some college/Associate’s degree	84	30.0
Bachelor’s degree/Some graduate school	85	30.4
Graduate degree	95	33.9
Combined Biological Parent Education Level
Less than high school/General educational	10	3.6
development (GED)/High school diploma
Some college/Associate’s degree	63	22.5
Bachelor’s degree/Some graduate school	83	29.6
Graduate degree	124	44.3
Household Income
0–25,000	19	6.8
25,001–50,000	42	15
50, 001–75,000	32	11.4
75, 001–100,000	39	13.9
100, 001–150,000	66	23.6
Over 150,001	75	26.8
Missing	7	2.5

Procedure

During the first visit, caregivers completed a sociodemographics form and a structured intervention history interview, while children completed diagnostic and cognitive assessments (McPartland et al., 2020).

Measures

Sociodemographic information.

Caregivers completed a written sociodemographics survey that elicited information on participant and caregiver race/ethnicity, annual household income, and parental educational attainment. The highest level of education reported for either the biological mother or father was included in analyses.

Intervention history.

We adapted the Autism Centers of Excellence (ACE) intervention history survey into a caregiver interview designed to improve reporting accuracy. The intervention history interview (Supplemental Document 1) included a comprehensive list of 18 ASD-specific interventions developed for use in large-scale multidisciplinary studies funded by the National Institutes of Health. Clinicians administered the interview to caregivers by first stating each intervention type and asking caregivers to indicate whether or not their child had received each intervention in the preceding six weeks. Clinicians then asked caregivers to provide specific names of any additional interventions that their child received and information regarding any caregiver-training the family received. If caregivers indicated that their child had received an intervention, they were asked to provide the total number of hours the child had received during the six-week period. A calendar was brought to the interview to provide caregivers with the specific dates of the six-week range. Thus, information on both the type of intervention(s) the participant received and the number of hours per intervention(s) was collected. Information on where interventions were delivered (e.g. home, school, center) was not collected, but parents were encouraged to report on all interventions received across settings. No data were missing for the intervention history interview. All reported intervention use was included in the present data analysis. Complete data were available from 273 children; 7 parents declined to provide income, and missing data were excluded listwise.

Cognitive ability.

The Differential Ability Scale, Second Edition (DAS-II; Elliot, 1990; Elliot, 2007; Early Years and School Age Levels) was utilized to measure cognitive ability. This instrument is normed for children between the ages of 2 and 17 years. The DAS-II yielded standard deviation IQ values based on the General Conceptual Ability (GCA) score (i.e. full-scale IQ), along with subtest mental ages. The mental ages were used to generate ratio full-scale IQ scores. The ratio IQ refers to an individual’s mean mental age across all subtests divided by chronological age, whereas the deviation IQ score measures the degree to which an individual’s IQ deviates from that of the average IQ. If both the deviation and ratio scores were considered valid (i.e., the majority of scores contributing to their calculation were above floor on the measure), the full-scale deviation IQ was utilized to determine study eligibility and entered into subsequent analyses. If the full-scale deviation IQ was invalid (i.e., if more than half the t-scores fell below the floor for a particular cluster), the full-scale ratio IQ score was utilized.

Autism symptom severity.

The ADOS-2 is a standardized play-based observation and measurement system of symptoms and symptom severity associated with ASD. Specifically, this tool allows for the measurement of social and communication skills, play skills, sensory interests and reactivity to sensory stimuli, and restricted and/or repetitive behaviors or interests. The assessment was administered by master’s or doctoral level clinicians who had achieved research reliability on the ADOS-2, and ongoing reliability was maintained consortium-wide. The comparison score provides a standardized measure of symptom severity across modules of the ADOS-2 (Gotham, Pickles, & Lord, 2009).

Coding Procedure

Two independent raters with expertise in ASD interventions, a doctoral student (AS) and a licensed psychologist (JK), coded the list of interventions that caregivers reported during their intervention history interviews. Interventions were coded as either EBI or Other [Appendix A; Appendix B]. Interventions coded as EBI were identified based on a published comprehensive review of ASD evidence-based practices (Steinbrenner et al., 2020); see Appendix A for full criteria utilized to identify EBIs. All other interventions and services that did not meet EBI criteria were collapsed into an Other category; these reported interventions lacked specificity to determine their level of empirical support (e.g., broad evidence-based practice such as speech therapy), lacked sufficient evidence to be categorized as EBI, or were classified as complementary and alternative medications or non evidence-based practices. Consensus coding was utilized, and discrepancies were resolved via discussion. While the intervention history form also gathered data on medication use, those data will be reported elsewhere.

Data Analysis

Two multivariable linear regressions were used to examine factors associated with the total number of evidence based and other intervention hours received during the 6-week period. Predictors were selected on the basis of previously reported factors examining intervention access and use and included child (age, cognitive ability, ASD symptom severity, race, ethnicity) and family (highest parent education level, household income level) characteristics, and data collection site. Categorical variables (i.e., race, ethnicity, parent education, income level, data collection site) were dummy coded and included as fixed effects. We chose this approach because the number of locations was too small to provide good estimates of the distribution of random effects and the number of additional degrees of freedom required was minimal. Because the number of intervention hours were positively skewed and some children received no intervention during the 6-week period, data were shifted by adding a constant of 1¹ to each total and then natural log transformed to normalize the data. Parallel analyses were run for numbers of hours of EBI and Other interventions to determine which factors were predictive of specific types of service usage. Finally, a multivariable Poisson regression was run with the same predictors, using the number of different interventions received by a child (non-transformed) as the outcome to examine determinants of variety in intervention profiles, as opposed to hours received. This analysis combined EBI and Other interventions. Observed values of this count variable ranged from zero to eight unique interventions for individual children out of 21 interventions reported in the sample overall.

Results

Overall Intervention Use for School-aged children

Total intervention hours over the 6-week measurement window ranged from 0 to 79.3 hours/week (mean = 6.0, SD = 9.8). Thirty-two (11%) of the children in the study reportedly received no behavioral interventions in the six weeks prior to the baseline study visit. Of these, 96 (34 %) children received no EBI hours and 59 (21%) received no Other hours. The National Research Council (2001) guideline of 25 hours/week of total intervention services continues to be a widely held guideline for young children (i.e., up to eight years) on the autism spectrum. Only fifteen (5.4%) of the 280 participants in this study of school-aged children reportedly received 150-hours of intervention or more (i.e. ≥ 25 hours per week); most of whom were enrolled in school or camp-based intervention programs or using Picture Exchange Communication System (PECS).

Our naturalistic design enabled examination of the types of intervention that were received most commonly and at the highest number of hours for children aged 6 to 11 years (Figure 1). The most common interventions received were speech and language therapy (59.6%), occupational therapy (41.8%), social skills therapy (39.6%), applied behavior analysis (ABA) (28.9%), and individual social skills therapy (12.1%). The remaining interventions were each received by fewer than 10% of the children in our sample. Comprehensive interventions (e.g. ABA) and use of communication systems (e.g. PECS) were delivered with the greatest number of hours. ABA was received on average 10.97 hours/week; PECS 9.36 hours/week; developmental, individual difference, relationship-based (DIR)/Floortime 4.17 hours/week; pivotal response training (PRT) 3.50 hours/week. All remaining interventions had an estimated average fewer than 1.5 hours/week. Therapies that were not explicitly queried by our interview that arose in parent reports anyway included specialized camps, talk therapy, recreational interventions, and tutoring or other academic support. These were received at an estimated average of 3.02 hours/week and by 21.8% of children in the sample.

Figure 1.

Number of Children Receiving Interventions and Hours Received by Intervention Type

Note: The total number of hours received during the 6-week period were divided by six to provide an estimate of hours per week

Descriptive analysis of total Evidence-Based Intervention and Other interventions indicated that children spent the majority (i.e., 71%) of their time engaged in evidence based interventions (e.g., ABA, social skills interventions). Twenty-nine percent of intervention hours were spent in Other interventions (e.g., counseling at school, occupational therapy). One hundred eighty-four children (66%) received evidence-based interventions; among these children, the combination of EBIs totalled 6.45 hours/week. Other interventions were received by 221 children (79%) with an estimated average of 2.25 hours/week.

Child and Family Characteristics and Location as Predictors of Evidence-Based Intervention Hours

To test the associations of child characteristics (age, IQ, autism severity, race, and ethnicity), family characteristics (income, education level), and data collection site with the overall number of EBI hours, we entered these factors as predictors in a linear regression model with natural log-transformed EBI hours as the dependent variable. These factors accounted for a significant fraction of the variation in EBI hours, R² = .167, F(17,255) = 3.00, p < .001, with more intervention hours reported for children with higher autism symptom levels as defined by ADOS-2 comparison scores, F(1, 255) = 4.8269, p=.029031, and for non-Hispanic children, F(1, 255) = 3.94, p=.048 (Table 3). For each point on the ADOS-2 comparison score, children received 0.15 more EBI hours in the 6-week period. Non-Hispanic received 0.41 more hours. Evidence-based intervention hours also significantly differed by study site, F(4, 255) = 3.92, p=.004, with participants at Boston Children’s Hospital, University of California, Los Angeles, and the University of Washington receiving significantly more hours (i.e., 1.40, 2.48, 1.80 hours, respectively) of EBI than participants at Duke University.

Table 3.

Prediction of Evidence-Based Intervention Hours: Child, Demographic, and Location Model Coefficients

Variable	B	SE B	F	p
Age [Years]	−.099	.061	2.65	.105
Cognitive ability	−.011	.006	3.54	.061
ADOS-2 comparison score	.136	.062	4.82	.029
Racial group (White)	−.083	.219	0.14	.706
Ethnic group (Non-Hispanic)	.526	.265	3.94	.048
Household income (USD)			0.86	.507
0–25,000	−.422	.453		.352
25,001–50,000	.125	.345		.718
50,001–75,000	−.160	.365		.661
75,001–100,000	.004	.334		.989
100,001–150,000	.355	.284		.213
Over 150,001	–	–		–
Bioparent education level			0.94	.420
High school	−.474	.605		.434
Associate’s degree	−.458	.293		.119
Bachelor’s degree	−.282	.250		.260
Graduate degree	–	–		–
Data collection location			3.92	.004
UCLA	1.248	.323		.000
Boston Children’s Hospital	.875	.338		.010
University of Washington	.699	.338		.040
Yale University	.658	.336		.051
Duke University	–	–		–

Notes: The dependent variable is the natural log of 0.5+total intervention hours for the six week period, thus coefficients are adjusted on a natural log scale and must be converted to determine the % change in hours associated with the predictor. B = unstandardized regression coefficient; SE = standard error; Significance is reported for tests of fixed effects. ADOS-2 = Autism Diagnostic Observation Schedule, Second Edition; USD = United States dollars

Child and Family Characteristics and Location as Predictors of Other Intervention Hours

To examine the factors associated with Other types of intervention hours, we fit a linear regression model with log transformed hours for these intervention types (Table 4). The model was significant, R² = .270, F(17, 255) = 5.55, p<.001. As with EBI hours, Other intervention hours were significantly predicted by autism symptom severity F(1, 255) = 18.35, p<.001, such that children who had higher levels of autism symptoms accessed 0.19 more intervention hours during the six week period for each point on the ADOS-2 comparison score. Additional factors that predicted Other intervention hours differed from those that predicted EBI hours, and included child age, F(1, 255) = 11.03, p=.001 and cognitive ability, F(1, 255) = 13.59, p<.001, such that younger children and children with lower cognitive ability received a greater number of Other intervention hours. For each IQ point, 0.01 more hours of other intervention was received in the 6-week period. Parental education level also significantly predicted Other intervention hours, F(3, 255) = 2.81, p=.04, with higher education levels corresponding to more intervention hours (i.e., relative to graduate level education, associates degrees corresponded to 0.37 fewer hours and bachelor’s corresponded to 0.33 fewer hours over the 6-week period).

Table 4.

Child Characteristics and Location Relate to Other Intervention Hours

Variable	B	SE B	F	p
Age	−.134	.040	11.03	.001
Cognitive ability	−.014	.004	13.59	<.001
ADOS-2 comparison score	.175	.041	18.35	<.001
Racial group (White)	.020	.145	0.02	.889
Ethnic group (Non-Hispanic)	.323	.175	3.43	.065
Household income (USD)			0.61	.694
0–25,000	.086	.298		.774
25,001–50,000	.293	.228		.200
50,001–75,000	.072	.241		.767
75,001–100,000	.000	.220		.998
100,001–150,000	−.095	.187		.612
Over 150,001	-	-		-
Bioparent education level			2.81	.040
High school	−.508	.399		.204
Associate’s degree	−.464	.193		.017
Bachelor’s degree	−.396	.165		.017
Graduate degree	-	-		-
Data collection location			1.62	.169
UCLA	.464	.213		.030
Boston Children’s Hospital	.290	.223		.194
University of Washington	.353	.223		.115
Yale University	.509	.222		.022
Duke University	-	-		-

Notes: The dependent variable is the natural log of 1+intervention hours for the six week period, thus coefficients are adjusted on a natural log scale and must be converted to determine the % change in hours associated with the predictor. B = unstandardized regression coefficient; SE = standard error; Significance is reported for tests of fixed effects. ADOS-2 = Autism Diagnostic Observation Schedule, Second Edition; USD = United States dollars

Child and Family Characteristics and Location as Predictors of Intervention Variety

The number of unique intervention services received was summed across all intervention types. For example, if a child received both social skills and speech therapy, their count of unique intervention services received would be two. The median number of interventions received was 3 and ranged from 0 to 8 interventions out of 21 interventions reported across the sample.

To examine the factors that predicted the number of interventions received, we computed a Poisson regression (Table 5). Again, autism symptom severity, χ² (1)=6.94, p=.008 significantly predicted the number of intervention types received. Additionally, the variety of intervention types was significantly predicted by age, χ² (1)=4.34, p=.04, child cognitive ability, χ² (1)=14.84, p<.001, and data collection site, χ² (4)=13.08, p=.01, such that younger children with lower IQ and greater autism symptom severity received a greater number of interventions. Children participating at Yale University, Boston Children’s Hospital, and University of Washington received a significantly greater variety of intervention relative to participants at Duke University.

Table 5.

Child Characteristics and Data Collection Site Relate to the Count of Unique Intervention Services Received

	Likelihood ratio χ2	B	SE	p	Rate Ratio
Age	χ2 (1)=4.34, p=.037	−.050	.02	.047	0.954
Cognitive ability	χ2 (1)=14.84, p<.001	−.009	.00	.000	0.991
ADOS-2 comparison score	χ2 (1)=6.94, p=.008	.066	.03	.009	1.068
Racial group (non-White)	χ2 (1)=1.11, p=.292	−.090	.09	.294	.914
Ethnicity (non-Hispanic)	χ2 (1)=.212, p=.645	.047	.10	.646	1.049
Household income (USD)	χ2 (5)=.1.78, p=.878
0–25,000		−.035	.19	.850	0.965
25,001–50,000		.069	.13	.603	1.072
50,001–75,000		−.123	.15	.412	0.885
75,001–100,000		−.039	.13	.765	0.962
100,001–150,000		.013	.11	.906	1.013
over 150,000		0	-	-	1
Bioparent education level	χ2 (3)=5.39, p=.145
High school		−.323	.24	.181	0.724
Some college		−.248	.12	.033	0.780
Bachelor’s degree		−.119	.10	.225	0.888
Graduate degree		0	-	-	1
Data collection location	χ2 (4)=13.08, p=.01
UCLA		.217	.14	.118	1.242
Boston Children’s Hospital		.284	.14	.041	1.328
University of Washington		.335	.14	.018	1.398
Yale University		.470	.14	.001	1.600
Duke University		0	-	-	1

Notes: B = unstandardized regression coefficient; SE = standard error; ADOS-2 = Autism Diagnostic Observation Schedule, Second Edition; USD = United States dollars

Discussion

Overall, this study provides an approach to understanding patterns of intervention use among school-aged children on the autism spectrum, with a focus on factors associated with the variety, amount, and evidence-base of interventions utilized. Our sample offers an opportunity to explore the variability in interventions accessed by children on the autism spectrum participating in a clinical trial across five different states, and to examine factors that influence intervention access. Specifically, these findings provide an understanding of service utilization patterns among school-aged children on the autism spectrum, including hours of intervention delivered, receipt of interventions with varying levels of empirical support, and the number and variety of interventions received. Moreover, this study investigated relations between child and family factors, data collection site, and the amount and type of intervention services utilized by this sample across five different United States sites.

Intervention Utilization in School-Aged Children on the Autism Spectrum

First, we examined patterns of psychosocial, educational, and behavior intervention use among this sample. Among children aged 6 to 11 years of age, caregivers reported a wide range of total intervention hours (0 to 79.3 hours/week) with children receiving 6 hours/week on average. Caregivers also reported accessing up to eight different interventions (of the 21 queried), with the most common intervention being speech and language therapy. Lastly, this sample spent the majority of their intervention time receiving evidence-based interventions; this finding is promising as it indicates that children within this sample are primarily receiving interventions with strong empirical support for improving ASD-related symptoms.

Child Characteristics Associated with Patterns of Intervention Utilization

We identified several child clinical and developmental characteristics that were associated with intervention amount and type. Specifically, school-aged children who exhibited more severe ASD symptoms received a higher number of EBI hours, Other intervention hours, and number of different intervention types, while younger children with lower cognitive ability received more Other intervention hours. Likewise, younger children and those with lower cognitive ability received a higher number of different intervention types. These results are consistent with previous research illustrating associations between child-related factors (e.g. symptom severity) and receipt of services (Dallman et al., 2020; Wei et al., 2013; Shivers et al., 2018; Shattuck et al., 2011). These findings contribute to the literature by examining these associations among school-aged children specifically, and by providing a greater understanding of service utilization patterns for school-aged children, when interventions are categorized according to evidence level by separating evidence-based intervention from Other types of interventions. Importantly, these findings are relevant to the design of clinical trials, as novel interventions are typically contrasted with a treatment as usual condition that may include evidence-based practices accessed in the community.

Child and Family Sociodemographic Characteristics Associated with Patterns of Intervention Utilization

Our findings are somewhat consistent with previous research highlighting disparities in access to autism-related services based on family and household characteristics (Dallman et al., 2020; Liptak et al. 2008; Irvin et al. 2012; Lin and Yu 2015). The proportion of racial and ethnic subgroups in our study sample roughly represents those in the U.S. overall. Yet, our study was limited by selection and sampling bias that led to the overrepresentation of parents who had obtained education beyond high school and with moderate-to-high household income levels. Furthermore, this study was not powered to detect differences based on racial and ethnic subgroups. These limitations resulted in an insufficient sample to detect suspected differences in intervention use based on racial/ethnic group or socioeconomic status, which are well-documented in extant literature (Smith et al., 2020).

In the context of these limitations, we did find that Non-Hispanic children received more EBI hours than Hispanic children; this is consistent with previous literature highlighting associations between ethnicity and access to autism-related interventions and specialty services (Smith et al., 2020). Our work also extends prior findings to specifically suggest ethnic differences in access to interventions that are evidence-based. Future work is needed to understand the reasons for this difference in access to evidence based-intervention.

We also found that higher parent educational attainment, but not household income level, predicted greater use of Other, non-EBI interventions. Notably, our study sample included only 10 families for which both parents’ highest level of education was a high school diploma or below, and thus overrepresents those with at least some education beyond high school (U.S. Census Bureau, 2017). In contrast to prior literature (Liptak et al. 2008; Patten et al. 2013), household income level did not significantly account for additional variability in intervention hours. Given that ASD-related interventions, especially EBIs, are increasingly provided with minimal financial demands on the family through insurance coverage and public special education services, family income may have been less related to accessing interventions for school-aged children than in past studies. For example, research indicates that unmet service needs for Black children on the autism spectrum were cut roughly in half due to Medicaid waivers, supporting the assertion that financial demands related to ASD interventions may be reduced for families utilizing programs such as Medicaid (LaClair et al., 2019). Additionally, our sample had higher income than that of previous samples which mitigates our ability to detect income-related disparities in accessing services. Overall, the sample included in this study is likely more representative of samples historically included in clinical trials, as compared to community or population-based samples. As a result, these findings may be particularly helpful in interpreting the many prior clinical trials with similar sample characteristics. At the same time, our limited power to detect racial, ethnic, and sociodemographic differences in intervention use highlights the need to over-sample under-represented racial and ethnic groups in future trials to allow for secondary analyses that examine heterogeneity by race/ethnicity. It is also important to consider the resource burdens that trial participation can place on families, which may inequitably prevent those with more limited time, transportation, and scheduling flexibility from clinical trial participation.

Geographical Patterns Associated with Patterns of Intervention Utilization

Our findings highlight the need to consider intervention use differences based on intervention site geography. We found that intervention hours and number of intervention types differed by data collection site, which may indicate systematic differences in accessing ASD services by U.S. state/region. Variability in intervention hours may be influenced by state-level differences in ASD prevalence (Baio, 2012), and the rate at which ASD prevalence increases across states (Sheldrick & Carter, 2012). For example, Sheldrick and Carter (2012) found that the prevalence of ASD in North Carolina had the highest annual growth rate compared to seven other states (i.e., Alabama, Colorado, Wisconsin, New Jersey, Georgia, South Carolina, Missouri). However, participants in our sample from North Carolina received the lowest mean number of hours, suggesting the possibility that state-led initiatives for increasing the supply of ASD interventions may be lagging behind rapidly growing prevalence rates.

Previous research also suggests an increase in service use among children on the autism spectrum in states with autism insurance mandates (Barry et al., 2017; Saloner & Barry, 2017). The five ABC-CT study sites were all located within states with mandates requiring medical insurance to cover intervention for ASD at the time data were collected (California: INS CA INS § 10144.5, 2016; Connecticut: CT G.S. § 38a-514b, 2012; Massachusetts: Mass. Gen. Laws. ch. 32A §25; North Carolina: NC G.S. § 58-3-192, 2014; Washington: RCW 48.20.580, 2008). However, the scope of insured ASD services and which children can access those services continue to vary among these states (Hoffman, 2011). Furthermore, the complex and unequal structure of public education funding across the U.S. results in widely varying funds for school-based ASD services at the state and local level (Digest of Education Statistics, 2017). As such, variability in public education funding and state-led initiatives for autism insurance mandates contribute to this study’s observed state-wide differences in intervention hours.

Implications

This study contributes to an understanding of service utilization patterns, such as concurrent use of interventions among trial participants – an important consideration for future designs and interpretability of clinical trials that are examining response to a specific intervention without always sufficiently considering potential confounding effects from other interventions received or potential variability in other interventions received between individuals or comparison groups. While future research is needed to further explain the state-level factors and policies that may influence receipt of autism-related interventions, these findings may indicate treatment as usual patterns for similar samples in future clinical trials, and provide an understanding of intervention use among school-aged children on the autism spectrum specifically. Additionally, these findings illustrate the extent to which intervention use corresponded with certain clinical characterisitcs. Such information may be valuable to interpreting and designing clinical trials, as these associations may impact participation, eligibility criteria, inclusion criteria, and response to treatment in clinical trials.

These findings may be helpful to consider when designing future multi-site clinical trials and considering the variability in access to and use of services across different sites. For example:

• The severity of ASD symptoms in a sample will likely impact the level of “treatment as usual” among controls. Thus, it is important to include in stratification or randomization protocols. Age, cognitive ability, ethnicity and parental education will potentially impact hours of treatment as usual differently depending on the degree of evidence base.

• The inclusion of school-aged participants in medication trials would benefit from thoughtful consideration of behavioral interventions that may contribute to “placebo” effects in control groups and include a means of capturing the type and hours of behavioral interventions received. Moreover, medications that might enhance response to behavioral therapies may similarly enhance behavioral intervention received in the community.

• Disparities in access to treatment and to participation in clinical trials must be included in study design. Efforts to increase inclusion in trials are critical and will likely increase the heterogeneity of treatment as usual based on baseline opportunities related to factors such as ethnicity and parental education level.

Limitations and Future Directions

The intervention history was collected as part of a study focused on exploring biomarkers in ASD. Thus, ascertainment bias may have limited the sample to families who are more likely to engage with EBIs than would be reported among families who are not involved in a research study with a strong biological focus. As previously discussed, our sample was not sufficiently powered to examine differences in intervention use based on racial/ethnic subgroups. Only 17 families reported household income below $25,000 and higher income groups were disproportionately represented in our sample, which may have limited the ability to detect effects of income. Although study findings advance the knowledge base on service utilization patterns among school-aged children on the autism spectrum, this study also highlights the need to make clinical trial participation more accessible to families from minoritized racial and ethnic groups, families with lower socioeconomic status, and families from a wider range of geographical locations.

There are several limitations regarding intervention history administration. Specifically, clinicians did not utilize a standardized script across sites to provide instructions or descriptions/definitions of the interventions, services, and therapies listed. While clinicians may have utilized all three terms (interventions, services, and therapies) when inquiring about intervention history, it is possible that some support services were missed, and that inconsistency in the administration and terminology may have impacted caregiver reports on intervention use. Likewise, the period surveyed was brief (i.e., six weeks) and thus may have missed services that were received over a longer period. Furthermore, although the intervention history was administered by clinicians to increase reporting accuracy, this study relied on caregiver reports of intervention history, which may include some inaccuracies due to caregivers being unsure or misremembering specific details of interventions received. It is possible that caregivers may have provided socially desirable answers or may not have been familiar with all interventions listed in the interview. There may have been site or interviewer differences in the amount of information provided to caregivers to clarify or answer questions about the interventions listed. Further, the intervention history interview utilized in this study did not elicit information about where services were provided (e.g. school, home), or collect details from caregivers who reported additional interventions that were not included in the pre-specified list. Consequently, some services delivered at school may have been missed and some interventions categorized as “Other” may have been categorized differently with additional information. While this study illustrates one possible method for collecting information related to intervention use, future studies should aim to collect information on intervention history with measures that do not rely entirely on caregiver report (e.g., verify with billing data or IEP service grids when possible; collecting multi-informant reports). Future caregiver interviews about intervention use could be improved with inclusion of questions regarding intervention service setting and structured follow-up questions about services that do not immediately map to pre-specified categories. Although this study aimed to utilize a systematic method for collecting intervention history data, this study highlights the many improvements that could be made to intervention history interviews, to best capture treatment and service utilization in this population.

Conclusion

These findings provide a landscape of the interventions used by a sample of school-aged children on the autism spectrum who are representative of samples participating in clinical trials. Therefore, findings may be utilized to better understand and interpret clinical trial findings, by providing context for service use patterns among school-aged children on the autism spectrum. This study provides insight into who receives which and what amount of interventions, an important reference for clinical trials focused on improving clinical outcomes for school-aged children on the autism spectrum. Future research with samples that better represent socioeconomic, racial and ethnic, and geographic diversity is needed to detect inequities in these nuanced aspects of service utilization. Such findings may inform key targets and approaches for efforts to address inequities in service utilization among children on the autism spectrum.

Highlights.

• Parents of school-aged autistic children were asked about intervention use

• Child, family, and geographic factors were associated with intervention use

• Children received interventions with varying levels of empirical support

• Children spent a wide range of hours receiving intervention

• Research should further explore and consider these associations in clinical trials

APPENDICES

Appendix A.

Intervention Category Criteria

Intervention Category	Criteria
Evidence Based Intervention (EBI)	At least 2 high quality experimental/quasi-experimental group design studies, conducted by at least two different research groups, OR At least 5 high quality single case design articles by three different research groups with at least 20 participants, OR A combination of 1 high quality experimental/quasi experimental study, and 3 high quality single case design studies by at least two different research groups
	(Steinbrenner et al., 2020)
Other	All interventions that did not meet criteria for EBI (Steinbrenner et al., 2020), or lacked sufficient detail to determine EBI status were categorized as “other”.

Appendix B.

Intervention Categories

Evidence Based Interventions (Steinbrenner et al., 2020)	Intervention/Service Received
Behavioral Interventions/Interventions composing ABA	ABA Recreation Camp
	Individual behavioral therapy
	ABA Hebrew Group
Cognitive Behavioral/Instructional Strategies	Cognitive Behavior Therapy
Exercise and Movement	Adaptive Swimming
	Adapted physical education
	Basketball class
	Gymnastics group
	School based yoga
Functional Communication Training	Picture Exchange Communication System (PECS)
Parent-Implemented Intervention	Caregiver-child interaction
	Social skills coaching (for caregivers)
	Caregiver delivered ABA with BCBA or Behavior
	Therapist
	Family training, parenting strategies
	Family based behavioral therapy
Peer-based Instruction and Intervention	PEERS Friendship training
	Peer mentor
	Lego group (OT, social skills, and language)
Sensory Integration	Sensory Integration Therapy
Social Narratives	Social Stories/Narratives
Social Skills Training	Social Skills
	Social skills groups
	Individual social skills instruction (with counselor/social worker)
	Social thinking curriculum
	Social skills summer camp (APEX)
	Camp counselors with social skills focus
	“In classroom” assistance with social skills
Technology-Aided Instruction and Intervention	Robot/tech aided social skills intervention
Comprehensive Treatment Models	TEACCH
	Early Start Denver Model (ESDM)
“Other” Interventions	Intervention/Service Received
Evidence-Based Practices	Speech/Language Pathology (SLP)
	Occupational Therapy (OT)
	Supportive Therapy
	Development of Individual Differences
	Autism Education
	Caregiver training on: Behavior problems
	Caregiver training on: Behavior and managing expectations
	Caregiver training on: Behavioral strategies or training
	Caregiver training on: Challenging behaviors and self-care
	Friendship club
	Caregiver training on: First Steps program at SCAC
	Caregiver training on: SCAC ASD Services
	Caregiver training on: Reward system, challenging behavior
	Individual, social & behavioral therapy
	School based counseling
	Individual counseling with social worker
	1:1 aide (behavioral, social, academic support)
	Special education
	Summer school with SLP, OT and physical therapy
	Talk therapy with psychiatrist/adjustment counselor
	Play therapy
	Psychiatrist
	Psychologist and psychiatrist
	Psychologist
	Counseling/Therapy (individual or group)
	ECPHP (UCLA full day integrated program)
	Family Therapy
	Feeding Therapy
	Functional skills group
	Girls Group (personal hygiene group)
	In home therapist (social skills, daily routine, play with siblings)
	PRT Play Sessions
	Tutoring/academic support
	Girls on the Run, afterschool program
	Tween girl power, YMCA program
	Reading/math special education
	Para support for independent living instruction
Insufficient Evidence or Specificity to Categorize	Individual consulting firm
	Lunch bunch
	After school program
	Pokemon social group
	Equine Animal Therapy
	Music Therapy
	Art Therapy
	Special Diets
	Acupuncture
	Caregiver information session on parenting and technology
	Caregiver training on: Consultation mood
	Caregiver training on: Home-training
	Caregiver training on: Problems with siblings
	Caregiver training on: Skills & strategies
	Caregiver training on: Small group
	Caregiver training on: SS/TAG Program
	Caregiver training on: Conscious discipline
	Caregiver training on: Relational Developmental
	Intervention
	Reading group intervention
	Child life services
	Easter Seals Family Outings
	Home instruction
	Mentoring program
	Military group at school
	Outside guidance
	Personal care provider
	Recreation therapy
	Summer camp for kids with ASD
	Therapeutic mentoring
	Vision therapy
	“Mind up”
	Study organization