Adult Medicare Beneficiaries with Autism Spectrum Disorder: Case Identification and Characterization of Young Adults in the 2010 Limited Data Set Medicare Claims

Background: Although Medicaid is the most well-known public insurer of individuals with ASD it is limited to individuals who can meet eligibility and means testing rules, and Medicaid covered services differ by state. Although less well known, persons younger than 65 are eligible for Medicare if they receive a disability determination. Approximately 25% of all beneficiaries in the 2008-2010 Medicare fee-for-service (FFS) files are younger than 65 and are eligible by virtue of their disability status.

Objectives: A challenge in using claims data is identification of ASD cases to ensure accurate characterization. In pediatric validation studies, some work suggests that relying on 1 claim could reliably identify probable ASD (e.g. Burke et al., 2014). The purpose of the current study was to compare the demographic characteristics and number of ASD cases identified using one claim versus a two claim case identification criterion in FFS Medicare files.

Methods: Claims files for Medicare Limited Data Sets for 2008-2010 were used. See Figure 1 for case identification process and exclusion criteria. After case identification, unique beneficiaries in the last claim year of 2010 were classified as having at least 1 ASD claim (1+ sample), or at least 2 ASD claims (2+ sample); intellectual disability (ICD9-CM 317.xx-319.xx) was also used to classify beneficiaries. We conducted univariate and bivariable descriptive statistics on each sample using Stata V14.1 (StataCorp, 2015). Proportions and exact binomial 95%CI for sample characteristics (age, gender, race, ethnicity, ACG® concurrent risk score, Medicaid state buy-in months) for ASD individuals with and without intellectual disability (ID) were produced and differences were examined with the two-sample test of proportions. Significance was set at p<.008 after Bonferroni correction.

Results: A total of 5,547 young adults with at least 1 ASD were identified, of whom 63% (n=3,499) had no ID claim, and 37% (n=2,048) had both an ASD and ID claim. When restricting the sample to those with a minimum of 2 claims, a total of 2,816 young adults with ASD were identified, of whom 66% (n=1,847) had no ID claim, and 34% (n=969) had both ASD and ID claims. No differences were found in proportional demographic characteristics for the 1+ versus the 2+ groups (Table 1). For both the 1+ and 2+ claim analyses, the ASD and ID group was more likely to be older, Hispanic, non-white, and have 12m of Medicaid state-buy in than the ASD-only group. Additional analyses (not presented) will be shared regarding differences in predictive models analyzing the impact of using 1 versus 2 claims for service utilization analyses.

Conclusions: Researchers, policy makers, and service providers need to be aware of the availability and types of utilization afforded by Medicare. This study is the first step in validating the identification of a national sample of publicly insured adults with autism. Implications of using the 1 versus 2 claim criteria for claims analysis using Medicare to understand service utilization will be discussed, along with comparisons of models which predict service utilization when using the 1 or 2+ criteria.