Examining the scope of homicide-suicide perpetrated by law enforcement: A population-based retrospective study using multiple frames, 2003 – 2017

Background: Domestic violence perpetrated by law enforcement is an understudied public health issue. Previous studies examining the prevalence domestic violence by law enforcement primarily rely on convenience sampling and self-reported measures, producing unreliable estimates. Homicide-suicide events, as a subset of domestic violence offenses, present a unique opportunity for study given the data collection practices of existing public health surveillance infrastructure. The present study contributes to the literature by using population-based data and vital statistics to estimate a single subset of domestic violence offenses: homicide-suicide perpetration.
Methods: We joined homicide-suicide perpetration data from the National Violent Death Reporting System (NVDRS) with population data from the American Community Survey for the years 2003 – 2017, analyzing the joined data as a multiple-frame sample. Our exposure of interest was occupation type, classified as employment in (a) law enforcement, (b) military occupations, (c) other life-threateningly dangerous or traumatic occupations. Our unexposed reference group was persons in all other general occupations. To address issues of exposure misclassification, we applied conservative and anti-conservative classification methods. Our outcome of interest was perpetration of single-suspect homicide-suicide. Primary analyses used robust Poisson regression to calculate incidence ratios and 95% confidence intervals. Secondary analyses described years-of-life-lost (YLLs) to homicide victimization by homicide-suicide perpetrators, stratified by perpetrator occupation.
Results: Compared to the incidence of homicide-suicide perpetration among persons employed in all other general occupations, law enforcement (aIRR = 1.6; 95% CI: 1.2 – 2.1), military personnel (aIRR = 3.7, 95% CI: 2.7 – 5.0), and persons employed other life-threateningly dangerous or traumatic occupations (aIRR = 2.9, 95% CI: 2.1 – 4.0) were at an elevated risk for homicide-suicide perpetration when exposure was classified conservatively. Using anticonservative exposure classification yielded higher point estimates with largely overlapping confidence intervals. Altogether, homicide-suicide perpetrators from law enforcement, military, and other life-threateningly dangerous occupations were responsible for approximately 6 – 7% of the total homicide-suicide victim count and 6 – 7% of the total victim YLLs.
Conclusions: Law enforcement officers commit homicide-suicide at a higher rate than the generally employed population, as do military personnel and persons employed in other life-threateningly dangerous or traumatic occupations. NVDRS and ACS data can be joined and analyzed as a multiple-frame sample for the study of rare health outcomes.