And Then You're Dead

The Autopsy / Case File · Violence & Homicide

The Second Death

The homicide is the part that makes the news. The second death — the one the killer inflicts on himself, usually minutes later, usually with the same gun — is treated almost as an afterthought, a loose end the story doesn't need to explain. But the pairing isn't incidental. It's a distinct, trackable category of American death with its own pattern: who does it, who they kill first, and what usually preceded it.

The Violence Policy Center's ongoing news-based tracking, published as "American Roulette," found 237 murder-suicide events across the U.S. in just the first six months of 2025 — nine a week — resulting in 526 deaths, extrapolated to roughly 1,052 for the full year. Eighty-seven percent involved a firearm. Sixty-five percent involved an intimate partner, and in 99 percent of those cases, the person killed was a woman.

And Then You're Dead · September 2026 · 10 min read

Join the discussion →
A line of orange police tape reading POLICE LINE DO NOT CROSS, stretched across a nighttime scene with police lights glowing red and blue in the background
Police tape at the scene of a shooting in Minneapolis. Most murder-suicides in America follow the same basic shape: a firearm, an intimate partner, and a home address — the scene rarely looks like anything the neighbors expected. Photo: Tony Webster, CC BY 2.0, via Wikimedia Commons.

No federal agency counts murder-suicides as their own category. The FBI's Uniform Crime Reporting program logs the homicide; the suicide, if it's logged as connected at all, ends up in a separate vital-statistics stream that rarely gets cross-referenced back to the killing that preceded it. The most complete national picture comes instead from the Violence Policy Center, a gun-violence-prevention research organization that has published a semiannual "American Roulette" report since the early 2000s, built by systematically searching English-language news coverage for incidents in which one person killed one or more others and then killed themselves within a short time frame, typically the same 24-hour period. It's a methodology with real limits — it depends on cases getting covered by identifiable news outlets, and it will miss incidents in areas with thin local news coverage — but it is, by a wide margin, the most complete accounting that exists, precisely because no government dataset has ever been built to link the two deaths together.

What the Ninth Edition Found

The most recent installment of American Roulette, its ninth edition, published April 16, 2026, covered the first half of 2025 — January 1 through June 30 — and identified 237 distinct murder-suicide events, an average of just over nine per week. Those 237 events produced 526 total deaths: 289 homicide victims and 237 suicides (one perpetrator's death per event, by definition). Doubling the six-month count to estimate a full year, the report's authors arrive at approximately 1,052 murder-suicide deaths annually — a figure that has stayed in a broadly similar range across multiple editions of the report, suggesting a stable, ongoing pattern rather than an anomaly tied to any single reporting period.

The method concentration is stark. Firearms were involved in 207 of the 237 events, 87 percent — far above the roughly 54 percent share firearms hold in U.S. suicides generally, and reflective of the fact that a murder-suicide, almost by definition, requires a method capable of killing more than one person in rapid succession without giving a victim time to flee or a bystander time to intervene. A knife, a car, an overdose — all documented methods in individual murder-suicide cases — are far less reliable at accomplishing that. A gun, loaded and in the room, is not.

0
estimated murder-suicide deaths in the U.S. in 2025, extrapolated from the first half-year
0
percent of documented murder-suicide events in 2025 that involved a firearm

Who Dies, and Who Does the Killing

The gender pattern in murder-suicide is one of the most consistent findings across every edition of the Violence Policy Center's tracking, and it runs in a specific, asymmetric direction. Of the 237 perpetrators identified in the first half of 2025, 92 percent were male. Of the 289 homicide victims, 76 percent were female. Sixty-five percent of all events involved an intimate partner relationship between perpetrator and victim, and within that intimate-partner subset, 99 percent of the people killed were women killed by a male partner or ex-partner — a rate of concentration that leaves almost no ambiguity about the dominant shape of the crime: not a random act of violence, but the fatal endpoint of a specific, gendered pattern of control that intimate partner violence researchers have documented in non-fatal cases for decades before it ever reaches this point. Eighty percent of the events happened inside a home. Forty-five children and teenagers under 18 were themselves killed as part of these incidents in just the first half of 2025, and 27 more survived as witnesses to some part of the event.

Ninety-nine percent of the intimate-partner victims were women. This isn't a random subset of American violence. It's the fatal endpoint of a pattern researchers have been documenting in non-fatal cases for decades.

The Overlap With a Documented Warning-Sign Literature

Murder-suicide involving an intimate partner doesn't emerge from nowhere, and a substantial body of domestic-violence research — separate from, but directly relevant to, the Violence Policy Center's incident tracking — has spent decades identifying the risk factors that most often precede a fatal escalation. Recent or pending separation is one of the strongest predictors: leaving, or preparing to leave, an abusive relationship is statistically one of the most dangerous moments in it, not the safest. Prior non-fatal strangulation is another flagged repeatedly in fatality-review literature as a marker of substantially elevated future lethality risk. Access to a firearm compounds every other risk factor rather than sitting alongside them independently — a documented history of threats with a gun already in the house is a different risk profile than the same threats made without one. None of these are secret or obscure findings; they're the basis for tools like the Lethality Assessment Program screening instrument some police departments now use at domestic-violence 911 calls specifically to flag which situations are approaching this threshold before the second death happens, not after.

That existing warning-sign research is why murder-suicide sits at an uncomfortable intersection for prevention efforts: the intimate-partner subset of it isn't a mysterious, unpredictable event in the way mass violence directed at strangers often gets discussed. Fatality review boards — multidisciplinary panels that many states convene specifically to examine domestic-violence deaths after the fact — repeatedly find documented contact with police, courts, or victim services in the weeks or months before a fatal event, contact that didn't or couldn't translate into an intervention that reached the firearm itself. The pattern is legible in hindsight in case after case. The harder, unresolved question the research keeps landing on is how to make it legible — and actionable — before the fact, at the specific moment a firearm is still in the house and a separation is still pending, rather than after.

A Category With No Federal Home

The absence of a dedicated federal surveillance system for murder-suicide is itself part of the story. The CDC's National Violent Death Reporting System (NVDRS) — the most detailed government dataset that exists on violent death circumstances, now covering all 50 states — does capture individual homicide and suicide cases with rich contextual detail, including, in many cases, whether a death was part of a multiple-fatality incident. But NVDRS was not built with murder-suicide as its own analytic category the way the Violence Policy Center's news-based tracking treats it, and linking a homicide record to the perpetrator's own subsequent suicide record within the same dataset requires case-by-case cross-referencing rather than a single queryable field. The result is a strange asymmetry: the government dataset built specifically to study violent death in granular detail exists, and could in principle answer exactly how often, where, and under what documented prior circumstances a murder-suicide occurs — but the accounting Americans actually rely on to know how often this happens at all still comes from an advocacy organization manually reading the news.

The Children Left in the Room

The 45 children and teenagers killed and 27 more who survived as witnesses, in just six months of tracked incidents, point to a secondary population the incident-count headline number doesn't capture: the people who live in the house where a murder-suicide happens and are not among either the homicide or suicide totals, but whose lives are permanently reorganized by it anyway. Child welfare and bereavement researchers who study these cases separately from the criminological literature describe a specific, compounding form of trauma — loss of both parents at once in the most common configuration, loss of the surviving family's primary source of housing and income in the same event, and a grief process complicated by the fact that one parent killed the other. Extended family members who step in as caregivers afterward are themselves navigating a loss that doesn't fit standard bereavement support models built around a single death from natural causes or accident. It's a population that shows up in family court records, in school counselor caseloads, and in kinship-care placement statistics far more often than it shows up in any public dataset labeled "murder-suicide."

And Then You're Dead

The second death gets almost none of the coverage the first one does — a line at the bottom of the article, "the suspect was later found dead," as if it settled something rather than compounded it. Nine times a week, somewhere in the country, it was a gun already in the house, a relationship already ending, and warning signs that were there to be read, whether or not anyone in a position to act on them ever did.

That's it. That's the whole thing.

Comments

Loading…

Comments are anonymous and moderated.

Sources