Inside the Perimeter: New Databases on Lethal Violence at Work, in Hospitals, and on Campus
Across three new U.S. homicide databases released by the Violence Prevention Project, one finding dominates. The killer already had a way in.
In 94 percent of workplace insider homicides, the killer walked through the front door on credentials he or she already had. In 95 percent of campus homicides, the perpetrator faced no access restrictions. In medical settings, almost all the killing happens inside buildings designed to stay open.
That is the security problem in one sentence: The people most likely to commit lethal violence at your facility already have a way in.
The Violence Prevention Project, a nonprofit research center, has released three new homicide databases covering 25 years of U.S. data that document this pattern in granular detail. The Workplace Homicide Database tracks 653 incidents from 2000 to 2025 in which the perpetrator was a current or former employee of the workplace. This is a deliberately narrow focus. Most U.S. workplace killings are robberies committed by strangers, and that broader category has declined sharply since the mid-1990s. The insider category has not.
The Medical Homicide Database covers 280 homicides inside U.S. medical settings, with perpetrators ranging from current patients to partners of staff to clinicians themselves. The Higher Education Homicide Database covers, by coincidence, also 280 homicides at U.S. colleges and universities, with perpetrators including current and former students, faculty, staff, partners, and outsiders.
A small subset of cases in each database meets the federal definition of an active shooter event: one or more people actively engaged in killing or attempting to kill others in a populated area. These make up 22 percent of workplace insider cases, 12 percent of medical cases, and 7 percent of campus cases. Active shooter incidents are disproportionately lethal, more often planned than impulsive, and follow patterns that diverge from the broader homicide data in ways that matter for prevention.
All three databases, plus additional ones focused on K-12 schools and houses of worship, are free and searchable online.
Most U.S. workplace killings are robberies committed by strangers, and that broader category has declined sharply since the mid-1990s. The insider category has not.
The Discipline Window
Across the three datasets, one of the most consistent risk indicators is recent workplace conflict, and the pattern intensifies in active shooter cases. Among all workplace insider perpetrators, nearly a quarter had received a reprimand shortly before the attack. Among active shooters in the workplace, the rate climbs to 38 percent. Former employees commit 21 percent of workplace insider homicides overall, but 33 percent of workplace active shooter incidents. The post-termination period is the single highest-risk window for the most lethal kind of insider attack.
Higher education shows the same pattern in a different vocabulary. Among faculty and staff perpetrators (a group that includes three of the dataset’s 20 active shooters), professional grievances dominate. The 2010 University of Alabama in Huntsville shooting followed a tenure denial. The 2010 Ohio State shooting followed a negative performance review. A failed dissertation defense, a passed-over promotion, a forced exit: These stressors look different from the trauma histories that drive violence in K-12 schools. The typical school shooter is younger, carrying childhood trauma that built over years. The faculty perpetrator is older, deep into a career, and reacting to a sudden loss of status. One is a life coming apart from the start. The other is a life’s work taken away.
The operational implication is straightforward. Discipline, termination, and adverse personnel decisions are security events as much as they are HR events. When they happen, how they are handled, and what monitoring follows are all safety decisions.
The Domestic Violence Spillover
In medical and higher education settings, intimate partner violence drives a substantial share of lethal incidents. Domestic incidents account for more than a third of medical homicides. On higher education campuses, intimate partner violence drives nearly one homicide in four. Most victims are women. The attackers are former or current partners. The locations are predictable: a hospital where she works, a dormitory where she lives, or a parking lot she walks through twice a day.
Perimeter security does not address this problem. The partner knows the target’s schedule, building, and entrance, and walks in without raising an eyebrow.
What helps is screening, information sharing, and protective planning. This can include private screening for intimate partner violence during medical intake, coordination between residential life and behavioral intervention teams on campus, and workplace awareness when an employee discloses a safety concern at home. None of this is a simple security technology purchase.
The Suicide Overlap
The third overriding pattern is the relationship between homicide and suicide. One in five workplace insider homicide perpetrators dies by suicide at the scene. Among workplace active shooters, the rate climbs to one in two. On higher education campuses, half of all active shooters die by suicide. A quarter of campus domestic violence perpetrators do.
Violence prevention and suicide prevention overlap heavily. A distressed, often hopeless person who concludes that someone at his or her place of work, hospital, or campus is responsible for his or her pain may turn that conclusion outward, inward, or both. Programs treating the two types of prevention as separate functions miss the link.
What helps is screening, information sharing, and protective planning.
What This Means for Security Practitioners
For security directors, the databases offer two practical uses.
First, they direct attention toward internal risk infrastructure. The most useful single tool at most facilities is a confidential, nonpunitive reporting channel where staff can flag concerning behavior, paired with a cross-functional threat assessment team that can manage cases over time. A new camera array ranks well below it. Half of insider perpetrators show warning signs before the attack. Fewer than one in 10 make explicit threats. Reporting prompts the need to capture the broader behavioral cues, such as fixation on a grievance, sudden personality change, paranoid statements, or social withdrawal. Explicit threats alone would catch a minority of cases.
Second, the data support local benchmarking. A hospital security director or campus public safety chief can pull his or her setting from the dataset and compare local incident history and current policy against the national picture. The recommendations attached to each dataset are specific: pretermination risk screening, layered access design for medical buildings, dormitory protocols for campuses, employee assistance programs promoted as protective infrastructure rather than a benefits afterthought, and so on.
These databases will not stop the next attack, and they cannot substitute for a working threat assessment program. But they offer the most detailed open-source picture available of insider violence in these settings and show where prevention efforts actually pay off. For security leaders making the case for investment in behavioral programs alongside hardware, the evidence base is worth downloading.
James Densley, PhD, is professor and department chair of criminology and criminal justice at Metropolitan State University, Minnesota.
Jillian Peterson, PhD, is professor of criminology and criminal justice at Hamline University in Minnesota.
Together, they cofounded the Violence Prevention Project Research Center. Their book is The Violence Project: How to Stop a Mass Shooting Epidemic.
