Heterogeneity of the impact of mental health parity
*Frank B. Yoon, Mathematica Policy Research, Inc. 
Haiden A. Huskamp, Harvard Medical School 
Alisa B. Busch, Harvard Medical School 
Sharon-Lisa T. Normand, Harvard Medical School 

Keywords: Joint models; matching; mental health parity; multiple outcomes; observational study

Policy interventions in health insurance typically target broad populations at the national or state level, but effects of such interventions may differ across smaller clinical subpopulations. For example, parity in mental health coverage aims to protect an enrollee from catastrophic losses due to illness; these protective benefits may manifest themselves in greater magnitude in those enrollees with severe illnesses such as bipolar disorder than those with less severe but acute illnesses such as adjustment disorder. In an observational study of the Federal Employees Health Benefits (FEHB) Program we explore the heterogeneity of parity’s impact in the severely ill compared to its impact in the less severely ill and illustrate new techniques for matched designs and analysis of multiple outcomes. Matching with fine balance ensures that the matched samples have the same joint distribution of important categorical covariates while other covariates are stochastically balanced. Multiple outcomes characterizing mental health service usage and intensity are analyzed with joint difference-in-difference models for zero-inflated counts that account for the correlations among different service types and repeated measures. The joint models provide a global test of parity’s overall effect on usage of mental health services and a test on the interaction term that assesses whether parity had a heterogeneous effect among the clinical subpopulations. We illustrate estimation of these models through likelihood and quasi-likelihood methods with available software. In our application we find that parity affects mental health service utilization to a larger extent for those with adjustment disorder, who are less severely ill than those with bipolar disorder; specifically, for those with adjustment disorder, outpatient services decreased significantly with parity’s implementation, a likely product of managed care.