Case Identification and Characterization of Autistic Young Adults in 2010 Medicare Fee-for-Service Claims.

Abstract

Medicare is a public insurer for whom many autistic adults are eligible in the United States, but little is known about autistic beneficiaries who are covered. A challenge in using claim data is identification of autism spectrum disorder (ASD) cases to ensure accurate characterization. Some work suggests that relying on one claim could identify probable ASD, although other works indicate that two claims are necessary for case identification. The purpose of the current study was to describe the sample of Medicare young adult beneficiaries, and determine whether using a 1+ versus 2+ claim case identification resulted in similar interpretation of sample demographic characteristics and primary care utilization patterns in Medicare professional service claims. We used Medicare Limited Data Sets (2008-2010) claims. After ASD case identification using ICD-9-CM (299.xx), 527 unique beneficiaries in the last claim year of 2010 professional service file were identified as having at least one claim of ASD. Of these, 69% (n = 364) had two or more claims. Proportions and zero-inflated negative binomial regression were used to examine differences in demographic characteristics and primary care utilization and costs for the 1+ and 2+ samples. Medicare claims contain a sample of autistic adults with expected demographics identified in historic prevalence cohorts. No differences in age, gender, race/ethnicity, Hispanic status, or dual-eligibility months or Adjusted Clinical Groups (ACG)® concurrent risk scores were identified between the 1+ and 2+ samples. No difference was found in the overall estimation of primary care use or costs between the 1+ and 2+ samples based on Zellner's seemingly unrelated regression methods. This study is the first to describe a national sample of Medicare-insured autistic adults. We found that using a 1+ case identification results in a sample that is demographically similar to a 2+ claim sample, and produces similar estimates of utilization as a 2+ claim sample.