The Housing and Health Connection from the Data Side
Katherine O’Regan, Assistant Secretary for Policy Development and Research. Earlier this summer, the Urban Institute and the MacArthur Foundation hosted “How Housing Matters for Healthy Child Development,” a roundtable that brought together researchers and practitioners from the fields of housing, community development, and from health to discuss the connections between housing and healthy outcomes for children. The two-day event revealed that those in the health field have long recognized the impact of home and environment on health outcomes. This insight, however, is relatively new to those in the housing field. In my remarks, I proposed that for many housing researchers, interest in the link between health and housing emerged following the results of the Moving to Opportunity (MTO) demonstration, a program that used vouchers to help poor families living in public housing in high-poverty neighborhoods move to lower-poverty neighborhoods. A key finding of MTO was of significant impacts on health — in particular, mental health. Further exploration suggested that MTO’s effects on children’s health outcomes may have varied according to the baseline health vulnerability of their families, as explored in the Office of Policy Development and Research’s (PD&R’s) Expert Convening on Gender, Neighborhood Context, and Youth Development.
The MTO findings are of great importance for at least two reasons. First, one of the largest (or earliest) impacts of providing quality, affordable housing or better neighborhood environments for families may be on their health. Historically, however, housing researchers have neither tracked health outcomes nor included them in evaluations. Second, health itself may actually be a mediating mechanism, meaning that the effects that affordable housing and neighborhood development programs have on other life outcomes may depend on health status or occur in part because of health effects. Recognizing these connections is important for evaluating the effectiveness of our programs and for better targeting our interventions.
These insights have broadened the scope of research in the housing and community development fields, including here at HUD. In fact, almost every large demonstration or evaluation funded through HUD over the last several years is exploring opportunities to observe health outcomes even though health was not a focus when the study began. We have also recently funded or partnered in three research projects where health is a key focus and we consider this nexus of health and housing to be one of our priorities for future work. There is a particular challenge, however, to our progress: our programmatic scope is housing and community development, and this is reflected in the data we collect. Even as we engage in more comprehensive, cross-sector programmatic work, our administrative data remains ‘siloed’, reflecting the general nature of federal agencies. We need better ways of capturing other domains, particularly health, in our data, and we need to do so more frequently and more broadly than we currently do through specific research projects.
PD&R is working on that now. Three main efforts are underway at HUD that match the administrative data on HUD-assisted renters with their health-related data. These efforts are low-cost initiatives that will greatly enhance our knowledge about health status and the healthcare use (and costs) of assisted households, including the possible cost savings of various policies.
In our first such effort, PD&R collaborated with the Office of the Assistant Secretary for Planning and Evaluation (ASPE) at the U.S. Department of Health and Human Services (HHS) and the Lewin Group to test the feasibility of matching our data to Medicare and Medicaid claims data. We successfully matched and analyzed health cost data in a sample of 12 geographic areas, as outlined in the full report available on ASPE’s web site (http://aspe.hhs.gov/daltcp/reports/2014/HUDpic.pdf). In the pilot study, we found that Medicare beneficiaries in HUD-assisted housing tend to use more health services than do individuals in the surrounding communities and that Medicare-Medicaid dual beneficiaries in HUD-assisted housing had more chronic conditions and higher health care utilization than did Medicare-Medicaid dual enrollees in the surrounding communities. We are in the process of analyzing these data further and partnering with HHS to consider other opportunities for matching data to improve both housing and health outcomes for HUD-assisted tenants.
In our second effort, which is also a collaboration with ASPE, we are evaluating Vermont’s SASH program (Support and Services at Home). SASH is a model that aims to help frail residents residing in federally assisted housing and in the community age in place. The evaluation employs a comprehensive strategy to match health care and housing administrative data, and is providing encouraging results in terms of reduced rates of growth in Medicare expenditures among the SASH program participants.
PD&R is also collaborating with the National Center for Health Statistics (NCHS) at HHS to link HUD data to two major national health surveys: the National Health Interview Survey (NHIS) and the National Health and Nutrition Examination Survey. These national surveys provide timely information on a range of health indicators, including health status and access to health care.
Although HUD has geocoded NCHS survey data for many years, the agency began to explore the possibility of matching NCHS survey data with HUD-assisted housing administrative data only 2 years ago. Under a Memorandum of Understanding, HUD and HHS researchers explored the feasibility of using data matched from the NHIS’ representative sample survey of 40,000 households to analyze selected measures of health and health care utilization among HUD-assisted households compared with other households. The data matching has provided an opportunity for stronger interagency partnerships, as HUD and HHS staff have worked to improve the data matching process, resolve challenges related to maintaining confidentiality, and understand the initial matched survey results. Over the next year, HUD will release a descriptive report of both the process and preliminary findings, and the two agencies are in the early stages of identifying priority research questions that they can jointly investigate.
We are joining the efforts of numerous others who are working to link administrative data to make them more useful for policymaking and research. Many of these efforts encompass housing and health outcomes. For example, the Actionable Intelligence for Social Policy Initiative, led by Dennis Culhane at the University of Pennsylvania, is providing several local jurisdictions with technical expertise to create integrated data systems. By linking data from different county- and state-level social service providers with housing data from HUD, this MacArthur Foundation-funded initiative will give local leaders the data needed for better decisionmaking. These data enable better integration of services to meet families’ and children’s needs and allow a better assessment of how well different approaches are working. As another example, the National Neighborhood Indicators Partnership (NNIP), a collaboration of the Urban Institute and local partners in 35 cities, furthers the development and use of neighborhood-level information systems for community building and local decisionmaking. Health and housing are featured as two of the partnership’s focus areas.
The process of linking administrative datasets comes with many challenges, and I do not mean to suggest that it is easy or always appropriate. Our own efforts will require both significant oversight to protect privacy and the development of new protocols to move forward. As HUD’s programmatic work increasingly recognizes the connections between the housing and health domains, we need new data tools to address this. Linking administrative data — data already collected and cleaned as part of existing government programs — is clearly one of the least expensive ways to accomplish these goals. In an environment of limited public resources, making the most out of the data we already collect must become part of our toolkit.