Neighborhood characteristics including housing status can profoundly influence health. Recently, increasing attention has been paid to present-day impacts of "redlining," or historic area classifications that indicated less desirable (redlined) areas subject to decreased investment. Scholarship of redlining and health is emerging; limited guidance exists regarding optimal approaches to measuring historic redlining in studies of present-day health outcomes. We evaluated how different redlining approaches (map alignment methods) influence associations between redlining and health outcomes. We first identified 11 existing redlining map alignment methods and their 37 logical extensions, then merged these 48 map alignment methods with census tract life expectancy data to construct 9696 linear models of each method and life expectancy for all 202 redlined cities. We evaluated each model's statistical significance and R2 values and compared changes between historical and contemporary geographies and populations using Root Mean Squared Error (RMSE). RMSE peaked with a normal distribution at 0.175, indicating persistent difference between historical and contemporary geographies and populations. Continuous methods with low thresholds provided higher neighborhood coverage. Weighting methods had more significant associations, while high threshold methods had higher R2 values. In light of these findings, we recommend continuous methods that consider contemporary population distributions and mapping overlap for studies of redlining and health. We developed an R application {holcmapr} to enable map alignment method comparison and easier method selection.
Keywords: Geospatial methods; Health equity; Population health; Redlining.
© 2024. The New York Academy of Medicine.