The Autopsy / Case File · Violence & Homicide
Killed By the Person Closest to Her
Nearly 4,000 women were killed by an intimate partner in four years of federal surveillance data. Two in three deaths involved a gun.
When a woman is murdered in the United States, the person most likely to have done it isn't a stranger. It's a current or former partner, most often at home, most often with a firearm. Here's a sourced look at what federal violent-death surveillance actually shows about the pattern, the specific window when the risk spikes, and the screening tool built to catch it before the fatal call.
Join the discussion →
The CDC's National Violent Death Reporting System pulls together coroner and medical examiner reports, police records, and death certificates from nearly every state, case by case, to answer a question ordinary crime statistics usually can't: not just how a woman died, but who did it, where, and what happened right before. Four years of that data, covering 2018 through 2021, produced a specific and consistent picture.
The Numbers
Among 3,991 women identified by NVDRS as victims of intimate partner homicide over that four-year span, the pattern held with unusual consistency year over year, rather than spiking around any single event or region.
The surveillance data also captures the setting and circumstance with a level of detail an ordinary crime blotter doesn't. The overwhelming majority of these killings weren't chance encounters or crimes of opportunity committed by a stranger — they happened in the one place a victim might otherwise expect to be safest.
The rate held essentially flat between the first two years of the study period and the last two — not a crisis spiking suddenly into view, but a stable, ongoing mortality pattern that federal surveillance has now spent years documenting in the same detail as any infectious disease outbreak.
A Gun in the House Changes the Math
The firearm figure isn't incidental to the story — in the epidemiology of intimate partner violence, it's close to the whole story. A landmark case-control study led by Johns Hopkins nursing researcher Jacquelyn Campbell, drawing on interviews with the family members of homicide victims and with survivors of near-fatal violence, found that a gun's mere presence in a home where abuse is already occurring transforms the risk calculation entirely.
Access to a firearm increases the risk that an abusive partner will kill their victim by roughly 500%, independent of every other risk factor researchers controlled for.
That figure comes from formal risk-factor modeling, not intuition — it holds after accounting for prior threats, prior violence, substance use, and the relationship's history. A firearm doesn't just provide a method; it changes the odds an existing pattern of abuse ends in a killing rather than continuing as one more incident in a long history of them.
The Federal Gun Law That Doesn't Cover Everyone
Federal law has prohibited people convicted of a qualifying domestic violence misdemeanor, or subject to certain protective orders, from possessing a firearm since 1996's Lautenberg Amendment. For nearly three decades, though, that prohibition applied only to spouses, people who lived with the victim, or people who shared a child with them — not to a dating partner who had never lived with the victim, regardless of how serious or violent the relationship had been. Advocates and researchers came to call the gap the "boyfriend loophole," and it mattered because dating relationships, not marriages or cohabitations, make up a substantial share of intimate partner violence cases nationally.
The 2022 fix closed the federal gap only for people actually convicted of a qualifying misdemeanor or under a qualifying protective order — it does nothing for the far larger number of abusive relationships that never produce a conviction or a court order at all, whether because a victim didn't report, a case wasn't prosecuted, or a judge declined to issue a restraining order. The law changed who counts as prohibited. It didn't change how many dangerous relationships actually reach that legal threshold in the first place.
The Most Dangerous Time Is Leaving
One of the most consistent, and most counterintuitive, findings in intimate partner homicide research is that the danger to a victim doesn't peak during the relationship's worst period of ongoing abuse — it peaks after the victim tries to leave. Separation represents a victim reclaiming control the abuser has spent the relationship establishing, and for a subset of abusers, that loss of control is itself the trigger for lethal violence.
Research funded by the National Institute of Justice, built on the same case-control methodology behind the firearm-risk finding, identified recent or attempted separation as one of the strongest predictors of intimate partner homicide across the cases studied — stronger than many of the risk factors that intuitively seem more alarming, like a documented history of the most severe prior violence. The take-home implication for anyone working with victims is uncomfortable: the advice to "just leave" is not straightforwardly safety advice. Leaving well, with a safety plan built around the specific danger period that follows it, is a different and more survivable thing than leaving abruptly.
A Screening Tool Built for the 911 Call
Some police departments have built that danger-period research directly into how officers respond to a domestic violence call. The Lethality Assessment Program, developed by the Maryland Network Against Domestic Violence, gives responding officers an 11-question screening instrument — has the abuser ever used a weapon, threatened to kill the victim or someone she loves, choked her, threatened suicide — that flags a subset of calls as carrying an elevated risk of homicide, in real time, before anyone leaves the scene.
A high-risk score triggers an on-the-spot warm handoff: the officer calls a domestic violence hotline directly from the scene and puts the victim on the phone with an advocate before leaving, rather than handing over a pamphlet and a number to call later. Outcome research has found the program measurably increases victims' safety planning and help-seeking, on top of the homicide-rate decline observed during Maryland's rollout. It's now used by police departments in most states — but adoption remains a department-by-department choice, not a national requirement, which means whether a given 911 call gets this specific screening still depends heavily on which jurisdiction answers the phone.
The Disparity Inside the Pattern
The overall numbers above describe a national average. They obscure a racial gap that CDC researchers have documented as persistent across more than two decades of data, not a recent development or a single bad year.
The same CDC analysis found the gap "virtually unchanged" across the study period — not a disparity actively narrowing as public health attention and hotline funding have grown, but one that has simply persisted underneath the national totals reported each year. Intimate partner circumstances specifically, not homicide broadly, is where researchers found the disparity concentrated: Black and Native women weren't just more likely to be murdered overall, they were more likely to have been murdered by a partner when the case was resolved with a known relationship between victim and suspect. A national mortality pattern this well-documented, in other words, is really several overlapping patterns of unequal risk, not one evenly distributed one.
And Then You're Dead
The data here doesn't describe a random hazard, the kind no one could have seen coming. It describes a pattern precise enough that a police department can build an 11-question checklist around it and measurably cut the number of women who don't survive. The mechanism is documented, the highest-risk moment is documented, even the specific weapon is documented two-thirds of the time. What's still missing, case by case, isn't data. It's the intervention arriving before the moment the data already knows is coming.
That's it. That's the whole thing.
Sources
- 3,991 women killed by an intimate partner 2018–2021, 66.6% involving a firearm, 68.0% occurring at the victim's residence, and 98.5% committed by a male suspect — CDC, "Notes from the Field: Intimate Partner Homicide Among Women — United States, 2018–2021," MMWR, August 2024
- National Violent Death Reporting System methodology and scope — CDC, "Surveillance for Violent Deaths — National Violent Death Reporting System, 50 States, the District of Columbia, and Puerto Rico, 2022," MMWR Surveillance Summaries
- Firearm access increasing intimate partner homicide risk by roughly 500%, and recent/attempted separation as a leading risk factor — Campbell et al., "Assessing Risk Factors for Intimate Partner Homicide," National Institute of Justice Journal, funded research summary
- The Lautenberg Amendment's original spouse/cohabitant scope, the "boyfriend loophole," and the 2022 Bipartisan Safer Communities Act's five-year dating-partner firearm prohibition — Everytown Research, "What Is the 'Boyfriend Loophole'?" and background legislative summary, Bipartisan Safer Communities Act, 2022
- 4.9 million prohibited firearm transactions blocked by the National Instant Criminal Background Check System since 1994 — Brady United, "Gun Sale Loopholes"
- Maryland Lethality Assessment Program's 11-question screening tool, its warm-handoff design, and the documented 34% decline in intimate partner homicides in Maryland during the 2007–2012 implementation period — National Institute of Justice, "How Effective Are Lethality Assessment Programs for Addressing Intimate Partner Violence?"
- Homicide rates of 4.4 per 100,000 for non-Hispanic Black women and 4.3 per 100,000 for American Indian/Alaska Native women, versus 1–2 per 100,000 for other groups, and the disparity's persistence from 2003–2014 — CDC, "Racial and Ethnic Differences in Homicides of Adult Women and the Role of Intimate Partner Violence — United States, 2003–2014," MMWR
Photo: U.S. Air Force photo by Airman 1st Class Gabrielle Spradling, via Wikimedia Commons / DVIDS (public domain).
Comments
Loading…