Key Points
Question
Did homicide of females aged 15 to 44 years (hereinafter, women of reproductive age) recover uniformly after the COVID-19 surge, or did post-2022 trajectories diverge across state abortion policy environments?
Findings
In this study of US mortality data from 2016 through 2024, national homicide mortality decreased after the COVID-19 surge, but post-Dobbs trajectories diverged between states with and without sustained abortion bans. The fitted homicide rate of women of reproductive age in states with total or near-total abortion bans was 7% higher than the modeled counterfactual in 2023 and 15% higher in 2024, corresponding to an estimated 48 and 87 excess deaths, respectively.
Meaning
The findings suggest that national decreases in homicide mortality may mask widening differences in homicide mortality among women of reproductive age across state abortion policy environments.
Abstract
Importance
Homicide is a leading cause of death among female adolescents and women aged 15 to 44 years (hereinafter, women of reproductive age) in the US. Whether homicide trajectories differed across state abortion policy environments after the COVID-19 surge and the Dobbs v Jackson Women’s Health Organization (hereinafter, Dobbs) decision is unclear.
Objectives
To characterize national homicide trends among women of reproductive age and assess whether post-2022 trajectories in homicide mortality among women of reproductive age diverged between states with and without sustained abortion bans.
Design, Setting, and Participants
This quasi-experimental study used national interrupted time series and comparative interrupted time series analyses of homicide mortality from January 1, 2016, through December 31, 2024, among women of reproductive age.
Exposure
The COVID-19 pandemic and sustained state abortion bans after Dobbs.
Main Outcomes and Measures
The main outcome was monthly homicide deaths. The population-offset Poisson model estimated differential post-Dobbs changes in homicide rates. Absolute rate differences, excess deaths, and years of life lost were estimated from fitted and counterfactual Poisson models. States with sustained total or near-total abortion bans after Dobbs were compared with states without such bans; states with partial or legally fluid restrictions were excluded. A triple-difference analysis comparing those aged 15 to 29 years with those aged 45 to 59 years served as a negative-control analysis.
Results
The analysis included 19 118 homicide deaths among women of reproductive age. Nationally, homicide increased sharply during the COVID-19 surge and subsequently decreased, with no clear national level or slope change beginning in July 2022. In policy-stratified analyses, however, homicide trajectories diverged. After July 2022, the exposed states (those with total or near-total abortion bans) vs unexposed states homicide monthly rate ratio (RR) was 1.006 (95% CI, 1.000-1.011; P = .03). Across calendar years 2023 and 2024, the fitted homicide rate in exposed states was higher than the modeled counterfactual (2023: RR, 1.07; 95% bootstrap interval, 1.01-1.13; and 2024: RR, 1.15; 95% bootstrap interval, 1.03-1.29). This corresponded to an estimated 48 excess deaths (95% bootstrap interval, 11-81) in 2023 and 87 excess deaths (95% bootstrap interval, 20-149) in 2024. A triple-difference model indicated greater post-Dobbs divergence among those aged 15 to 29 years than those aged 45 to 59 years (monthly ratio of RRs, 1.017; 95% CI, 1.000-1.034; P = .046); placebo Dobbs dates yielded nonsignificant results.
Conclusions and Relevance
In this study, national homicide mortality rates among women of reproductive age returned toward prepandemic levels after COVID-19, but post-2022 trajectories diverged across state abortion policy environments. The findings suggest that national improvements in homicide mortality rates among women of reproductive age may mask less-favorable homicide trends among women of reproductive age in states with sustained abortion bans.
This study examines national homicide trends among women of reproductive age and assesses whether post–COVID-19 trajectories in homicide mortality among women of reproductive age diverged between states with and without sustained abortion bans.
Introduction
Homicide is consistently among the top 5 leading causes of death among female adolescents and women aged 15 to 44 years (hereinafter, women of reproductive age) in the US and occurs at substantially higher rates than in other high-income countries.1 During the COVID-19 pandemic, many countries experienced marked increases in not only disease-related mortality but also in external causes of death, including homicide.2,3 The pandemic may have heightened violence risk for women in particular through school and workplace closures, reduced access to social services, and lower mobility outside the home compared with males.3,4 COVID-19 also emerged during an ongoing US maternal health crisis, with homicide serving as an important cause of pregnancy-associated death.5,6 The pandemic was followed by unprecedented changes in state reproductive policy after Dobbs v Jackson Women’s Health Organization (hereinafter Dobbs) overturned the federal constitutional right to abortion. More broadly, research shows US mortality is increasingly stratified by place, with widening survival gaps reflecting economic and social conditions as well as divergent policy environments.7,8,9 Against this backdrop, it remains unclear whether homicide among women of reproductive age followed a common trajectory after the COVID-19 surge or diverged across state abortion policy environments after Dobbs.
Evidence suggests that abortion restrictions shape the risk of violence, including lethal violence, for women of reproductive age.10,11,12,13 Being denied a wanted abortion has been associated with prolonged exposure to violence and persistent economic hardship.10,14,15 Prior to Dobbs, each additional targeted regulation of abortion providers law was associated with a 4.4% increase in total homicide and a 3.4% increase in intimate partner violence (IPV)–related homicide among women and girls aged 10 to 44 years.11 In the post-Dobbs era, greater increases in travel distance to abortion facilities were associated with 7% to 10% increases in reported IPV-related crime.12 These burdens may be unevenly distributed: most US patients seeking abortion have low incomes, and many pay out of pocket.16,17,18 Greater travel and financial strain may compound vulnerability to violence and its progression to lethality.19,20,21 Together, this evidence raises concern that homicide trajectories among women of reproductive age may have differed across state reproductive policy environments during the same period.
We therefore examined homicide mortality among women of reproductive age in the US across 2 major disruptions: the COVID-19 pandemic and the post-Dobbs reorganization of abortion policy. This temporal overlap complicates efforts to distinguish changes after the COVID-19 homicide surge from subsequent policy divergence. As a result, we first descriptively characterized national homicide trends before, during, and after the COVID-19 surge. Second, as the primary inferential analysis, we tested whether the national post–COVID-19 surge trajectory masked divergence between states that implemented sustained total or near-total abortion bans after Dobbs and states without such bans, using a population-adjusted comparative interrupted time series model that allowed COVID-19–era dynamics to differ between policy groups. Third, we assessed whether post-Dobbs divergence was concentrated among women of reproductive age and quantified the associated population burden in excess deaths and years of life lost.
Methods
Data Source
This population-level quasi-experimental study used interrupted time series and comparative interrupted time series analyses of aggregate monthly homicide mortality. Monthly homicide mortality counts were obtained from the Centers for Disease Control and Prevention’s Wide-Ranging Online Data for Epidemiologic Research (WONDER) Underlying Cause of Death database, which compiles death certificate data from the National Center for Health Statistics. Deaths were classified using the International Statistical Classification of Disease and Related Health Problems, Tenth Revision; homicide mortality was defined using underlying cause-of-death codes X85 to Y09. Population denominators were obtained from the US Census Bureau Population Estimates Program. The study period spanned January 1, 2016, through December 31, 2024, reflecting the most recent complete year available and allowing multiple pre–COVID-19 baseline years. The study followed Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. The University of Arizona institutional review board determined that this analysis of publicly available, deidentified data was not human participant research; approval and informed consent were not required.
Population
The primary analytic population comprised female decedents aged 15 to 44 years residing in the 50 US states or District of Columbia whose underlying cause of death was homicide; US territories were excluded. For policy-stratified analyses, exposed states were those with sustained total or near-total abortion bans in effect during 2023 and 2024: Alabama, Arkansas, Idaho, Kentucky, Louisiana, Mississippi, Oklahoma, South Dakota, Tennessee, Texas, and West Virginia. The unexposed cluster included the remaining states and District of Columbia that were not excluded. Florida, Georgia, Iowa, and South Carolina were excluded because early-gestation or comparable partial restrictions created partially exposed or legally fluid policy environments (eTable 1 in Supplement 1).
Because monthly state-level homicide counts are frequently suppressed, states were aggregated into prespecified, fixed policy clusters, preserving complete monthly series without imputation; inference was therefore at the policy-cluster level. Texas was included in the primary exposed cluster and also examined separately. Women aged 45 to 59 years were examined as a negative-control population.
Statistical Analysis
Analyses were conducted from May 2025 through June 2026 and updated as complete 2024 mortality data became available. We first characterized national homicide trends among decedents aged 15 to 44 years using a national segmented interrupted time series regression. The model included a continuous monthly trend, calendar month indicators, and level and slope terms for the COVID-19 surge beginning April 2020, the postsurge recovery beginning January 2022, and the post-Dobbs period beginning July 2022.22,23
The primary state policy–stratified analysis used Poisson regression with the logarithm of the female population aged 15 to 44 years as an offset. The model included an exposed-cluster indicator, common COVID-19 surge and postsurge recovery level and slope terms, exposed-cluster interactions for COVID-19–period level and slope changes, calendar month indicators, and an exposed cluster × months since Dobbs interaction beginning in July 2022.22,24 These interactions allowed the COVID-19 surge to alter homicide levels and trends differently in exposed and unexposed states, rather than assuming common pandemic-era changes across policy groups. Identification therefore relies on the assumption that—after accounting for preexisting group differences, differential COVID-19–era level and slope changes, and the modeled January 2022 recovery shift—the exposed-vs-unexposed trajectory would not otherwise have undergone an additional systematic change after July 2022.25,26 Because a population-level association with abortion policy would plausibly emerge gradually as pregnancies and related exposures accumulated, rather than produce an immediate discontinuity, the primary estimand was the post-Dobbs difference in slopes, expressed as the multiplicative monthly change in the exposed vs unexposed rate ratio (RR). Inference used calendar month–indexed Newey-West covariance estimates with 12 monthly lags.
Pre–COVID-19 comparability was assessed using visual inspection, a differential linear pretrend test, and a joint test of year-specific exposed group interactions (eTable 2 in Supplement 1). Because homicide is rare and analyses required policy cluster aggregation, the linear slope was used as a parsimonious summary of the overall post-Dobbs differential trajectory; a flexible specification replacing the monthly slope interaction with separate 6-month exposed group intervals assessed departures from linearity (eMethods, eTable 6, and eFigure in Supplement 1). We also examined alternative parameterizations of the post–COVID-19 recovery period.
To aid interpretation, primary Poisson estimates were translated to calendar year relative and absolute scales. For 2024, the aggregate RR compared fitted exposed state deaths with counterfactual deaths obtained after setting the exposed group post-Dobbs slope term to 0. The corresponding absolute rate difference was expressed as annualized homicide deaths per 100 000 women of reproductive age. Excess deaths were estimated from age-specific Poisson models for decedents aged 15 to 29 and 30 to 44 years by comparing fitted with counterfactual exposed state deaths in 2023 and 2024. Years of life lost were estimated using remaining life expectancy from the National Center for Health Statistics female period life table at midpoint ages of 22 and 37 years. Uncertainty in burden estimates was assessed using a 12-month moving-block residual bootstrap with 5000 replications.
Complementary analyses included a linear count model using the same slope-only timing structure and a month-year fixed-effects specification. A sensitivity analysis shifted the post-Dobbs breakpoint to May 2022 to reflect the leak of Supreme Court Justice Samuel Alito’s draft opinion27 (eTable 10 in Supplement 1). Timing-specific placebo analyses moved the intervention to July 2021, January 2022, and April 2022 and used data only through June 2022 to test whether comparable differential slopes were evident before Dobbs (eTable 7 in Supplement 1). We also separated Texas from the remaining exposed states and assigned Texas a September 2021 policy onset under Senate Bill 8 (the Texas Heartbeat Act).26 Exploratory analyses examined firearm homicide as a share of all homicide and estimated a Texas-separated firearm count model using the Senate Bill 8 timing.
Finally, we estimated a prespecified triple-difference negative-control model comparing those aged 15 to 29 years with those aged 45 to 59 years across exposed and unexposed clusters. Women aged 45 to 59 years share state-level social, economic, and violence environments but are substantially less likely to be directly affected by changes in abortion access. The triple-difference estimand assessed whether post-Dobbs divergence was greater among younger women of reproductive age than among the older comparison group. Analyses were conducted using Stata, version 17.0 (Stata Corp), and Python, version 3.14 (Python Software Foundation), and figures were created using Matplotlib, version 3.10.8, in Python. All P values were from 2-sided tests and results were deemed statistically significant at P < .05.
Results
From January 2016 through December 2024, the primary analysis included 216 monthly policy-cluster observations and 19 118 homicides among women of reproductive age; all monthly series were complete. Before COVID-19, homicide mortality was higher in states that later implemented sustained abortion bans than in unexposed states (Table 1). From January 2016 through March 2020, exposed states had an annualized homicide rate of 4.61 per 100 000 women of reproductive age, compared with 2.99 per 100 000 in unexposed states. Despite these differences in baseline levels, pre–COVID-19 trajectories were similar across policy groups (Table 1; eTable 2 in Supplement 1).
Table 1. Pre–COVID-19 Characteristics of the US Population and State Policy Clusters, January 2016 to March 2020.
| Characteristic | All US states and DC (including excluded states) | Exposed states (sustained abortion bans)a | Unexposed states (no sustained abortion bans)b,c |
|---|---|---|---|
| Population and homicide burden | |||
| No. of jurisdictions | 51 | 11 | 36 |
| Annual population of women of reproductive age, mean (SD), millionsd,e | 63.8 (0.3) | 12.7 (0.1) | 43.6 (0.1) |
| Monthly homicide deaths among women of reproductive age, mean (SD), No.e,f | 183.0 (17.3) | 48.8 (7.6) | 108.5 (12.9) |
| Annualized homicide measuresf | |||
| Annualized homicide deaths among women of reproductive age, No. | 2195.5 | 586.1 | 1301.6 |
| Annualized homicide rate among women of reproductive age, per 100 000 | 3.44 | 4.61 | 2.99 |
| Annualized homicide rate among female adolescents and women aged 15-29 y, per 100 000 | 3.45 | 4.47 | 3.07 |
| Annualized homicide rate among women aged 30-44 y, per 100 000 | 3.42 | 4.75 | 2.90 |
Abbreviation: DC, District of Columbia.
Exposed states were Alabama, Arkansas, Idaho, Kentucky, Louisiana, Mississippi, Oklahoma, South Dakota, Tennessee, Texas, and West Virginia. These states implemented sustained total or near-total abortion bans during the post–Dobbs v Jackson Women’s Health Organization period.
Unexposed jurisdictions were the 35 states and District of Columbia that did not meet the exposed-state definition and were not excluded.
Florida, Georgia, Iowa, and South Carolina were excluded from policy-stratified analyses because early-gestation or comparable partial restrictions resulted in partially exposed or legally fluid policy environments. The all-US column includes these 4 excluded jurisdictions.
Women of reproductive age comprise female adolescents and women aged 15 to 44 years.
Population mean (SD) values were calculated across the 51 monthly observations using the corresponding annual population denominator for each month. Monthly homicide death mean (SD) values were calculated across the same 51 months. Annualized homicide deaths were calculated by multiplying mean monthly deaths by 12. Annualized homicide rates were calculated using the corresponding mean female population.
Values are based on monthly homicide deaths from January 2016 through March 2020, before the COVID-19 surge period. Homicide deaths were defined as deaths due to assault using International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, codes X85 to Y09.
Nationally, homicide deaths among US women of reproductive age increased markedly during the COVID-19 surge and subsequently decreased (Figure 1). Beginning in January 2022, the national trajectory shifted into a postsurge recovery period. There was no evidence of an additional national level change at July 2022 (9.8 deaths/mo; 95% CI, −7.7 to 27.3 deaths/mo; P = .27) or change in slope (−0.2 deaths/mo/mo; 95% CI, −3.9 to 3.5 deaths/mo/mo; P = .92). Thus, the national decrease continued after Dobbs without a detectable population-wide discontinuity (eResults in Supplement 1).
Figure 1. Interrupted Time Series Plot of National Homicide Mortality Among US Women of Reproductive Age, 2016-2024.
Points show observed monthly homicide deaths among US women of reproductive age (defined as female adolescents and women aged 15 to 44 years). The solid line shows the modeled postsurge recovery trajectory from the national interrupted time series model with the July 2022 level and slope terms set to 0; the dashed line shows the projection of the pre–COVID-19 trend. In the full national model, neither the July 2022 level change nor slope change was statistically significant. Dobbs indicates Dobbs v Jackson Women’s Health Organization.
In the primary policy-stratified model, post-Dobbs homicide trajectories differed between exposed and unexposed states (Figure 2A and B; Table 2). After July 2022, the exposed vs unexposed homicide monthly RR was 1.006 (95% CI, 1.000-1.011; P = .03). Figure 2B shows the overall fitted exposed vs unexposed RR, which was already elevated before Dobbs because homicide rates were higher in exposed states at baseline; the post-Dobbs slope term estimates additional widening in that relative trajectory after July 2022. The calendar year estimates in Table 2 isolate this additional post-Dobbs component by comparing fitted exposed state outcomes with the modeled counterfactual obtained after setting the post-Dobbs slope term to 0. On this counterfactual scale, the fitted homicide rate in exposed states was 7% higher in 2023 (RR, 1.07; 95% bootstrap interval, 1.01-1.13) and 15% higher in 2024 (RR, 1.15; 95% bootstrap interval, 1.03-1.29), corresponding to absolute rate differences of 0.34 and 0.61 additional homicide deaths per 100 000 women per year, respectively. Using the same counterfactual framework, age-specific Poisson models estimated 48 excess deaths (95% bootstrap interval, 11-81) in 2023 and 87 (95% bootstrap interval, 20-149) in 2024, corresponding to 2447 years (95% bootstrap interval, 466-4244 years) and 4432 years (95% bootstrap interval, 875-7644 years) of life lost, respectively (Figure 2C; Table 2).
Figure 2. Comparative Time Series Plots of Homicide Mortality in Exposed and Unexposed States, 2016-2024.
A, Observed and seasonally standardized fitted homicide rates in exposed and unexposed states. Seasonally standardized fitted rates come from the primary population-offset Poisson model, and shaded bands indicate 95% CIs. The model allows COVID-19–era level and slope changes to differ between exposed and unexposed policy clusters; the January 2022 recovery terms add a common postsurge change in trajectory while retaining those differential COVID-19–era dynamics. Dobbs indicates Dobbs v Jackson Women’s Health Organization. B, Fitted homicide rate ratio (RR) in exposed vs unexposed states. The line shows the fitted exposed vs unexposed homicide RR from the primary population-offset Poisson model across the full study period. The RR was already greater than 1 before Dobbs, reflecting the preexisting difference in homicide rates; changes before July 2022 reflect modeled differential COVID-19–era dynamics. The post-Dobbs slope interaction estimates additional change in this relative trajectory beginning in July 2022, rather than an immediate level shift. The fitted line is model based rather than a direct month-specific ratio of observed rates. C, Fitted and counterfactual homicide deaths in exposed states. Points show observed monthly homicide deaths in exposed states. The counterfactual was obtained from the same fitted models after setting only the exposed group post-Dobbs slope term to 0; baseline group differences, seasonality, differential COVID-19–era dynamics, the modeled postsurge recovery, and population offsets are otherwise retained. Fitted and counterfactual trajectories are therefore identical before July 2022 and diverge only through the additional modeled post-Dobbs trajectory. The shaded area shows the difference between these trajectories and forms the basis for the model-estimated excess death burden.
Table 2. Primary Post-Dobbs Comparative Estimates and Model-Estimated Population Burden.
| Estimand | Estimate (95% CI or bootstrap interval) |
|---|---|
| Primary post-Dobbs comparative analysis: women aged 15-44 ya | |
| Post-Dobbs difference in slopes, monthly RRb | 1.006 (95% CI, 1.000-1.011); P = .03 |
| Model-estimated calendar year RR, 2023c | 1.07 (1.01-1.13) |
| Model-estimated absolute homicide rate difference, 2023, deaths per 100 000 women of reproductive age/yc | 0.34 (0.03-0.65) |
| Model-estimated calendar-year RR, 2024 | 1.15 (1.03-1.29) |
| Model-estimated absolute homicide rate difference, 2024, deaths per 100 000 women of reproductive age/y | 0.61 (0.05-1.18) |
| Population-adjusted burden in exposed statesd | |
| Estimated excess deaths, 2023, No. | 48 (11-81) |
| Years of life lost, 2023, No. | 2447 (466-4244) |
| Estimated excess deaths, 2024, No. | 87 (20-149) |
| Years of life lost, 2024, No. | 4432 (875-7644) |
Abbreviations: Dobbs, Dobbs v Jackson Women’s Health Organization; RR, rate ratio.
Estimates from the primary population-offset Poisson comparative interrupted time series model of monthly homicide deaths among women of reproductive age (defined as female adolescents and women aged 15 to 44 years) in exposed and unexposed policy clusters from January 2016 through December 2024. The model accounted for seasonality, common COVID-19 and postsurge recovery changes, and differential COVID-19–era changes between policy groups.
The post-Dobbs difference in slopes is the primary estimand. The monthly RR represents the multiplicative monthly change in the exposed vs unexposed homicide RR after July 2022; values greater than 1 indicate progressively less favorable relative homicide trends in exposed states.
Calendar year RRs and absolute rate differences translate the same model to annual scales. Counterfactual exposed-state outcomes were estimated by setting the exposed group post-Dobbs slope term to 0 while retaining baseline group differences and other modeled terms. Absolute differences are annualized deaths per 100 000 women of reproductive age.
Excess deaths and years of life lost from age-specific Poisson models for female adolescents and women aged 15 to 29 and women aged 30 to 44 years using the same counterfactual. Newey-West covariance estimates with 12 monthly lags were used for the primary slope; annual and burden intervals used a 12-month moving-block residual bootstrap with 5000 replications. Years of life lost used the 2023 National Center for Health Statistics female period life table at midpoint ages 22 and 37 years.
Several analyses assessed whether the estimated post-Dobbs divergence reflected preexisting differential trends, pandemic-era group differences, or the imposed linear functional form. Formal pre–COVID-19 tests found no evidence of differential trends, and differential COVID-19–era changes did not indicate greater worsening in exposed states (eTable 2 and eTable 3 in Supplement 1). A flexible 6-month specification showed little immediate relative change followed by larger contrasts in 2023 and early 2024; the lower estimate in late 2024 was sensitive to a single unusually low exposed-cluster month (eTable 6 and eTable 6A; eFigure in Supplement 1). Nearby pre-Dobbs placebo dates did not produce significant differential slopes (eTable 7 in Supplement 1). In the prespecified negative-control analysis, post-Dobbs divergence was greater among those aged 15 to 29 years than among those aged 45 to 59 years (monthly ratio of RRs, 1.017; 95% CI, 1.000-1.034; P = .046; eTable 5 in Supplement 1). Results were directionally consistent in complementary count and Texas-separated analyses (eTables 4 and 8 in Supplement 1). Exploratory firearm analyses indicated increasing firearm involvement across policy groups, reducing concern that the findings are solely explained by contemporaneous firearm policy changes (eTable 9 in Supplement 1).
Discussion
Nationally, homicide among women of reproductive age increased markedly during the COVID-19 surge and subsequently decreased toward prepandemic levels. However, this national recovery was not uniform across state abortion policy environments. In the policy-stratified model, homicide trajectories became less favorable in states with sustained abortion bans after July 2022, with the fitted 2024 homicide rate 15% higher than the modeled counterfactual, which corresponded with an estimated 87 excess deaths and 4432 years of life lost in 2024. The divergence appeared to emerge gradually and was greater among younger women of reproductive age than the older negative-control group.
Several pathways may link abortion restrictions with women’s exposure to violence. Pregnancy and the postpartum period are established times of vulnerability to IPV, and unintended pregnancy and IPV are linked bidirectionally.28,29 A prior study found that 6% to 22% of women having abortions report recent IPV,10 and nearly one-third of women in the Turnaway Study30 reported partner-related reasons for seeking abortion. In addition, women denied abortion may experience more prolonged exposure to violence than those who obtain care.10,30 The relevant pathways may also extend beyond IPV or pregnancy-associated deaths to economic hardship, household instability, and constrained autonomy, with homicide representing the most extreme end point of broader violence-related harm.14,15,19,20,21 These mechanisms were not directly measured in the mortality data and should be interpreted as plausible pathways rather than demonstrated mediators. The estimates of years of life lost additionally underscore that the observed divergence reflects premature mortality concentrated in younger adulthood.
These results should also be interpreted within broader geographic differences in violence risk. States with sustained abortion bans entered the study period with higher homicide mortality and differ from unexposed states in firearm access and other social and policy characteristics associated with violence.31,32 Exploratory firearm analyses showed that firearm homicide comprised a high and increasing share of homicide in both exposed and unexposed states. In Texas, firearm-specific divergence was evident using the September 2021 Senate Bill 8 timing, before the June 2022 New York State Rifle & Pistol Association v Bruen decision33 and Bipartisan Safer Communities Act.34 This reduces concern that the primary finding is solely an artifact of assigning exposure near contemporaneous national firearm policy changes, but it does not eliminate confounding by firearm policy or distinguish the independent effects of correlated state policy environments. Similarly, the Texas-specific Poisson estimate was directionally consistent but imprecise, whereas the non-Texas–exposed-state estimate remained positive and statistically significant.
Homicide mortality among women of reproductive age is also marked by profound racial and ethnic inequality, with Black women facing substantially higher risk.35 Because abortion bans are concentrated in regions shaped by longstanding structural inequities, changes in reproductive policy may interact with preexisting racialized patterns of violence.6 Our cluster-level data did not permit sufficiently stable race- and ethnicity-stratified estimates, and future work should assess whether the post-Dobbs divergence has widened racial and ethnic disparities in homicide.
Limitations
Several limitations merit consideration. First, homicide classification may vary across jurisdictions or population groups, and death certificate data lack information on pregnancy status, perpetrator relationship, IPV context, and economic circumstances. As an ecological analysis, the study cannot identify individual exposure to abortion restrictions or establish the mechanisms linking policy environments with homicide mortality. Second, monthly state-level homicide counts are frequently suppressed, requiring prespecified policy clusters rather than state-level monthly classifications. This preserved complete time series without imputation but obscured within-state heterogeneity and simplified a fluid post-Dobbs policy landscape. Third, although pre–COVID-19 trends were similar, nearby pre-Dobbs placebo dates did not reproduce the primary finding, and models allowed COVID-19–era dynamics to differ by policy group, unmeasured contemporaneous policy changes or broader shifts in violence environments could still contribute to the observed divergence. Because only 6 months separated the January 2022 recovery transition from the July 2022 Dobbs breakpoint, the data provide limited information for distinguishing continued differential recovery from a new post-Dobbs divergence. Estimates were also sensitive to how the post–COVID-19 recovery period was parameterized; the January 2022 recovery transition was retained because it reflected the national postsurge decrease, but this dependence limits causal attribution. Fourth, the primary model summarized post-Dobbs divergence with a linear slope, and flexible estimates did indicate an uneven pattern over time but were sensitive to one unusually low exposed-cluster month. Fifth, implementation timing varied across states, and Texas was exposed earlier under Senate Bill 8. The estimates therefore characterized divergence across post-Dobbs policy environments rather than the causal effect of any single policy action.
Conclusions
In this study, national homicide mortality among women of reproductive age decreased after the COVID-19 surge, but post-2022 trajectories diverged across state abortion policy environments. The findings suggest that national improvements may mask less-favorable homicide trends among women of reproductive age in states with sustained abortion bans. Continued follow-up is needed to determine whether this divergence persists and to clarify mechanisms underlying it.
eMethods.
eResults.
eTable 1. Classification of US States Into Exposed, Unexposed, and Excluded Analytic Clusters
eTable 2. Preperiod Trend Diagnostics
eTable 3. Differential COVID-19 Dynamics in the Slope-Only Models
eTable 4. Complementary Linear Count and Month-Year Fixed-Effects Sensitivity Analyses
eTable 5. Age-Specific and Negative-Control Post-Dobbs Estimates From Population-Offset Poisson Models
eTable 6. Flexible 6-Month Post-Dobbs Estimates From the Population-Offset Poisson Model
eTable 6A. Leave-One-Month-Out Diagnostic for the July-to-December 2024 Flexible Estimate
eTable 7. Placebo Intervention-Date Analyses Before Dobbs
eTable 8. Texas-Separated 3-Group Sensitivity Analysis, 2018-2024
eTable 9. Exploratory Firearm Analyses, 2018-2024
eTable 10. National Interrupted Time-Series Sensitivity Analysis With May 2022 Breakpoint
eFigure. Flexible 6-Month Post-Dobbs Estimates From the Population-Offset Poisson Model
eReferences.
Data Sharing Statement
References
- 1.Grinshteyn E, Hemenway D. Violent death rates in the US compared to those of the other high-income countries, 2015. Prev Med. 2019;123:20-26. doi: 10.1016/j.ypmed.2019.02.026 [DOI] [PubMed] [Google Scholar]
- 2.Curtin SC. Trends in death rates for leading methods of injury: United States, 2003-2023. NCHS Data Brief. 2025;(526):CS357496. doi: 10.15620/cdc/174582 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Piquero AR, Roman JK. Firearm homicide demographics before and after the COVID-19 pandemic. JAMA Netw Open. 2024;7(5):e2412946. doi: 10.1001/jamanetworkopen.2024.12946 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Woskie L, Wenham C. Do men and women “lockdown” differently? examining Panama’s COVID-19 sex-segregated social distancing policy. Fem Econ. 2021;27(1-2):327-344. doi: 10.1080/13545701.2020.1867761 [DOI] [Google Scholar]
- 5.Wallace ME. Trends in pregnancy-associated homicide, United States, 2020. Am J Public Health. 2022;112(9):1333-1336. doi: 10.2105/AJPH.2022.306937 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wallace ME, Jahn JL. Pregnancy-associated mortality due to homicide, suicide, and drug overdose. JAMA Netw Open. 2025;8(2):e2459342. doi: 10.1001/jamanetworkopen.2024.59342 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Chetty R, Stepner M, Abraham S, et al. The association between income and life expectancy in the United States, 2001-2014. JAMA. 2016;315(16):1750-1766. doi: 10.1001/jama.2016.4226 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Machado S, Kyriopoulos I, Orav EJ, Papanicolas I. Association between wealth and mortality in the United States and Europe. N Engl J Med. 2025;392(13):1310-1319. doi: 10.1056/NEJMsa2408259 [DOI] [PubMed] [Google Scholar]
- 9.Meara ER, Richards S, Cutler DM. The gap gets bigger: changes in mortality and life expectancy, by education, 1981-2000. Health Aff (Millwood). 2008;27(2):350-360. doi: 10.1377/hlthaff.27.2.350 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Roberts SC, Biggs MA, Chibber KS, Gould H, Rocca CH, Foster DG. Risk of violence from the man involved in the pregnancy after receiving or being denied an abortion. BMC Med. 2014;12:144. doi: 10.1186/s12916-014-0144-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Wallace ME, Stoecker C, Sauter S, Vilda D. States’ abortion laws associated with intimate partner violence-related homicide of women and girls in the US, 2014-20. Health Aff (Millwood). 2024;43(5):682-690. doi: 10.1377/hlthaff.2023.01098 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Dave D, Durrance CP, Erten B, Wang Y, Wolfe B. Abortion restrictions and intimate partner violence in the Dobbs era. J Health Econ. 2025;104:103074. doi: 10.1016/j.jhealeco.2025.103074 [DOI] [PubMed] [Google Scholar]
- 13.Boyle KM, Regoeczi W, Meyer CB. State divorce laws, reproductive care policies, and pregnancy-associated homicide rates, 2018-2021. JAMA Netw Open. 2024;7(11):e2444199. doi: 10.1001/jamanetworkopen.2024.44199 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Foster DG, Biggs MA, Ralph L, Gerdts C, Roberts S, Glymour MM. Socioeconomic outcomes of women who receive and women who are denied wanted abortions in the United States. Am J Public Health. 2018;108(3):407-413. doi: 10.2105/AJPH.2017.304247 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Grace KT, Anderson JC. Reproductive coercion: a systematic review. Trauma Violence Abuse. 2018;19(4):371-390. doi: 10.1177/1524838016663935 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Jones RK, Upadhyay UD, Weitz TA. At what cost? payment for abortion care by U.S. women. Womens Health Issues. 2013;23(3):e173-e178. doi: 10.1016/j.whi.2013.03.001 [DOI] [PubMed] [Google Scholar]
- 17.Barr-Walker J, Jayaweera RT, Ramirez AM, Gerdts C. Experiences of women who travel for abortion: a mixed methods systematic review. PLoS One. 2019;14(4):e0209991. doi: 10.1371/journal.pone.0209991 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Addante AN, Paul R, Dorsey M, McNicholas C, Madden T. Differences in financial and social burdens experienced by patients traveling for abortion care. Womens Health Issues. 2021;31(5):426-431. doi: 10.1016/j.whi.2021.06.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Campbell JC, Glass N, Sharps PW, Laughon K, Bloom T. Intimate partner homicide: review and implications of research and policy. Trauma Violence Abuse. 2007;8(3):246-269. doi: 10.1177/1524838007303505 [DOI] [PubMed] [Google Scholar]
- 20.Glass N, Laughon K, Campbell J, et al. Non-fatal strangulation is an important risk factor for homicide of women. J Emerg Med. 2008;35(3):329-335. doi: 10.1016/j.jemermed.2007.02.065 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Zeoli AM, Malinski R, Turchan B. Risks and targeted interventions: firearms in intimate partner violence. Epidemiol Rev. 2016;38(1):125-139. doi: 10.1093/epirev/mxv007 [DOI] [PubMed] [Google Scholar]
- 22.Bernal JL, Cummins S, Gasparrini A. Interrupted time series regression for the evaluation of public health interventions: a tutorial. Int J Epidemiol. 2017;46(1):348-355. doi: 10.1093/ije/dyw098 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Baicker K, Svoronos T. Testing the validity of the single interrupted time series design. NBER Working Paper No. 2019-97. Published online July 31, 2019.
- 24.Dimick JB, Ryan AM. Methods for evaluating changes in health care policy: the difference-in-differences approach. JAMA. 2014;312(22):2401-2402. doi: 10.1001/jama.2014.16153 [DOI] [PubMed] [Google Scholar]
- 25.Bertrand M, Duflo E, Mullainathan S. How much should we trust differences-in-differences estimates? Q J Econ. 2004;119(1):249-275. doi: 10.1162/003355304772839588 [DOI] [Google Scholar]
- 26.Fry CE, Hatfield LA. Birds of a feather flock together: comparing controlled pre-post designs. Health Serv Res. 2021;56(5):942-952. doi: 10.1111/1475-6773.13697 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Reingold RB, Gostin LO. The leaked Supreme Court ruling opinion: implications for abortion access. JAMA. 2022;328(1):13-14. doi: 10.1001/jama.2022.10204 [DOI] [PubMed] [Google Scholar]
- 28.Miller E, McCauley HL, Tancredi DJ, Decker MR, Anderson H, Silverman JG. Recent reproductive coercion and unintended pregnancy among female family planning clients. Contraception. 2014;89(2):122-128. doi: 10.1016/j.contraception.2013.10.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Alhusen JL, Ray E, Sharps P, Bullock L. Intimate partner violence during pregnancy: maternal and neonatal outcomes. J Womens Health (Larchmt). 2015;24(1):100-106. doi: 10.1089/jwh.2014.4872 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Chibber KS, Biggs MA, Roberts SCM, Foster DG. The role of intimate partners in women’s reasons for seeking abortion. Womens Health Issues. 2014;24(1):e131-e138. doi: 10.1016/j.whi.2013.10.007 [DOI] [PubMed] [Google Scholar]
- 31.Keegan G, Hoofnagle M, Chor J, et al. State-level analysis of intimate partner violence, abortion access, and peripartum homicide: call for screening and violence interventions for pregnant patients. J Am Coll Surg. 2024;238(5):880-888. doi: 10.1097/XCS.0000000000001019 [DOI] [PubMed] [Google Scholar]
- 32.Dholakia A, Monuteaux MC, D’Ambrosi G, McLone SG, Fleegler E, Lee LK. Firearm homicide in pregnant women and state-level firearm ownership. JAMA Netw Open. 2025;8(11):e2542447. doi: 10.1001/jamanetworkopen.2025.42447 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.New York State Rifle & Pistol Association, Inc v Bruen, 597 US 1 (2022).
- 34.Bipartisan Safer Communities Act, Pub L No. 117-159, 136 Stat 1313 (2022).
- 35.Waller BY, Joseph VA, Keyes KM. Racial inequities in homicide rates and homicide methods among Black and White women aged 25-44 years in the USA, 1999-2020: a cross-sectional time series study. Lancet. 2024;403(10430):935-945. doi: 10.1016/S0140-6736(23)02279-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods.
eResults.
eTable 1. Classification of US States Into Exposed, Unexposed, and Excluded Analytic Clusters
eTable 2. Preperiod Trend Diagnostics
eTable 3. Differential COVID-19 Dynamics in the Slope-Only Models
eTable 4. Complementary Linear Count and Month-Year Fixed-Effects Sensitivity Analyses
eTable 5. Age-Specific and Negative-Control Post-Dobbs Estimates From Population-Offset Poisson Models
eTable 6. Flexible 6-Month Post-Dobbs Estimates From the Population-Offset Poisson Model
eTable 6A. Leave-One-Month-Out Diagnostic for the July-to-December 2024 Flexible Estimate
eTable 7. Placebo Intervention-Date Analyses Before Dobbs
eTable 8. Texas-Separated 3-Group Sensitivity Analysis, 2018-2024
eTable 9. Exploratory Firearm Analyses, 2018-2024
eTable 10. National Interrupted Time-Series Sensitivity Analysis With May 2022 Breakpoint
eFigure. Flexible 6-Month Post-Dobbs Estimates From the Population-Offset Poisson Model
eReferences.
Data Sharing Statement
