Mortality Classification for Deaths With Nonfirearm Force by Police, 2012-2021.

Importance: For deaths preceded by nonfirearm force by law enforcement officers, mortality classification has potential implications for public accountability and epidemiologic surveillance. Objective: To characterize the proportion of in-custody deaths for which cause and manner of death reflected the use of nonfirearm force by police officers and to assess factors associated with mortality classification. Design, Setting, and Participants: This cross-sectional study examined national data from the Associated Press's Lethal Restraint database. The analysis included deaths that occurred in 2012 to 2021 following the use of nonfirearm force by state or local police officers and that did not occur in prisons or jails. Exposures: Death investigation system (medical examiner, coroner, or sheriff-coroner), racial and ethnic bias, and county political context (percentage of Republican presidential votes). Main Outcomes and Measures: Manner of death classification (homicide vs accident, undetermined, natural, suicide, and any manner), mention of force-related injuries or conditions in the cause-of-death statement (yes vs no), and mention of any force in the cause-of-death statement (yes vs no) were assessed using logistic regression models. Results: A total of 940 decedents (mean [SD] age, 39 [11] years; 909 men [97.0%]; 297 identified as Black [32.4%], 179 as Hispanic or Latinx [19.6%], 401 as White [43.9%], and 37 as other [4.0%] race and ethnicity) were included. Of the 940 deaths, 268 (28.5%) were classified as homicide, and 155 (16.5%) were mentioned as a force-related injury or condition and 400 (42.6%) as any force in the cause-of-death statement. In contrast, 695 cause-of-death statements (73.9%) mentioned drugs. Unadjusted results showed that homicide classification increased from 25.0% (66 of 264 deaths) during 2012-2014 to 32.2% (123 of 382 deaths) during 2018-2021. Models estimating adjusted prevalence differences showed that compared with medical examiner jurisdictions, coroners (-0.19; 95% CI, -0.31 to -0.06) and sheriff-coroners (-0.17; 95% CI, -0.28 to -0.05) were less likely to classify deaths as homicides. Model results also showed that classifications for incidents occurring in counties with the lowest percentage of Republican voters were the most likely to reflect force across all 3 manner and cause outcomes (0.17 [95% CI, 0.05-0.28] for homicide, 0.22 [95% CI, 0.06-0.38] for any mention of force, and 0.14 [95% CI, 0.023-0.26] for force-related injuries or conditions). Conclusions and Relevance: In this cross-sectional study of deaths following use of nonfirearm force by police officers, nonhomicide classifications and cause-of-death statements making no mention of force were widespread. These findings suggest that inconsistent classification of the cause and manner of deaths that follow nonfirearm force by police officers is an issue of public safety and health with profound social implications.