Online Program

Designing a multilevel evaluation of health care systems: the Medicaid Emergency Psychiatric Services Demonstration

Priyanka Anand, Mathematica Policy Research 
Melissa Azur, Mathematica Policy Research 
Crystal Blyler, Mathematica Policy Research 
*Frank Yoon, Mathematica Policy Research 

Keywords: hierarchical models, multiple control groups, systems of care, psychiatry, mental health

From the inception of Medicaid, institutions for mental diseases (IMDs), such as hospitals, nursing facilities or other institutions of more than 16 beds that primarily provide care to individuals with mental disorders, have been excluded from receipt of federal matching Medicaid funds for inpatient care of adult beneficiaries. This Medicaid IMD exclusion applies even in the case of beneficiaries with acute psychiatric needs, such as those expressing suicidal or homicidal thoughts or gestures. The lack of federal reimbursement in such cases may contribute to critical service gaps and inefficiencies, which may in turn result in diminished outcomes for low income adults with severe mental illness and overall increases in public expenditures for their health, mental health, and social services.

We will present our plan for evaluating a CMS national demonstration authorized by the Patient Protection and Affordable Care Act of 2010 that waives the Medicaid IMD exclusion policy. Multiple administrative databases, such as CMS and state claims data bases, present an opportunity to apply hierarchical longitudinal models to study the demonstration’s impact on patients, facilities, and systems of care. A design using multiple control groups that are defined, for example, by geography or state-specific policy, enhances a multifaceted approach for analyzing extremely complicated, highly confounded data. We will focus on addressing biases due to unobserved differences in case-mix, such as those induced by self-selection into treatment facilities or by geographic variation.