Presentation (CTI)

OBJECTIVES: Obtaining individual-level health and biomarker data for small areas are challenging to obtain. For example, population examinations and surveys are expensive to produce at small areas and claims data are utilization-biased, often lack demographics, and may not capture low-income and rural populations. We generated and validated individual-level datasets containing biomarker data and including uninsured and undiagnosed individuals using a synthetic population reconstruction approach.
METHODS: Using data from the 2022 U.S. Census American Community Survey and three combined waves of the National Health and Nutrition Examination Survey (NHANES; 2013-2018), we generated populations representative of all U.S. counties, states, and Washington, DC. We conducted external validation for state-level prevalence estimates of dyslipidemia, hypertension, and asthma using two one-sided t tests to assess statistical equivalence using an equivalence margin of ±0.05 between synthetic population estimates and state-level prevalence estimates from the Behavioral Risk Factor Surveillance System (BRFSS).
RESULTS: Statistically equivalent estimates were observed for 40 (78%) of 51 tests for asthma, 39 (76%) for hypertension, and 17 (33%) for dyslipidemia. Mean percent differences between synthetic and BRFSS estimates were 2.6% for asthma, 2.0% for hypertension, and 4.7% for dyslipidemia. Validation against BRFSS demonstrated strong agreement for asthma and hypertension across most states but lower equivalence for dyslipidemia.
CONCLUSIONS: This approach addresses a critical gap not filled existing methods and data. By incorporating biomarker data, sociodemographic characteristics, and uninsured populations into an individual-level dataset for any U.S. state, these estimates provide a more comprehensive view of disease burden. Discrepancies observed in validation likely reflect differences in survey design, as evidenced by differences between NHANES and BRFSS within national-level estimates, or the inability to incorporate spatially varying factors. External validation of estimates for county-level data is currently underway. Policymakers can use these data to better assess prevalence and design interventions that reach populations typically excluded from claims-based surveillance.