We used data from the first wave of the National Longitudinal Transition Study-2 (NLTS2), a nationally representative study of adolescents in special education. A two stage sampling procedure was used and estimates were weighted to generalize to all students age 13–16, in special education in 2000. Further details of the study sample design and weighting have been previously reported (9
Data were collected in 2001 through telephone interviews with parents or guardians of youths age 13–17 administered in English or Spanish. Parents who were not reached by telephone were mailed a shortened self-administered questionnaire. Parents or guardians of 920 youths in the special education autism enrollment category responded, a response rate of 83.5%. Use of these data is governed by a data use agreement with the U.S. Department of Education and was approved by the University Institutional Review Board. All unweighted sample size numbers were rounded to the nearest ten as required by the data use agreement.
Youths in the sample were selected based on classification into the special education reporting category of autism which does not require a DSM-IV diagnosis of autism, hence we do not have information about the specific type of ASD (i.e. Aspergers, Autistic Disorder or PDD-NOS). Epidemiological surveillance data have found that 99% of children served under the autism educational designation also meet DSM-IV criteria for an ASD (10
). However, some youths who meet diagnostic criteria for an ASD may be served under another eligibility category and would not be included in this analysis.
Mental health service use was assessed by asking, “During the past 12 months, has (youth) received any psychological or mental health services or counseling?” Respondents that answered affirmatively were asked a follow-up question to determine whether the services had been through the school. The data did not allow us to determine if youths receiving services in school were also receiving services outside of school. We included gender, race, ethnicity, and parental education as predisposing variables since these often serve as proxies for beliefs about mental health treatments. Language impairment was also considered a predisposing characteristic as lack of speaking ability may limit the perceived appropriateness of mental health services. An indicator for severe language impairment was created for youths who have a lot of trouble speaking clearly or don’t speak at all.
Enabling resources included income, health insurance status, case management, having a diagnostic medical evaluation in the past year, and parent and youth involvement in individualized education planning (IEP) meetings. A sequence of questions about insurance status was recoded into private and government/other insurance for analysis. Two dichotomous indicators asked about parent and youth attendance at the most recent IEP meeting.
Measures of need included parent-reported co-morbid Attention Deficit Hyperactivity Disorder (ADHD), social skills, and experiences of bullying. We included ADHD because it is a common co-morbidity among youths with an ASD (2
). Unfortunately, the survey did not directly ask parents about other types of comorbidities. Social skills were measured using eleven items drawn from the Social Skills Rating System (SSRS) Parent Form (11
), with higher scale scores indicating greater skill. Bullying victimization was measured by collapsing three questions about whether youth had been bullied, teased, or physically attacked at school. Another question asked whether the youth had bullied others.
One logistic regression model examined use of mental health services among all youths with an ASD. A second logistic regression model examined the use of school-based mental health services among the subset of youths that had received any mental health services. Twenty multiply imputed data sets were created using sequential regression in IVEware to handle missing data (12
). All estimates are population estimates. Analyses were weighted and variances were adjusted to account for the sampling design and imputation using Stata 11.