Evaluation of the Independence at Home Demonstration
Under Section 3024 of the Patient Protection and the Affordable Care Act (PL 111-148), the U.S. Congress mandated that the Centers for Medicare & Medicaid Services conduct a demonstration designed to test the ability of the home-based primary care delivery model to reduce health care costs and improve health outcomes of high-need Medicare beneficiaries. The 18 practices chosen to participate in the three-year (2012–2015) Independence at Home (IAH) Demonstration are expected to design and implement coordinated care plans tailored to individual beneficiaries’ chronic conditions and responsive to their preferences, make in-home primary care visits to these patients, and be available 24 hours per day to meet their health needs. The goal of the demonstration is to improve the provision of comprehensive, coordinated, continuous, and accessible care to chronically ill, disabled beneficiaries. This evaluation is determining whether—and, if so, how—sites achieve success.
Mathematica is using a mixed-methods design that centers on a set of three key research questions:
- What is the effect of the demonstration on processes of care? Our implementation analysis is using qualitative techniques and descriptive methods to assess practices’ actual experience and the factors that helped or hindered their efforts to improve outcomes. Much of the data will be obtained from three rounds of site visits to the 18 demonstration practices and telephone interviews with key staff at those practices.
- What is the effect of the demonstration on outcomes of interest—health status, utilization of services, costs and savings, and enrollees and their caregiver’s experiences with the program? Our mixed methods approach includes difference-in-difference regression analyses with an appropriate comparison group to control for other simultaneous changes that might be fully or partially responsible for any observed improvement in key outcomes. Using propensity score matching, we are matching samples of Medicare beneficiaries to create a virtual comparison group for each of the 18 demonstration practices. The results provide an estimate of what the costs and utilization of demonstration enrollees would have been if they had not received primary care in the home. To measure outcomes of interest that are not captured in administrative data, we have developed a beneficiary/family and caregiver survey. The survey uses a mail with telephone follow-up design. We are collecting data from 7,500 beneficiaries and their caregivers. We will conduct a descriptive analysis of these data and use cross-sectional regression analysis to compare treatment group measures with a national benchmark constructed from the Medicare Current Beneficiary Survey.
- What features of the demonstration interventions are the most important predictors of study outcomes of interest? We will integrate findings from the site visits with those from the impact analysis to determine features that are the most important predictors.
Jesse (Jay) Crosson