Association for Public Policy Analysis & Management

Homeless individuals suffer from much higher than average rates of both morbidity and mortality. Caring for homeless people places substantial burdens on safety net providers, including Medicaid payers, and it has been argued that using some Medicaid funds on supportive housing for medically needy homeless individuals will improve their health and lower total Medicaid costs by increasing the quality and appropriateness of the care provided.

While some studies suggest that moving homeless individuals to housing can reduce health care costs and inappropriate utilization, few studies have included an appropriate comparison group. This can lead to overestimating the impacts of supportive housing, since high medical utilization can “regress to the mean” over time even in the absence of a program intervention.

In this study, we present the central findings from an evaluation of the Corporation for Supportive Housing’s Social Innovation Fund intervention, funded by the Corporation for Neighborhood and Community Services and several foundations.  Four sites were funded: San Francisco, Los Angeles, a consortium of Connecticut cities, and Washtenaw County, Michigan.  In three of the sites, the evaluation randomly assigned homeless high utilizers to receive supportive housing combined with staff focused on helping clients obtain needed medical care, or to receive “usual care,” in order to test whether the intervention significantly reduced health care costs.

Data-driven approaches were used to identify the target population.   Approximately 1,000 individuals were randomized to either the program or control groups in three of the four sites, and about 500 individuals were housed across all sites. The sites varied in regard to the type of housing provided (congregate vs. scattered site), and in the staffing of their support services.  Randomization to the treatment condition ensured that there were no systematic difference between treatment and control groups, making this one of the most rigorous tests to date of the potential for supportive housing to reduce medical costs and improve health care quality for homeless, high cost users of the health care system. Health care utilization and cost data were collected for the year prior to and the year following random assignment, and housed participants were surveyed at baseline and one year after being housed. In addition, regular site visits were made to each of the sites.

In this paper we will use an intent to treat approach with difference-in-differences analyses to estimate impacts of the intervention.  Analyses to date indicate that each site was highly effective in targeting and engaging clients with substantial health care utilization at baseline.  Further, preliminary analyses indicate that costs are reduced in the year after assignment but reductions are found among the control groups, as well.  In this paper, we will carefully examine variations across sites and, most especially, across types of health care conditions.  We hypothesize that many homeless individuals face complex medical conditions that will result in increased costs when treated appropriately once housed. Such increased costs may offset savings from those for whom housing and case management can reduce costs while increasing health status.