*Names in bold indicate Presenter
We utilized the data from the NY State Department of Health on all MBI participants and a matching random sample of the non-MBI Medicaid beneficiaries for the period between 2006 and 2010 for examining our hypothesis. These confidential datasets consist of longitudinal information on program enrollment and eligibility dates, socio-demographic characteristics, service utilization (e.g., inpatient/outpatient treatment and diagnostic services, prescription drugs, long-term care), health care expenditure for each claim including service dates and ICD-9 diagnosis. The analytical sample consisted of 11,220 MBI participants (treatment group) and 47,552 non-MBI Medicaid beneficiaries (control group). The unconditional comparison of healthcare expenditures revealed a mean differential of USD 600 per month between the two groups.
After conditioning the participation in MBI by a set of observed individual and location specific characteristics (using propensity score method), we continue to observe a substantial differential in healthcare utilization between treatment and control groups. In particular, the average MBI participant has 55 percent lower health care expenditure, 3.4 percent lower likelihood of utilizing emergency care and 2 percent lower prevalence of hospitalization compared to the average non-MBI Medicaid beneficiary. Further utilizing the Instrumental Variable (IV) approach we examined the impact of unobserved factors in explaining the observed differences in healthcare expenditures. In particular, we introduce a set of exclusion restrictions (instrumental variables) in our empirical model that don’t have direct impacts on health care outcome but have direct impacts on the decision to participate in Medicaid Buy-In. The set of exclusion restrictions includes the amount of monetary resources received by state through Medicaid Infrastructure Grant (MIG) in a given year, the number of vocational rehabilitation counselors in the district office in a given period of time, and the number of counselors participated in the MIG related trainings in the district office and finally, macroeconomic variables such as county monthly unemployment rate. Using this set of instruments, we observe substantial differences in health care related outcomes between treatment and control groups. Policy and methodological implications of evaluation of MBI programs will be discussed in the presentation.