Research Letter
Deaths Caused by Law Enforcement in the US
Caroline Raymond-King, Molly M. Jeffery, Sirry Alang, et al
JAMA 2026 Aug 11;336(6):512-514
doi: 10.1001/jama.2026.9981
Deaths caused by law enforcement have received increasing public attention in the US. However, national surveillance of these deaths remains challenging because vital statistics data substantially undercount fatalities caused by law enforcement compared with the National Violent Death Reporting System, federal reporting, and media-based datasets.1-3 Nevertheless, mortality files from the National Center for Health Statistics (NCHS) allow for national surveillance of these deaths using the same methodology over more than 20 years, whereas other datasets do not. This study was designed to describe long-term trends in deaths caused by law enforcement.
Methods
We conducted a serial cross-sectional analysis of national mortality data from 2003 to 2024 using NCHS Mortality Multiple Cause-of-Death records.4 Decedents were included if the underlying or contributing cause of death included legal intervention (International Classification of Diseases and Related Health Problems, Tenth Revisioncodes Y35 or Y89.0), excluding legal execution (Y35.5).
We calculated population-based crude death rates per 100 000 person-years using population denominators from Centers for Disease Control and Prevention WONDER (Wide-ranging Online Data for Epidemiologic Research) and stratified analyses by age (children were aged <18 years), sex, and race and ethnicity (Supplement 1).
Temporal trends were modeled using Poisson regression with a log link and population offset to estimate annual percent change (APC). Segmented models with prespecified break points at 2015 and 2020 were used to assess changes in slope over time based on hypothesized shifts in national attention to police use of force following the deaths of Michael Brown in 2014, Freddie Gray in 2015, and George Floyd in 2020. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline. The threshold for significance was .05 and tests were 2-sided.
Results
From 2003 to 2024, US death certificates indicated that 11 775 deaths were caused by law enforcement, including 11 505 adults and 270 children (Table). Annual deaths increased from 363 (0.17 per 100 000 person-years) to 809 (0.31 per 100 000 person-years) among adults (absolute rate change [ARC], 0.14 per 100 000 person-years; APC, 3.5% [95% CI, 2.9%-4.2%]) and from 8 (0.01 per 100 000 person-years) to 37 (0.05 per 100 000 person-years) among children (ARC, 0.04 per 100 000 person-years; APC, 5.6% [95% CI, 2.7%-8.7%]; Figure). Firearms accounted for 10 314 deaths (87.6%).
Table. Demographics of Deaths Caused by Law Enforcement, 2003 to 2024a
| No. (crude rate per 100 000 person-years) | |||||
|---|---|---|---|---|---|
| All races and ethnicities | Non-Hispanic | Hispanic | Otherb | ||
| Black | White | ||||
| Age, median (IQR), y | 35 (27-45) | 30 (23-40) | 35 (27-46) | 31 (25-39) | 34 (27-43) |
| All persons | 11 775 (0.17) | 2, 759 (0.32) | 5357 (0.17) | 2387 (0.21) | 589 (0.13) |
| Age group, y | |||||
| 0 to 12 | 11 (0.001) | <10 | <10 | <10 | <10 |
| 13 to 17 | 259 (0.06) | 113 (0.17) | 54 (0.02) | 64 (0.07) | 15 (0.05) |
| 18 to 29 | 3718 (0.32) | 1189 (0.75) | 1173 (0.19) | 960 (0.42) | 194 (0.23) |
| 30 to 44 | 4749 (0.34) | 990 (0.57) | 2184 (0.27) | 1003 (0.40) | 255 (0.24) |
| 45 to 64 | 2665 (0.15) | 419 (0.21) | 1671 (0.14) | 332 (0.16) | 113 (0.11) |
| ≥65 | 373 (0.04) | 42 (0.05) | 273 (0.04) | 26 (0.03) | 11 (0.02) |
| Sex | |||||
| Male | 11 204 (0.34) | 2638 (0.65) | 5048 (0.24) | 2313 (0.40) | 561 (0.26) |
| Female | 571 (0.02) | 121 (0.03) | 309 (0.01) | 74 (0.01) | 28 (0.01) |
| Age group and sex | |||||
| 0 to 12 y, Female | <10 | <10 | <10 | <10 | <10 |
| 0 to 12 y, Male | <10 | <10 | <10 | <10 | <10 |
| 13 to 17 y, Female | 18 (0.01) | <10 | <10 | <10 | <10 |
| 13 to 17 y, Male | 228 (0.10) | 106 (0.31) | 50 (0.04) | 58 (0.12) | 14 (0.09) |
| 18 to 29 y, Female | 149 (0.03) | 46 (0.06) | 65 (0.02) | 27 (0.02) | 11 (0.03) |
| 18 to 29 y, Male | 3367 (0.60) | 1143 (1.45) | 1108 (0.35) | 933 (0.78) | 183 (0.44) |
| 30 to 44 y, Female | 194 (0.03) | 43 (0.05) | 115 (0.03) | 28 (0.02) | <10 |
| 30 to 44 y, Male | 4238 (0.64) | 947 (1.14) | 2069 (0.52) | 975 (0.74) | 247 (0.48) |
| 45 to 64 y, Female | 143 (0.02) | 21 (0.02) | 104 (0.02) | 11 (0.01) | <10 |
| 45 to 64 y, Male | 2392 (0.29) | 398 (0.43) | 1567 (0.27) | 321 (0.31) | 106 (0.22) |
| ≥65 y, Female | 23 (0.004) | <10 | 19 (0.005) | <10 | 0 |
| ≥65 y, Male | 329 (0.08) | 39 (0.11) | 254 (0.08) | 25 (0.08) | 11 (0.05) |
Figure. Line Graphs of Rates of Deaths Caused by Law Enforcement Among Children and Adults
The absolute rate is the mean rate over the period. The annual rate change (ARC) is the difference in rates at the beginning and end of the period. Annual percent change (APC) represents the model-estimated annual relative change in the mortality rate during each period, estimated using Poisson regression with a log link and population offset. Data were analyzed using segmented regression models that included prespecified break points at 2015 and 2020, selected based on hypothesized shifts in national attention to police use of force following high-profile deaths of Michael Brown, Freddie Gray, and George Floyd. These analyses assess whether the change in slope is significant at the break points (whereas the change in rate is the difference in rates within time periods).
aIn 2015, the change in slope for adults was significant (P = .005).
bIn 2020, the change in slope for adults was significant (P = .049).
Among adults, mortality rates were increasing before 2015 (ARC, 0.03 per 100 000 person-years; APC, 2.5% [95% CI, 1.7%-3.3%]) and increased more rapidly between 2015 and 2020 (ARC, 0.09 per 100 000 person-years; APC, 6.5% [95% CI, 4.2%-8.8%]; P = .005 for change in slope). After 2020, the rate of increase was lower (ARC, 0.04 per 100 000 person-years; APC, 1.8% [95% CI, −0.7% to 4.4%]), with a change in slope at 2020 (P = .049). Among children, rates were stable before 2015 (ARC, 0; APC, 0.7% [95% CI, −3.6% to 5.2%]) and increased through 2015 to 2020 (ARC, 0.02 per 100 000 person-years; APC, 12.2% [95% CI, 1.0%-24.7%]) and after 2020 (ARC, 0.03 per 100 000 person-years; APC, 17.3% [95% CI, 0.9%-36.4%]); changes in slope at 2015 and 2020 were not statistically significant.
Nearly all deaths were among males (n = 11 204 [95.2%]), and non-Hispanic Black males had the highest mortality rate in every year examined (0.65 per 100 000 person-years) (Figure), including during adolescence (aged 13 to 17 years; 0.31 per 100 000 person-years). Across all groups, the highest mortality rate was among Black males aged 18 to 29 years (1.45 per 100 000 person-years).
Discussion
From 2003 to 2024, annual deaths caused by law enforcement and reported in death certificate data increased from 363 to 809 among adults and from 8 to 37 among children. Non-Hispanic Black males had the highest mortality rates each year, indicating persistent racial disparities, even in adolescence. Firearms caused more than 87% of deaths.
Few previous studies have reported on deaths in children caused by law enforcement. In this study, the relative increase in pediatric deaths caused by law enforcement exceeded that in adult deaths. Non-Hispanic Black and Hispanic persons were at higher risk of death by police than White persons, with increasing disparities observed over time. These disparities are consistent with other studies.5
This study has limitations. First, deaths caused by law enforcement are underreported in vital statistics data—by as much as 55% in prior work—both from misclassification and incomplete reporting.2 Comparative studies suggest that vital statistics data have limited sensitivity for identifying deaths caused by law enforcement relative to other systems, although specificity is likely high given standardized International Classification of Diseases and Related Health Problems, Tenth Revision coding for legal intervention.1 Thus, rates observed in death certificate data should be considered underestimates. Second, temporal variation in death certificate coding is likely because media attention may influence documentation of police involvement. Observed increases in 2015 and 2020 could reflect true increases in deaths or improvements in reporting. For example, Mapping Police Violence reports higher annual counts, although the increases in deaths through 2024 appear similar.6
These findings highlight the need for more accurate mortality surveillance and for interventions to reduce deaths caused by law enforcement, particularly efforts to prevent firearm deaths.
