The Individuals with Disabilities Education Act (IDEA) of 1990 required that schools provide a free and appropriate education to qualified children with autism. It also requires that the education system provide assistance to these children even during preschool years to ready them for school. In the last decade, the number of children identified with autism in the special education system has risen dramatically from 54,064 to 258,305, an average increase of 17% a year (www.ideadata.org). While little data are available on the educational costs associated with these children, a report commissioned by the Government Accountability Office found that the average annual expenditure during the 1999–2000 academic year for children with autism was $18,000, almost three times that of children not in special education, and the second highest average expenditure among all special education categories (Chambers et al. 2004). These services often include specialized classrooms, one-on-one instruction, intensive behavioral interventions (Lord et al. 1989), and speech, language and occupational therapy, for children to receive an education in the least restrictive environment appropriate to their individual needs (Koppelman 2004; Noel and Shreve 2006; Parrish et al. 2004; Rodman et al. 1999).

Because many of these services can be construed as medical, states often turn to Medicaid to help offset the economic burden associated with this care. In fact, some have argued that Congress’s original intent with both Medicaid and IDEA legislation was for Medicaid to cover these medically necessary expenses (Mandell et al. 2008a; Noel and Shreve 2006). States have incentive to rely on Medicaid rather than IDEA funds, because the federal match for IDEA services is 17–19%, compared with between 50 and 75% for Medicaid services. The Center for Medicaid Services, however, has clarified that they consider themselves the payer of last resort for many of these services, especially behavioral therapies provided in educational settings, creating mounting tension and confusion regarding who should pay for care (Shattuck and Grosse 2007).

States have set various policies to serve children with ASD through their Medicaid programs. Most have expanded eligibility through waiver services for children with developmental disabilities and arrangements between Medicaid and special education (Parrish et al. 1998, 2004; Stahmer and Mandell 2007; Thurlow et al. 2002). It is unclear, however, whether autism services offered through state Medicaid programs are a complement to the resources devoted to autism services delivered through the education system or whether they act as a substitute in places with fewer education resources.

In the adjusted analysis, some education variables continued to be associated with Medicaid ASD prevalence. Counties with greater per-student education expenditures had a lower Medicaid ASD prevalence (11.2% lower in the 4th quartile than in the 1st quartile). While not every quartile comparison was statistically significant at p < 0.05 for these two variables, counties with more students ≤12th grade and those with a higher proportion of students receiving special education services also had a higher Medicaid ASD prevalence.

Variables measuring healthcare resources maintained significant associations with Medicaid autism prevalence in the adjusted analysis. Counties with higher per capita number of pediatricians (15.1% higher in counties with “high” vs. “none”) and pediatric specialists (8.9% higher in counties with “any” vs. “none”) had greater Medicaid ASD prevalence.

Medicaid-enrolled ASD prevalence varied greatly as a function of other county characteristics. Even after adjusting for poverty, the proportion of Medicaid-enrolled children in the county was associated with prevalence. While states have considerable latitude in how they use Medicaid to fund healthcare services for children, our analysis included state as a fixed effect, meaning that our results represent variation within states. The observed association between Medicaid enrollment and ASD prevalence may be due to counties taking differential advantage of their state’s Medicaid policies, perhaps because of local knowledge or the availability of education or other resources unmeasured in this study. Alternatively, families may change county of residence to be closer to better healthcare resources. The finding that counties with more pediatricians and pediatric specialists per capita enrolled a greater proportion of children with ASD in Medicaid may be due to families moving to be closer to these resources or a higher prevalence of pediatric specialists resulting in more diagnosing.

Counties with a greater proportion of white residents had higher Medicaid-enrolled ASD prevalence, which is in line with previous findings about racial and ethnic disparities in the diagnosis of autism (Mandell et al. 2002, 2006, 2009). Membership in a traditionally underserved minority group may play a more important role in accessing services than income, which was not associated with ASD prevalence. While there is no known difference in the population prevalence of ASD among ethnic groups, one study found that African American and Latino children meeting research criteria for ASD were much less likely than white children to be identified as such in their healthcare or education records (Mandell et al. 2009). Researchers have hypothesized that physicians may be less likely to diagnose ASD among ethnic minorities; there may also be more stigma or other barriers associated with ASD and accessing publicly funded services among ethnic minority parents.

Interpretation of study findings are limited by a number of factors, primary among them that the ASD diagnosis in the Medicaid claims has not been validated. While its accuracy has not been specifically examined, Fombonne et al. (2004) found 97% positive predictive value for chart diagnoses and a diagnosis of ASD administered by a trained research team, and Yeargin-Allsopp et al. (2003) found that 98% of children with a chart diagnosis met research criteria for ASD A second limitation is the absence of other county variables (e.g., ASD-specific intervention resources, county-level administrative prevalence of autism in the special education system) that may relate to identification in the Medicaid system.

Despite these limitations, there are important implications related to these findings. Within states, counties appear to differ in how they take advantage of Medicaid policies. The impact of these different arrangements between Medicaid and education on the coordination, intensity and quality of care for children with autism has yet to be determined. When Medicaid and education resources both are tapped to provide services for children with ASD, it can increase the potential volume of care and range of services these children receive. On the other hand, a large body of research suggests that when multiple systems are responsible for providing related care to the same child, it can result in considerable gaps and inefficiencies in care (American Academy of Pediatrics 2002; Dickens et al. 1992; US Department of Health and Human Services 1999).

Our findings also provide further evidence for the need for outreach to minority groups and clinicians working with them to improve recognition of the developmental differences associated with ASD. Our focus on the county—rather than the individual—as the unit of analysis suggests that outreach might be geographically as well as ethnically targeted.