Expert Commentary

Deaths in police custody in the United States: Research review

We summarized studies that look at deaths in police custody from multiple angles, including restraint methods and police force demographics.

Deaths in police custody research George Floyd
Memorial for George Floyd outside the Cup Foods convenience store in Minneapolis in May 2020. (Flickr/Lorie Shaull)

The deaths of black men at the hands of white police officers in recent years have raised a number of questions about the treatment of racial minorities within the criminal justice system, as well as about patterns of arrest-related deaths more generally. Some researchers have called for Congressional-mandated government databases to be more thorough so they can better find patterns in the violent interactions between police and civilians.

The recent death of George Floyd in Minneapolis and the mass protests that followed have pushed the issue back into the national spotlight. Floyd, a black man accused of buying cigarettes with a counterfeit $20 bill, died after a white police officer kneeled on his neck for nearly nine minutes to restrain him. That officer, Derek Chauvin, has been fired and charged with second-degree murder. The three other officers who were involved were fired and charged with aiding and abetting second-degree murder.

In response to the renewed media attention, we have significantly updated this research roundup — last updated in 2016 — with new information. We will continue to update this piece with new research in the weeks to come.

For years, legislators, community leaders and others have wanted to know: How many black Americans have died while being apprehended, arrested or transported by law enforcement officers? And how does that number compare with the number of men and women of other races and ethnicities killed each year in police custody?

No one knows the official answers to those questions. It has been five years since Congress passed the Death in Custody Reporting Act of 2013, which went into effect in December 2014, but federal officials have not yet gathered the data and made it public. Rep. Robert C. “Bobby” Scott of Virginia introduced the legislation, created based on an earlier law that expired in 2006 and had required states to submit quarterly reports on deaths in police custody.

The new law requires the Attorney General to collect from each state as well as all federal law enforcement agencies “information regarding the death of any person who is detained, under arrest, or is in the process of being arrested, is en route to be incarcerated, or is incarcerated at a municipal or county jail, state prison, state-run boot camp prison, boot camp prison that is contracted out by the state, any state or local contract facility, or other local or state correctional facility (including any juvenile facility).”

Earlier this year, two members of Congress — the chairman of the House Judiciary Committee, Jerrold Nadler of New York, and the chairwoman of the House Subcommittee on Crime, Terrorism and Homeland Security, Karen Bass of California — wrote a letter to the Inspector General requesting an investigation into the Department of Justice’s failure to implement the Death in Custody Reporting Act of 2013.

“The United States continues to face a persistent crisis of deaths in custody, the true scope of which remains unknown,” Nadler and Bass wrote.

A 54-page report from the Office of the Inspector General chronicles problems the U.S. Department of Justice has had implementing the law. It notes that state-level data collection “will be delayed until at least FY 2020,” which ends Sept. 30, 2020.

The report, released in December 2018, also points out that the Department of Justice “does not have plans to submit a required report that details results of a study on DCRA [Death in Custody Reporting Act] data. DCRA required that such a report be submitted to Congress no later than 2 years after December 18, 2014.”

In the absence of a government database, several organizations have created their own. For example, Fatal Encounters, founded by former editor and publisher of the Reno News & Review D. Brian Burghart, maintains a searchable database of people who died during interactions with police. The Washington Post’s national database cataloging fatal shootings by police was last updated June 1.

Academic research provides insights but offers an incomplete picture. A study published in The New England Journal of Medicine in 2016 finds there were 222 “legal intervention” deaths in 2013, or cases in which someone was killed by an on-duty law enforcement or other peace officer. The study is based on data from just 17 states, however, and none of the largest states — California, Florida, Illinois, New York, Pennsylvania and Texas — were included.

According to the paper, nearly everyone killed by on-duty officers in those states that year were male and between the ages of 20 and 54 years old. It finds that black people were most likely to die in police custody.

“The rates [of death] were higher among non-Hispanic blacks (0.6 per 100,000 population) and Hispanics (0.3 per 100,000) than among non-Hispanic whites (0.1 per 100,000),” the authors write. “Multiple factors were found to be associated with the circumstances of these events, such as a crisis (e.g., the victim had had a bad argument, had divorce papers served, was laid off, faced foreclosure on a house, or had a court date for a legal problem within 2 weeks before the death), a current mental health problem, and intimate partner violence.”

A newer study, published in the Proceedings of the National Academy of Sciences in 2019, estimates that black men have a 1 in 1,000 chance of being killed by police during their lifetimes. That’s 2.5 times the odds for a non-Hispanic white man, the authors find.

Arrest-related deaths

A string of high-profile deaths of black men in police custody in recent years has raised questions about how black Americans are treated during and immediately after their arrest. But a December 2016 report on arrest-related deaths from the U.S. Bureau of Justice Statistics notes the agency receives data on only a portion of deaths.

“The ARD [Arrest-Related Deaths] program captured about half of the estimated number of justifiable homicides in the United States from 2003 through 2011, excluding 2010,” according to the report. “In general, the incomplete coverage each year was due in part to the unstandardized data collection process across states. However, program coverage increased to a high of 69% in 2011, when the program began to rely more on open information sources to identify potentially eligible deaths.”

The news media and Fatal Encounters project are open information sources that have become key to the Bureau of Justice Statistics’ data collection effort. A technical report the agency released in July 2019 illustrates the importance of these sources. From 2003 to 2009, the report states, the agency’s Arrest-Related Death program identified 375 to 496 arrest-related deaths per year. In 2011, when program officials began collecting data from open information sources, they identified 689 deaths that year.

The new data collection method helped the agency find and track down information on 424 deaths in police custody during a three-month study period, June 1, 2015 through Aug. 31, 2015.  Officials used news media coverage to find the vast majority of deaths – 89%, according to the technical report. Fewer than 50 deaths — 11% — were reported by law enforcement agencies.

More than 42% of arrest-related deaths identified during the study period occurred in five states — California, Florida, Georgia, Ohio and Texas. Of the 424 deaths in total, 63% were categorized as homicides.

The federal government also tracks fatalities in local jails and state and federal prisons through its Deaths in Custody Reporting Program. Suicides accounted for 31% of deaths in local jails from 2000 to 2016, a federal report released earlier this year shows.  About half were due to illness, including heart and liver disease and cancer.

The report reveals white inmates are more likely to die in jail than black or Hispanic inmates. “The mortality rate for white jail inmates in 2016 (240 deaths per 100,000 white inmates) was more than double the rate for black inmates (118 deaths per 100,000 black inmates) and almost triple the rate for Hispanic inmates (87 deaths per 100,000 Hispanic inmates),” it concludes.

Reporting on incidents

Experts involved in analysis of these incidents caution that the numbers can often hide meaningful context, and reporters would be well served to go beneath the surface and ask about how data is collected — and any potential holes or weaknesses in the data. Overall, states have varied in their methods of reporting law enforcement-related incidents of many kinds to the federal government.

For an example of how data, or the lack of it, can matter — and mislead — see the series on prison rape written by David Kaiser for The New York Review of Books. Finally, for news reporters covering individual incidents, context can be crucially important, from the degree to which a neighborhood is a high-crime area, or where assaults on officers are common; to the level of police training to deal with, for example, violent and mentally ill persons; to the precise nature of the incident and whether it involved a suspect threatening public safety at the time of a violent intervention by authorities.

There’s also a substantial body of government and academic research on these issues. Below, we’ve rounded up studies that look at deaths in police custody, the use of different types of restraint and whether changing police force demographics might result in fewer black men dying during interactions with law enforcement.


Whose Death Matters? A Quantitative Analysis of Media Attention to Deaths of Black Americans in Police Confrontations, 2013–2016
Zuckerman, Ethan; et al. International Journal of Communication, September 2019.

Summary: The news media became more likely to cover black men’s fatal encounters with police following the August 2014 death of 18-year-old Michael Brown, fatally shot by a police officer after an altercation in Ferguson, Missouri, this study finds.

A team of researchers, led by media scholar Ethan Zuckerman of MIT, analyzed news coverage of black people who were killed by police or died in police custody in the U.S. from January 2013 through the end of June 2016. Researchers also queried Facebook to gauge the number of times those stories were liked and shared.

In total, the researchers examined news coverage of “343 deaths of unarmed Black Americans at the hands of police, as well as the number of media articles, the sharing of those articles, and the framings used in those articles to investigate differences in coverage before and after Michael Brown’s death.”

“Based on analysis of these data, we found evidence for a significant shift in attention to stories about police killings of unarmed Black men in the wake of Michael Brown’s death, which we considered a ‘key event’ in changing the media framing of these deaths,” Zuckerman and his colleagues write. “In addition to an increased volume of coverage, we saw evidence that police-involved deaths after Michael Brown’s death were significantly more likely to be reported on as part of a larger pattern of police violence against Black citizens, and some evidence that these stories were more likely to be shared on social media.”


“Will More Black Cops Matter? Officer Race and Police‐Involved Homicides of Black Citizens”
Nicholson-Crotty, Sean; Nicholson-Crotty, Jill; Fernandez, Sergio. Public Administration Review, March/April 2017.

Summary: In this study, researchers investigate whether adding more black officers to local police forces would reduce the number of black people dying in police custody. The key takeaway: It depends on the size of the police department and how many black officers it already employs.

For most cities, simply increasing the share of black officers would not be an effective policy solution, the researchers find after examining arrest-related deaths in the 100 most populous cities in the U.S. in 2014 and 2015 and the demographics of local police departments. The analysis is based in part on data collected by The Washington Post and the advocacy organization Mapping Police Violence.

The researchers, from Indiana University, Bloomington and the University of Johannesburg in South Africa, discovered that as the percentage of black officers working in large city grows, so does the number of black men dying in police custody. “In 2015, we see a positive and significant relationship between the percentage of black officers and fatal encounters between police and black citizens in the vast majority of the largest U.S. cities,” they write. “The results also suggest that there is an inflection point at which black officers may become less likely to discriminate against black citizens and more inclined to assume a minority advocacy role or to become neutral enforcers of the law, though the limited number of observations makes it impossible to know for sure.”

The researchers explain that increasing the share of police officers “is only a solution for the small handful of large departments around the country that already have a high percentage of black officers.”

“In places like Ferguson, where blacks are only 11 percent of the police force, efforts to double or even triple the share of the police force that is black would not reach this threshold and may actually be associated with an increase in violent interactions between police and black citizens until the critical mass is achieved,” they write.


Deaths in Police Custody”
Heide, Steffen; Chan, Theodore. The Journal of Forensic and Legal Medicine, February 2016.

Summary: The authors of this study found that it’s difficult to compare deaths in police custody across countries for three reasons: 1) a lack of a uniform definition of the term “custody” 2) country-specific differences in the structures of police custody and 3) variations in the way scholars have designed their studies of the issue.

For example, in Europe, people are generally considered to be in police custody for fewer than 48 hours, the authors explain. But in the U.S., someone is typically considered to be in police custody during the process of arrest through the duration of any incarceration, which can last days to decades. “Some studies only record deaths in the police cell, with suicide as an exclusion criteria in some cases,” write the authors, from the University of Halle-Wittenberg in Germany and the University of California, San Diego. “Other studies also include deaths linked to arrests by the police, up to and including deaths caused by the use of firearms or fatal traffic collisions of police vehicles.”

The authors analyzed academic studies of deaths in police custody in Europe, North America and Australia. Because of drastic differences in the data, the authors concluded they could only make a few broad generalizations. They found that men were more likely than women to die in police custody, for instance. “The male dominance is essentially due to the fact that around the world women are much less frequently taken into police custody than men,” they write.

They note the need for further investigation of “sudden death” incidents in police custody — unexpected deaths often during or immediately after a confrontation with law enforcement in which there is not clear cause of death such as a gunshot wound.


“Homicides by Police: Comparing Counts From the National Violent Death Reporting System, Vital Statistics, and Supplementary Homicide Reports”
Barber, Catherine; et al. The American Journal of Public Health, May 2016, Vol.106, doi: 10.2105/AJPH.2016.303074.

Summary: The report compares how different government databases track homicides by law enforcement officers and finds the National Violent Death Reporting System (NVDRS), run by the Center for Disease Control, the best available because it draws from multiple sources and “captures detailed coded data and rich narratives that describe the precipitating circumstances and incident dynamics for all suicides and homicides occurring in participating states.” The authors recommend the NVDRS be expanded to all 50 states. It currently tracks 32.


“Race, Crime, and the Micro-Ecology of Deadly Force”
Klinger, David; Rosenfeld, Richard; Isom, Daniel; Deckard, Michael. Criminology & Public Policy, February 2016, Volume 15, doi: 10.1111/1745-9133.12174.

Research Summary: “Limitations in data and research on the use of firearms by police officers in the United States preclude sound understanding of the determinants of deadly force in police work. The current study addresses these limitations with detailed case attributes and a microspatial analysis of police shootings in St. Louis, MO, between 2003 and 2012. The results indicate that neither the racial composition of neighborhoods nor their level of economic disadvantage directly increase the frequency of police shootings, whereas levels of violent crime do—but only to a point. Police shootings are less frequent in areas with the highest levels of criminal violence than in those with midlevels of violence. We offer a provisional interpretation of these results and call for replications in other settings.”


“Pattern of law enforcement–related injuries in the United States”
Chang, David C.; et al. Journal of Trauma and Acute Care Surgery, Volume 80(6), June 2016, doi: 10.1097/TA.0000000000001000

Conclusion: “The majority of law enforcement–related injuries are among white or black young men. Hispanic patients are more likely to be injured by a firearm than struck. When injured by firearm, white and black patients are more likely to die compared with Hispanic patients. Unfortunately, data about these injuries are scattered across multiple data systems. A uniform national system to aggregate these data sources is needed to better understand the scope of the problem, for both law enforcement personnel and civilians.”


“Arrest-Related Deaths, 2003-2009: Statistical Tables”
Burch, Andrea M. U.S. Department of Justice, Bureau of Justice Statistics, November 2011.

Introduction: “From 2003 through 2009, a total of 4,813 deaths were reported to the Bureau of Justice Statistics’ (BJS) Arrest-Related Deaths (ARD) program. Of these, about 6 in 10 deaths (2,931) were classified as homicide by law enforcement personnel, and 4 in 10 (1,882) were attributed to other manners of death. Suicide and death by intoxication each accounted for 11 percent of reported arrest-related deaths, accidental injury for 6 percent, and natural causes for 5 percent (figure 1). Deaths with manners classified as undetermined or those in which manners were unknown represented about 6 percent of reported arrest-related deaths.”


“Unexpected Arrest-Related Deaths in America: 12 Months of Open Source Surveillance”
Ho, Jeffrey D. Western Journal of Emergency Medicine, May 2009, Volume X, No. 2.

Abstract: “Introduction: Sudden, unexpected arrest-related death (ARD) has been associated with drug abuse, extreme delirium or certain police practices. There is insufficient surveillance and causation data available. We report 12 months of surveillance data using a novel data collection methodology. Methods: We used an open-source, prospective method to collect 12 consecutive months of data, including demographics, behavior, illicit substance use, control methods used, and time of collapse after law enforcement contact. Descriptive analysis and chi-square testing were applied. Results: There were 162 ARD events reported that met inclusion criteria. The majority were male with mean age 36 years, and involved bizarre, agitated behavior and reports of drug abuse just prior to death. Law enforcement control techniques included none (14 percent); empty-hand techniques (69 percent); intermediate weapons such as Taser device, impact weapon or chemical irritant spray (52 percent); and deadly force (12 percent). Time from contact to subject collapse included instantaneous (13 percent), within the first hour (53 percent) and 1-48 hours (35 percent). Significant collapse time associations occurred with the use of certain intermediate weapons.”


“On the Problems and Promise of Research on Lethal Police Violence”
Klinger, David A. Homicide Studies, 2012, 16(1) 78-96, doi: 10.1177/1088767911430861.

Abstract: “We presently have little information about how frequently police officers shoot citizens or are involved in any sort of interaction in which citizens die. Despite this, however, researchers persist in using the limited data available on fatal police violence in various sorts of analyses. The current article outlines the liabilities in available counts of fatal police action, describes some of the problems posed by using such data, discusses why counting citizens killed by police bullets is not a sound way to measure deadly force, and offers some ideas for improving measurement of the use of deadly force and other police actions that lead to the death of citizens.”


“Arrest-Related Deaths Program Assessment: Technical Report”
Banks, Duren; Couzens, Lance; Blanton, Caroline; Cribb, Devon. Bureau of Justice Statistics, U.S. Department of Justice, March 2015, NCJ 248543.

Executive summary: “The Bureau of Justice Statistics (BJS) designed the Arrest-Related Deaths (ARD) program to be a census of all deaths that occur during the process of arrest in the United States…. We found that over the study period from 2003 through 2009 and 2011, the ARD program captured, at best, 49 percent of all law enforcement homicides in the United States. The lower bound of ARD program coverage was estimated to be 36 percent. These findings indicate that the current ARD program methodology does not allow a census of all law enforcement homicides in the United States. The ARD program captured approximately 49 percent of law enforcement homicides, while the SHR captured 46 percent. An estimated 28 percent of the law enforcement homicides in the United States are not captured by either system. However, the methodology for identifying ARD cases has changed over the observation period. In 2011, the ARD program was estimated to cover between 59 percent and 69 percent of all law enforcement homicides in the United States, depending on the estimation method used. While this coverage estimate still does not result in a census, it does suggest improvements over time in the overall approach to identifying law enforcement homicides and reporting them to the ARD program.


“Can TASER Electronic Control Devices Cause Cardiac Arrest?”
Kroll, Mark W.; et al. Circulation, 2014. 10.1161/circulationaha.113.004401.

Introduction: “The electronic control device (ECD) has gained widespread acceptance as the force option for law enforcement because of its dramatic reduction in both suspect and officer injury. At the same time, advocacy groups post statements on the Internet listing the hundreds of arrest-related deaths after ECD use with the implication that the ECD involvement was causal. Studies covering a total of >48 000 forceful arrests have consistently found suspect injury rate reductions of ≈65 percent. Of the 250 000 annual ECD field uses in the United States, only 1 in 4000 is involved in an arrest-related death.” The research looks at 12 cases of cardiace after ECD application. Results: ” These data suggest that the threshold of factual evidence for blaming a cardiac arrest on an ECD should be set very high. The published case reports have not met that threshold.”


“The Effect of the Prone Maximal Restraint Position with and without
Weight force on Cardiac Output and other Hemodynamic Measures”

Savaser, Davut J.; et al. Journal of Forensic and Legal Medicine, August 2013, Vol. 30

Abstract: “Background: The prone maximal restraint (PMR) position has been used by law enforcement and emergency care personnel to restrain acutely combative or agitated individual. The position places the subject prone with wrists handcuffed behind the back and secured to the ankles. Prior work has indicated a reduction in inferior vena cava (IVC) diameter associated with this position when weight force is applied to the back. It is therefore possible that this position can negatively impact hemodynamic stability. Objectives: We sought to measure the impact of PMR with and without weight force on measures of cardiac function including vital signs, oxygenation, stroke volume (SV), IVC diameter, cardiac output (CO) and cardiac index (CI). Conclusions: PMR with and without weight force did not result in any changes in CO or other evidence of cardiovascular or hemodynamic compromise.”


“Effect of Position and Weight Force on Inferior Vena Cava Diameter: Implications for Arrest-related Death”
Ho, Jeffrey D.; et al. Forensic Science International, 2011. doi: 10.1016/j.forsciint.2011.07.001.

Abstract: “Introduction: The physiology of many sudden, unexpected arrest-related deaths (ARDs) proximate to restraint has not been elucidated. A sudden decrease in central venous return during restraint procedures could be physiologically detrimental. The impact of body position and applied weight force on central venous return has not been previously studied. In this study, we use ultrasound to measure the size of the inferior vena cava (IVC) as a surrogate of central venous return in the standing position, prone position, and with weight force applied to the thorax in the prone position…. Conclusions: The physiology involved in many sudden, unexpected ARDs has not been elucidated. However, in our study, we found a significant decrease in IVC diameter with weight force compression to the upper thorax when the subject was in the prone position. This may have implications for the tactics of restraint to aid in the prevention of sudden, unexpected ARD cases.”


“Evaluation of the Ventilatory Effects of the Prone Maximum Restraint (PMR) Position on Obese Human Subjects”
Sloane, Christian; et al. Forensic Science International, April 2014, 237. doi: 10.1016/j.forsciint.2014.01.017.

Abstract: “The study sought to determine the physiologic effects of the prone maximum restraint (PMR) position in obese subjects after intense exercise. We designed an experimental, randomized, cross-over trial in human subjects conducted at a university exercise physiology laboratory. Ten otherwise healthy, obese (BMI > 30) subjects performed a period of heavy exertion on a cycling ergometer to 85 percent of maximum heart rate, and then were placed in one of three positions in random order for 15 min: (1) seated with hands behind the back, (2) prone with arms to the sides, (3) PMR position. While in each position, mean arterial blood pressure (MAP), heart rate (HR), minute ventilation (VE), oxygen saturation (SaO2), and end tidal CO2 (etCO2) were measured every 5 min. There were no significant differences identified between the three positions in MAP, HR, VE, or O2s at at any time period. There was a slight increase in heart rate at 15 min in the PMR position over the prone position (95 vs. 87). There was a decrease in end tidal CO2 at 15 min in the PMR over the prone position (32 mmHg vs. 35 mmHg). In addition, there was no evidence of hypoxia or hypoventilation during any of the monitored 15 min position periods. Conclusion: In this small study of obese subjects, there were no clinically significant differences in the cardiovascular and respiratorymeasures comparing seated, prone, and PMR position following exertion.”


“Excited Delirium Syndrome (EXDS): Defining Based on a Review of the Literature”
Vilke, Gary M.; et al. Clinical Reviews, February 2011.

Abstract: “Patients present to police, emergency medical services, and the emergency department with aggressive behavior, altered sensorium, and a host of other signs that may include hyperthermia, ‘superhuman’ strength, diaphoresis, and lack of willingness to yield to overwhelming force. A certain percentage of these individuals will go on to expire from a sudden cardiac arrest and death, despite optimal therapy. Traditionally, the forensic community would often classify these as ‘excited delirium’ deaths. Objectives: This article will review selected examples of the literature on this topic to determine if it is definable as a discrete medical entity, has a recognizable history, epidemiology, clinical presentation, pathophysiology, and treatment recommendations. Discussion: Excited delirium syndrome is characterized by delirium, agitation, acidosis, and hyperadrenergic autonomic dysfunction, typically in the setting of acute-on-chronic drug abuse or serious mental illness or a combination of both. Conclusions: Based upon available evidence, it is the consensus of an American College of Emergency Physicians Task Force that Excited Delirium Syndrome is a real syndrome with uncertain, likely multiple, etiologies.”


This image was obtained from the Flickr account of Lorie Shaull and is being used under a Creative Commons license. No changes were made.

About the Authors