Using Public Health Data to Monitor Killings by Law Enforcement in the United States

Injuries inflicted by police constitute an important cause of death in the United States, yet public health data collection systems routinely undercount these incidents. This dissertation presents three epidemiologic studies that measure and analyze police-related deaths. The first is a matched-pair analysis of Massachusetts data (2004-2016) comparing news media-derived databases on police-related deaths (The Guardian, Washington Post, WGBH News, and Fatal Encounters) to Massachusetts vital statistics (N = 84 decedents). Demographic data reported in all four databases were highly concordant vis-à-vis death certificates, while vital statistics misclassified the cause of death (i.e. reported as diagnostic codes other than “legal intervention”) for 24% of deaths. The second study matched 2015 US data from The Guardian to National Vital Statistics System (NVSS) mortality records. Capture-recapture analysis estimated the total number of police-related deaths in 2015 was 1,166 (95% CI: 1,153, 1,184). NVSS classified 44.9% (95% CI: 44.2%, 45.4%) of these as legal intervention, and The Guardian documented 93.1% (95% CI: 91.7%, 94.2%) of deaths. The third paper analyzed the association between rates of police-related deaths and spatial social polarization throughout the entire United States, with polarization defined as census tract concentrations of privileged and deprived residential groups and measured with the Index of Concentration at the Extremes (ICE). We identified deaths using 2015-2016 Guardian data (N = 2119). In multilevel models incorporating random intercepts for cities and states, concentrated deprivation was associated with higher rates of deaths, while concentrated privilege was associated with lower rates. For the total population, associations for ICE measures that defined privilege/deprivation using income, solely or in combination with race/ethnicity, were stronger versus ICE measures quantifying concentrations of racial/ethnic groups alone. In stratified models, rates of police-related deaths were lowest for non-Hispanic white persons in the quintile of census tracts with the greatest concentrations of non-Hispanic white persons, but were highest in the same quintile of census tracts for non-Hispanic black persons. Public health data monitoring systems can be improved to better track police-related deaths, and existing systems can monitor the number of deaths, as well as analyze inequities in death rates