Therapy and Psychotropic Medication Use in Young Children With Autism Spectrum Disorder

This study is a cross-sectional analysis of data from the AS-ATN, a research registry of 2- to 18-year-old children with ASD enrolled through a network of 17 academic health centers in the United States and Canada. Participation in the AS-ATN is approved by the institutional review board at each site. Written informed consent was obtained from parents of all participants. This secondary analysis study was deemed exempt by the institutional review board at the Children’s Hospital of Philadelphia.

To participate in the AS-ATN registry, participants needed a caregiver to be fluent in English or Spanish and have a clinical diagnosis of ASD (based on Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision criteria by a developmental pediatrician, psychiatrist, neurologist, or psychologist with expertise in ASD) after a standardized diagnostic evaluation that included a demographic and medical history, a physical examination, an Autism Diagnostic Observation Schedule, and standardized measures of cognitive ability and behavioral symptoms. More than 90% of cases scored ASD on the Autism Diagnostic Observation Schedule. Participants who enrolled in the registry between December 2007 and December 2013 were eligible for this study if they were 36 to 72 months of age at the time of registry enrollment and had follow-up data available for analysis (Fig 1). We analyzed the first available follow-up data on therapy and medication use at least 6 months after the date of initial enrollment to ensure that families had at least 6 months after diagnostic evaluation to obtain the interventions recommended by AS-ATN clinicians. We excluded participants >72 months of age at the time of follow-up to focus on children who were likely still in preschool.

Parents completed a demographics questionnaire that included sex, race, ethnicity, insurance type, and primary caregiver education level. After AS-ATN evaluation, participants received treatment recommendations. Follow-up visits were determined on the basis of clinical need (with a goal of participants being seen at least annually), and registry data were collected at each follow-up time point. Data on medication use was collected by the clinician during in-person visits. Data on therapy use was collected by using a parent questionnaire completed at an in-person visit or by mail. Parents indicated on the follow-up questionnaire whether their children were receiving each of the following therapies: speech therapy, occupational therapy, physical therapy, behavioral therapy (including ABA, the Lovaas model, and Discrete Trial Training), verbal behavior therapy, Developmental Individual Difference Relationship-Based Model and/or Floortime, social skills training, family therapy, learning center and/or resource room access, academic tutoring, and other. Duration and frequency of each therapy type were included. Total therapy hours were calculated by adding the hours per week of each individual therapy. Psychotropic medications were categorized as stimulants, α-agonists, selective serotonin reuptake inhibitors, atypical antipsychotics, or antiepileptics. Additional prescription psychotropic medications (including atomoxetine) were grouped into a category called other. Antiepileptic medications were included for mood or behavior indications but not for seizures. Other prescription psychotropic medications (including atomoxetine) were only included if an indication was specified and relevant (attention-deficit/hyperactivity disorder [ADHD], anxiety, irritability, aggression, or sleep).

Data analysis was performed by using SPSS Statistics 22 (IBM SPSS Statistics, IBM Corporation). Variables pertinent to this study were examined for skewed distributions, and those that were not normally distributed were analyzed by using a nonparametric method. The median hours per week of behavioral-based therapy and total therapy were compared between sociodemographic groups by using the nonparametric Kruskal-Wallis or Mann-Whitney U tests. If an overall significant difference in hours per week between the groups was found, a post hoc analysis between pairs of socioeconomic status–based groups was performed by using the Kruskal-Wallis or Mann-Whitney U test. The Pearson χ² test was used to compare the percentage of young children treated with ≥1 psychotropic medications among sociodemographic categories. A post hoc analysis was used to do pairwise comparisons between socioeconomic status–based categories by using the Pearson χ² test. Applying Bonferroni’s principle separately for each of 9 socioeconomic variables, a P <.006 was considered statistically significant.

The multiple linear regression model was used to examine the relationship between the number of therapy hours children with ASD received and variables that were individually associated with hours of therapy (P ≤.01). We used multiple logistic regression to further evaluate the significant factors that were associated with psychotropic medication use (yes or no) using a list of variables that were individually associated with psychotropic medication use (P ≤.05).

Study sample characteristics appear in Table 1. Participants were predominately male and white. Most were from the United States, and 15% were from Canada. Fewer than half had an ASD diagnosis before AS-ATN evaluation. The mean age at the time of AS-ATN diagnosis was 47.3 (SD 7.1) months. The mean age at the follow-up visit from which data on therapy use were extracted was 59.6 (SD 7.2) months and for medication use was 59.7 (SD 7.0) months. When comparing the 965 children evaluated in this study with the 2840 children in the larger AS-ATN registry sample (Fig 1), participants with follow-up data available were significantly more likely to be of non-Hispanic and/or Latino origin (92% vs 89%; P < .001), have a primary caregiver with a college degree or higher (54% vs 42%; P < .001), and to receive an AS-ATN diagnosis at a younger age (47.3 months versus 54.5 months; P < .001). There were no differences in sex, percentage white, or insurance status.

Data on therapy use at follow-up was available for 805 participants (Table 2). The most common therapies were speech and occupational therapy. Most children who received verbal behavior therapy also received behavioral therapy, so overall, only 33.4% of the sample reported receiving a behaviorally based therapy and only 22.4% received training in social skills. Of the 672 participants with data available to calculate the total hours per week of therapy received, the median hours per week of therapy was 5.5 hours (interquartile range 2.0–15.0). Only 14.1% of participants received >25 hours per week (Table 2). Fifty-six children (8.3%) received 0 hours of therapy. Compared with other children with therapy hours available, those receiving 0 therapy hours per week were more likely to have public insurance (66% vs 43%; P = .003) but did not differ on sex, ethnicity, race, or parental education.

Sociodemographic factors were associated with hours per week of total nonpharmacologic interventions (Table 3). Specifically, age at the time of follow-up, having private insurance, having an ASD diagnosis before registry enrollment, and site region were significantly associated with total therapy hours.

We examined the relationship between hours per week of therapy (log transformed to meet the regression analysis assumption of normal distribution) and age at follow-up, ethnicity, insurance status, site region, previous ASD diagnosis, and IQ (selected from Table 3 on the basis of P ≤ .01). The log-transformed hours per week eliminated 56 cases with zero hours of therapy. The multiple regression model showed that ethnicity, previous ASD diagnosis, and IQ were significant predictors (P < .02), but site region, insurance status, and follow-up age were not (P > .10). The adjusted r² for the model was 0.092, indicating that a small proportion of the variance (<10%) in hours per week of therapy that children with ASD received was explained by this model. The reduced model using ethnicity, previous ASD diagnosis, and IQ explained 6% of the hours per week of therapy that children with ASD received. Hispanic participants received 1.80 fewer therapy hours than non-Hispanic participants. Children with a previous ASD diagnosis received 1.75 more therapy hours than children without a previous ASD diagnosis; children with an IQ <70 received 1.25 more therapy hours than children with an IQ >70.

Psychotropic medication use data were available for 613 participants. Of these, 100 (16.3%) received ≥1 psychotropic medication and 20 (3.3%) received ≥2. α-agonists were the most commonly reported (43 participants; 7.0%), followed by stimulants (26 participants; 4.2%), atypical antipsychotics (25 participants; 4.1%), and selective serotonin reuptake inhibitors (19 participants; 3.1%). Few participants were on antiepileptics for nonseizure indications or other psychotropic medications (8 participants; 1.3%).

After adjusting for multiple comparisons, site region did not reach the statistical significance threshold but was included in the multiple logistic regression analysis because of varying frequencies between the different sites in psychotropic medication use. To further evaluate factors associated with psychotropic medication use, we entered variables (univariate P <.05) into a multiple logistic regression model: age at follow-up (continuous variable), race, primary caregiver education, and site region. Follow-up age, primary caregiver education, and site region were statistically significant predictors of psychotropic medication use (Table 4).

More than 6 months after receiving an AS-ATN ASD diagnosis, we found that most participants were receiving <25 hours per week of therapy and approximately one-third of participants were receiving behaviorally based therapies. As previously discussed, the emerging consensus is for young children with ASD to receive at least 25 hours per week of intervention.^3,4  The low rate of behaviorally based therapy was noteworthy and is consistent with findings in previous multistate studies suggesting that only ∼29% to 64% of children and 14% of preschoolers with ASD receive behavioral therapy.^7,8,11,13,21  Furthermore, the low number of treatment hours per week is concerning given the strong evidence of a dose-response relationship for intensive behavioral interventions in language and adaptive skills.²²  Recognizing that behavioral therapies are a scarce resource in some communities, these findings highlight the need for more research to develop innovative models of delivering evidence-based therapies to the broader population of children with ASD.

Through univariate analysis, we found variations in therapy intensity to be associated with age at follow-up, previous diagnosis, AS-ATN site region, and insurance type. Of these variables, only previous diagnosis remained significant in a multiple regression model. This finding that children with a previous diagnosis were receiving more hours per week of therapy than those without a previous diagnosis could be related to the amount of time it takes for families to access therapies. We chose to analyze data from follow-up visits >6 months after AS-ATN diagnosis to account for time needed for families to access recommended therapies. However, this process may take >6 months, as described by Piccininni et al²³  in Canada (mean waiting time for intensive behavioral intervention was 2.7 years). Children with an ASD diagnosis before enrollment may have had more time to establish interventions compared with those who were first diagnosed at enrollment. However, the multivariable model explained <10% of the variance in therapy hours received. These variables are not strong predictors of therapy hours, and further research is needed to clarify factors that influence therapy intensity for young children with ASD.

Estimates of the frequency of psychotropic medication use for children with ASD vary greatly. The finding that 16.3% of children used psychotropic medication is lower than the 32% of 3- to 5-year-old Medicaid-enrolled participants with ASD reported previously.¹⁷  However, psychotropic medication use in this young cohort is higherthan the 10% reported by Coury et al¹⁹  at the time of enrollment in the AS-ATN registry and higher than the 11% reported for 3- to 5-year-olds enrolled in a national database.¹⁸ 

The most commonly prescribed class of psychotropic medications for this young cohort were α-agonists. In one randomized controlled study, α-agonists were effective in decreasing hyperactivity in children with ASD, but most of the children were older than those in this study.²⁴  Although long-acting α-agonists are approved by the Food and Drug Administration for treatment of ADHD in children >6 years old, α-agonists are not yet approved for young children or for children with ASD. Use of α-agonists in 7.0% of our sample suggests the need to study these medications in younger children with ASD and to conduct comparative effectiveness studies with stimulant medications, which have been shown to reduce ADHD symptoms in children with ASD.^25–27 

In the multiple logistic regression model, psychotropic medication use was associated with site region. Participants in Canada were the least likely to be taking psychotropic medication, perhaps reflecting differences in the health care system there. Within the United States, fewer participants in the Northeast and West reported psychotropic medication use compared with those in the Midwest and South, which is similar to the pattern reported by Rosenberg et al¹⁸  in a national sample of children with ASD. Further research is needed to clarify the reasons for such geographic variation in medication use, which could include differences in state and county health care policies, local characteristics (like urban density), and access to nonpharmacologic therapies and subspecialty health care providers.¹⁷ 

Our study uses data from a large, multisite network registry, and all participants have rigorously confirmed ASD diagnoses as well as detailed clinical assessments. Despite these strengths, this study has several limitations. The participants in the AS-ATN are self-referred, must speak English or Spanish, have agreed to participate in a research study at an academic health center, and are predominantly white. This limits this study’s generalizability to other populations. Additionally, participants were enrolled in the registry between December 2007 and December 2013, and access to services and prescribing trends may have changed. It is also important to note that the subset of participants who had follow-up data available for analysis in this study differed significantly from the larger AS-ATN registry population in this age range on several sociodemographic characteristics. These biases in our population toward a subset of highly motivated participants might bias our results in the direction of increased therapy use (eg, parents who follow-up may be more likely to follow treatment recommendations and advocate strongly for services). If so, our results suggest that even in this best-case–scenario group of highly motivated participants, actual use of therapy is lower than expected.

Additionally, our analysis is limited by parental report of services. Although the questionnaire completed by parents asked about multiple therapy types, it did not explicitly ask about where the services were provided or the time spent in preschool or early intervention–based classrooms. Thus, it is possible that some children did receive additional hours of intervention in such a setting. However, even if the participants spent an additional 13.5 hours per week in a special education classroom (as was found in a national sample of 3–5-year-old children with ASD⁹ ), the total hours per week of active intervention for many children with ASD still falls short of the 25 hours per week recommended by guidelines.^3,4  Furthermore, even if school-based hours are undercounted, attendance in a school-based setting does not necessarily imply the use of evidence-based intervention practices for the treatment of ASD. For example, a study of Georgia public schools demonstrated that <10% of teachers reported using any evidence-based strategies for children with ASD.²⁸  Future studies should include exploring the total time that a child is engaged in any type of direct intervention, including autism-specific school services and private therapies.

Despite some limitations, our findings have several important implications. The total number of hours per week of therapy for young children with ASD is lower than what is recommended to promote optimal outcomes, and a minority of children receive behaviorally based therapies. Efforts are needed to improve access to these interventions. Furthermore, medications are widely used, even in young children, and the most commonly used medications need to be better studied in this population.

We thank Avi Ziskind, PhD, for his assistance with programming and statistical analysis; Paige Fischer, MS, for her assistance with programming; and David Mandell, ScD, of the University of Pennsylvania Perelman School of Medicine for his assistance with the conceptualization of this study.

ABA

applied behavioral analysis

ADHD

attention-deficit/hyperactivity disorder

ASD

autism spectrum disorder

AS-ATN

Autism Speaks Autism Treatment Network