Frederick H. Hooven, MMHS, Jianying Zhang, MPH, Alexis D. Henry, ScD, OTR/L, and Jay Himmelstein, MD MPH. Center for Health Policy and Research, University of Massachusetts Medical School, 222 Maple Ave, Shrewsbury, MA 01545, 508-856-8221, jianying.zhang@umassmed.edu

Medicaid buy-in programs allow people with disabilities to return to work without fear of losing Medicaid benefits. Evaluations of these programs rely heavily on Medicaid claims data, but these databases usually lack a variable describing a person’s disabling condition. Medicaid evaluations based on claims are therefore rarely able to account for the effect of specific disabilities on program outcomes. We assessed the accuracy of a claims-based method (Claims) for identifying people with either physical or mental health related disabilities, by first comparing the classification to the disability assigned by the Massachusetts Disability Evaluation Services (DES), and second, to the disabling condition reported in a survey of Medicaid members with disabilities. Claims disagreed with the DES classification 25% of the time for mental illness, and 37% for physical disability. When DES was the classification standard, the sensitivity for the Claims classification of mental illness was 83%, and the specificity was 69%. For physical disability, the sensitivity and specificity were 60% and 65%, respectively. Using self–report of disability as the standard, we found lower sensitivity (76%) but higher specificity (81%) for mental illness classification, and very low sensitivity (22%), but higher specificity (70%) for classification of physical disability. This study provides important information for quality assessment and program evaluation efforts which use claims data to evaluate Medicaid programs for people with disabilities. Caution should be employed when Medicaid data is used to assess the relationship between people’s disabling conditions and outcomes such as cost, service use, and employment.

Learning Objectives: