Officer Injury and Death: What 30 Years of FBI LEOKA Data Reveals

A review of the Law Enforcement Officers Killed and Assaulted program, the long-run trends, the circumstances that most threaten officers, and the training implications.

The FBI's Law Enforcement Officers Killed and Assaulted (LEOKA) program is the closest thing the United States has to a continuous, decades-long record of the dangers of police work. It is also one of the most misread datasets in the field, because the headline number, officers feloniously killed, is small enough to swing sharply year to year, and the underlying assault data is large but unevenly reported. This article works through what LEOKA can and cannot support, with an emphasis on the circumstances that recur and what they imply for training.

What LEOKA measures, and how well

LEOKA has two distinct components. The first is an annual count of law enforcement employees, sworn and civilian, reported by participating agencies as a staffing snapshot each fall.¹ The second, and the one of interest here, is incident data: officers feloniously killed (intentional homicides), officers accidentally killed in the line of duty, and officers assaulted.¹ Assault records are broken down by the type of call the officer was responding to, the weapon used, the officer's assignment, and the time of day.¹

The employee counts are reliable. The incident counts require care. Reporting coverage has been unstable, particularly through the FBI's transition away from the older Summary Reporting System: the number of agencies submitting LEOKA assault data has run in the range of roughly 5,000 to 6,000 in recent years, far short of the roughly 19,000 agencies nationally.¹ Known undercounting is severe enough that in one recent year the assault and accidental-death tallies were obviously incomplete relative to officer deaths documented elsewhere.¹

The honest framing, then, is this. Felonious-killing counts are close to a complete national figure, because a line-of-duty homicide of an officer is almost always documented. Assault counts are a large but partial sample. Both are useful; they are useful in different ways.

The 30-year arc: a long decline, with a reversal

Across three decades, the dominant LEOKA story is a substantial long-run decline in officers feloniously killed. The 1990s were the high-danger period of the modern era. Felonious killings then fell roughly 20 percent from the 1990s to the 2000s, and roughly another 25 percent from the 2000s to the 2010s.² The decline was not driven by a single cause. Analysts attribute it to a combination of better vehicle safety, dramatic improvements in emergency trauma care that converted what would once have been fatal wounds into survivable ones, and the spread of mandatory bullet-resistant vest policies.²

That long decline then partially reversed. Beginning in the early 2010s, accidental line-of-duty deaths kept falling while felonious killings ticked back upward, narrowing a gap that had been wide for decades.² The year-to-year figures in the 2020s show how volatile a small number can be. In the first nine months of 2024, 54 officers were feloniously killed, a 12.5 percent increase over the 48 killed in the same period of 2023.³ Then in 2025 the figure fell to 53 officers feloniously killed for the full year, reported as the lowest annual total since 2020 and a 17.2 percent decrease from 2024.⁴

The methodological point matters more than any single year. A swing from 48 to 54 to 53 is a swing of a handful of human beings. It is real, and every one of those deaths is a tragedy, but it is not a statistically stable trend, and it should not be presented as one. The defensible reading of LEOKA fatality data is the multi-decade arc, not the annual headline.

The circumstances that most threaten officers

The training value of LEOKA lies less in the death counts than in the circumstance breakdowns, because circumstances recur and can be rehearsed.

Disturbance calls are the most common assault context. When LEOKA assaults are sorted by call type, disturbance calls rank first.⁵ The LEOKA definition of a disturbance call is broad: it covers breaches of the peace including disorderly persons, fights, persons under the influence, verbal altercations, and it explicitly includes domestic violence calls and reports of dangerous individuals such as a person carrying a gun in public.⁵ This is the single most important circumstance finding in the dataset. The encounters that most often produce an assault on an officer are not, in the main, dramatic ambushes. They are the high-volume, emotionally charged, low-information calls that fill a patrol shift.

Arrest and custody situations rank high. Assaults are also concentrated in the moment of taking a subject into custody.⁵ This is intuitive: the point at which an officer asserts physical control is the point at which a subject who is going to resist will do so.

Ambushes are rare but disproportionately lethal. An ambush, defined as an unexpected assault resulting from premeditated design by the offender, is among the least frequent call types in the assault data.⁵ But because the offender controls the timing and has decided in advance to attack, an ambush carries a far higher chance of officer death per incident than its frequency suggests. Rarity and lethality point in opposite directions here, and training has to account for both.

Most assaults involve unarmed offenders. A consistent and somewhat counterintuitive LEOKA finding is that the majority of officer assaults are committed with "personal weapons," the offender's hands, fists, and feet, rather than firearms or edged weapons.⁵ Firearm and knife assaults are less frequent. The data also shows that gun and knife assaults without injury exceed those with injury, which is consistent with weapons being displayed or threatened more often than actually used to wound.⁵ The practical reading: the typical officer assault is a close-quarters physical struggle, often during a disturbance call or an arrest, not a shooting.

Traffic stops deserve a specific note because of their place in officer-safety culture. Among officers feloniously killed during traffic stops, the deaths concentrated in earlier decades, with roughly 40 percent of such killings in the 1990s, 32 percent in the 2000s, and 24 percent in the 2010s.⁶ Traffic stops remain a genuine hazard, but the long-run trend within that category has been downward, consistent with the overall felonious-killing decline.

The connection to communication and de-escalation

Here intellectual honesty requires precision. LEOKA is a descriptive dataset. It records where and how officers were assaulted. It does not, by itself, prove that any particular training would have prevented any particular assault. A claim that "de-escalation reduces officer injury" cannot be read directly off LEOKA.

What LEOKA does support is narrower and still useful. It identifies that officer assaults concentrate in two phases: the verbal and contact phase of disturbance calls, and the moment of taking custody. Both of those phases are precisely the phases that communication and control training addresses. The disturbance call is, by definition, an encounter still in its verbal stage, where tone, positioning, and pace are the officer's working tools. The custody moment is a transition that can be planned, sequenced, and rehearsed.

There is also a growing research literature directly examining the relationship between officer training and felonious officer deaths, including work specifically on traffic-stop fatalities and the role of education and retraining.⁶ That literature is the appropriate place to look for causal claims; LEOKA is the place to look for the pattern that tells you where to aim. The defensible synthesis is this: officer-safety training and de-escalation training are not opposing priorities. The same data that shows officers are most often assaulted on disturbance calls also shows those calls are still in their communication phase when the assault occurs. Skill in that phase is an officer-safety skill.

What this means for your agency's training

LEOKA, read with discipline, supports a clear set of conclusions.

The long-run decline in officer deaths is real and is partly the product of deliberate interventions: vests, vehicle safety, trauma care. That is a proof of concept. Officer fatality risk responds to systematic effort, and training is one more lever in that system.

The circumstance data tells agencies where the risk concentrates. Officers are assaulted most often on disturbance calls and during custody transitions, most often by unarmed subjects, in close quarters. A training program that spends its scenario time on those high-frequency contexts is matching its effort to the data. A program that spends most of its time on the rare ambush, while neglecting the disturbance call, is training for the exception and underpreparing for the norm. Both belong in the curriculum; the proportions should reflect the frequencies.

The annual fatality headline should be handled carefully in any agency communication. A 12 percent year-over-year change in a number near 50 is not a trend, and presenting it as one erodes credibility with anyone who knows the data.

And LEOKA's own coverage gaps argue, once again, for rigorous internal tracking. An agency that records its own officer assaults by call type and phase can see its own risk pattern with a clarity LEOKA's partial national sample cannot offer.

CodeBlu's scenario-based model is well suited to rehearsing the high-frequency assault contexts the data identifies: the disturbance call still in its verbal phase, and the planned custody transition. The claim made here is bounded. LEOKA shows where assaults happen; it does not certify any program's effect on assault rates. Agencies should rehearse the contexts the data flags and measure their own outcomes.

Recommended CodeBlu scenarios this article supports

1. Disturbance call, verbal phase: the highest-frequency assault context, rehearsed while the encounter is still resolvable through communication and positioning.
2. Planned custody transition: sequencing and communicating the move to physical control to reduce the assault risk concentrated at that moment.
3. Close-quarters unarmed resistance: reflecting the LEOKA finding that most officer assaults involve personal weapons in a physical struggle, not a shooting.
4. Ambush awareness and recovery: rehearsing the rare but high-lethality scenario without letting it crowd out the high-frequency ones.
5. Traffic stop approach and positioning: addressing a genuine, if declining, fatality context with deliberate approach planning.

---

Footnotes

1. FBI, "Law Enforcement Officers Killed and Assaulted (LEOKA)." https://www.fbi.gov/how-we-can-help-you/more-fbi-services-and-information/ucr/leoka (accessed May 2026). Data-quality and reporting-coverage observations also drawn from "Decoding FBI Crime Data, Chapter 7: LEOKA." https://ucrbook.com/leoka.html ↩ ↩² ↩³ ↩⁴ ↩⁵

2. "Decoding FBI Crime Data, Chapter 7: LEOKA." https://ucrbook.com/leoka.html ; long-run decline and contributing factors (vehicle safety, trauma care, vest policies) summarized from FBI LEOKA historical reporting. ↩ ↩² ↩³

3. FBI, "Statistics on Law Enforcement Officer Deaths in the Line of Duty from January through September 2024." https://le.fbi.gov/cjis-division/cjis-link/statistics-on-law-enforcement-officer-deaths-in-the-line-of-duty-from-january-through-september-2024 ↩

4. 2025 full-year felonious-killing figure (53 officers, lowest since 2020) reported from FBI preliminary LEOKA data. https://www.ammoland.com/2026/04/police-murders-fell-in-2025-down-again-in-early-2026/ ↩

5. FBI LEOKA assault breakdowns by call type and weapon, as compiled in "Decoding FBI Crime Data, Chapter 7: LEOKA." https://ucrbook.com/leoka.html ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

6. "Education, (re)training, and traffic stops: Felonious law enforcement officer deaths in the United States," ScienceDirect, 2023. https://www.sciencedirect.com/science/article/pii/S1756061623000447 ↩ ↩²