Using Small Area Estimation at a Census Tract Level to Create an Index of Economic, Behavioral Health, and Public Health Conditions in King County, Washington

BACKGROUND: King County, Washington is the 13^th most populous county in the United States, home to 2.1 million residents. At a county level, King County tends to have better health outcomes and lower risk factors as compared to Washington State or the rest of the US. Previous work on health disparities below the county level showed marked disparities between income, race/ethnicity, and geographies that rival some of the worst disparities seen in US. Public Health, Seattle & King County (PHSKC) set out to find if there were particular geographies that tended to have worse mental, economic, and health outcomes, with the intent to leverage the information to form partnerships with communities and leaders to improve outcomes.

METHODS: Using a cross-disciplinary group, with representatives from the Department of Community and Health Services and PHSKC, over 60 indicators were discussed based on availability (geographic, temporal, trend), applicability, and relevance to on-going work. The indicators spanned vital records (birth, death, hospitalization), economic (American Community Survey household income; unemployment), quality of housing, and behavioral risk factors and health outcome status from the Behavioral Risk Factor Surveillance System (BRFSS). Since 2005, King County has asked a nearest intersection for BRFSS respondents, and has geocoded the data to a census tract level. However, even combining all years only produced a robust sample size for about 60% of tracts. In order to produce the data at a granular level, PHSKC needed to implement small area estimates (SAE). ZIP-based data used the HUD tract to ZIP correspondence file to assign tract.

RESULTS: After looking at several SAE methods, including an unweighted mixed effect model and the resulting validation, correlation, and summary measures, PHSKC used the mixed effect model. Current efforts are examining hiearchical Bayesian models for comparison. The committee settled on 10 indicators, which were ranked, rank order summed, and mapped by deciles for individual indicators as well as the index.

CONCLUSIONS: Almost regardless of indicator, a clear pattern started to emerge, and a targeted geography was selected. Funds from an executive-sponsored transformation project with the goal of better integration of mental and public health, the Seattle Foundation, and Living Cities are currently being leveraged to provide communities in the opportunity area to submit for project funds to improve the economic, physical, and mental health of their residents.