Monitoring Deaths in Police Custody: Public Health Can and Must Do Better

Police accountability and transparency have been key demands of the Black Lives Matter protests that first took to the streets in 2014 and continued in 2020. However, six years after a police officer killed Michael Brown in Ferguson, Missouri, the United States does not even have an official nationwide system that documents all deaths in police custody, much less one that provides detailed and timely data to the public. This lack of transparency can, in part, be seen as a failure of public health infrastructure. ¹ Mortality surveillance is a key function of public health, one that entails enumerating deaths, describing the circumstances under which deaths occur, and using these data to inform systemic changes that can prevent future harm. The American Public Health Association adopted an official policy statement in 2018 that, among other recommendations, called on public health agencies to collect better data on police killings. ² We build on this call to action by describing the processes that lead to undercounting of deaths in custody in national mortality data and the paucity of critical details about circumstances that led to those deaths. We outline three suggestions that would improve data collection on deaths in custody and strengthen efforts toward police accountability.

CURRENT STATE OF MONITORING DEATHS IN POLICE CUSTODY

Under federal law, the Deaths in Custody Reporting Act (DCRA) defines a “death in custody” as the death of a civilian who was being detained, pursued for arrest, or transported by law enforcement. (The definition also includes death during incarceration, which is beyond the scope of this editorial.) Whereas the term “legal intervention” connotes that use of physical force by an officer caused the death, “death in custody” is a broader category that includes legal intervention as well as deaths whose link to use of force is unclear or contested (e.g., deaths following the use of a Taser, whose link to mortality is often disputed) and deaths that occurred in the absence of use of force (e.g., the decedent was struck by a car while pursued by law enforcement).^3,4

The two national public health data sets that record certain US deaths in police custody are the National Vital Statistics System (NVSS) and the National Violent Deaths Reporting System (NVDRS), which we describe in detail. Beyond the realm of public health, several other national efforts aim to collect data on deaths in custody. The US Department of Justice (DOJ) administers two systems—the Arrest-Related Deaths Program (enabled by the DCRA) and the Supplementary Homicide Report—that rely on voluntary reporting; when assessed by the DOJ, both systems counted fewer than half of deaths in custody. ⁵ Various nongovernmental organizations also collect data on deaths in custody by compiling news media reports, an approach that has been previously assessed to capture more than 90% of police killings, ⁶ although their ability to capture other deaths in custody (i.e., deaths without a clear link to use of force) has not been assessed. Ongoing nongovernmental efforts include, among others, the Washington Post’s Police Shootings Database, ⁷ which only records fatal shootings, and Fatal Encounters, which reports a broader set of deaths in custody, including those that did not involve use of force. Finally, in 2016, the DOJ redesigned its Arrest-Related Deaths Program to incorporate news media sources along with voluntary law enforcement reporting. ⁸ However, it appears that the new program never went into effect during President Trump’s administration. ⁹

Investigating DEATHS IN POLICE CUSTODY

Public health documentation of deaths in police custody begins with a death investigation by a coroner or medical examiner (CME). A coroner is an elected county-level official who typically has little medical training, whereas a medical examiner is a physician who is appointed rather than elected. CMEs conduct or oversee autopsies, write narrative reports about the circumstances that led to death, and fill out the cause-of-death section on death certificates. These documents serve as the raw data that are later processed in state and national vital statistics systems.

Several issues with CME practices lead to poor data quality regarding deaths in police custody. The net effect of these practices is that deaths are undercounted and causal links between use of force and death are obscured. First, when filling out death certificates for individuals killed by police, CMEs often fail to indicate police involvement, even when they correctly describe police involvement in separate narrative reports. ¹⁰ This practice leads to undercounting in death certificate–based vital statistics systems. Second, a more complex set of issues arises for nonfirearm deaths in custody, such as death after Taser shocks, chokeholds, prone restraint (i.e., holding a civilian face down on the ground), or chemical restraint (e.g., forcible injection with ketamine at the direction of police). Autopsies after such deaths in custody often yield inconclusive results, and the CME must base determinations on other evidence, which may consist solely of officer testimony. In such instances, the medical cause of death is often unclear, and the manner of death (particularly, whether the death was an accident versus homicide) may be contested. ¹¹ These nonfirearm deaths in police custody are rarely reported as homicides, even when they follow use of force, and the manner of death is often classified as accidental or undetermined.^6,12 Reporting practices regarding the manner and cause of death are often idiosyncratic and can be influenced by pressure that police and other government officials exert on death investigators.^13,14 Additionally, much of the forensics research that informs CME determinations regarding so-called “sudden deaths in police custody” involves conflict of interest, such as funding from Axon (formerly called Taser International) and funding from city governments in the context of wrongful death lawsuits against police. ¹⁵ Much of this research has involved attributing deaths in custody to a contested medical condition called “excited delirium” rather than alternative explanations, such as positional asphyxia, for which police would be held responsible. ¹⁶

THE NATIONAL VITAL STATISTICS SYSTEM

The NVSS is maintained by the Centers for Disease Control and Prevention (CDC) through a compact with state governments and contains data on virtually all deaths in the United States. Police killings have been identifiable in the NVSS since it adopted the sixth revision of the International Classification of Diseases (ICD) in 1949. The ICD codes for police killings fall under the category of “legal intervention,” defined as: “Injuries inflicted by the police or other law-enforcing agents, including military on duty, in the course of arresting or attempting to arrest lawbreakers, suppressing disturbances, maintaining order, and other legal action” (ICD, 10th edition, Geneva, Switzerland, 2010). Other deaths in custody (i.e., those for which the CME does not determine that use of force caused the death) are not considered legal intervention and are therefore not identifiable in the NVSS. The system’s nationwide data collection over a long historical period allows for comparisons of legal intervention mortality rates between locations and over time. However, ascertainment for legal intervention in the NVSS is poor: one study found that nationally, 55% of deaths that met the criteria for legal intervention were misclassified in 2015 and instead were typically reported as assault-related injuries (i.e., the same category used for homicides perpetrated by civilians). ⁶ Whereas some states correctly reported more than 75% of legal intervention deaths, other states reported 0%. ⁶ As described previously, underreporting is largely the result of CMEs failing to indicate police involvement on the death certificate, particularly in the text field labeled “Describe how injury occurred.” The National Center for Health Statistics uses the text fields to assign an ICD code and cannot assign legal intervention if sufficient data do not exist. NVSS misclassification rates are particularly high for legal intervention that does not involve a gunshot wound. ⁶

The NVSS has several shortcomings in addition to under-ascertainment of legal intervention. One is the lack of detailed data on the circumstances of the death—beyond reporting the mechanism of death (specific ICD codes describe whether the legal intervention injury was inflicted by a firearm, blunt object, and certain other subcategories), there are no other details about the events that led to the death. Additionally, there is a long lag between a death and the availability of that death in NVSS mortality data—the data for a given year are typically only available toward the end of the next calendar year.

THE NATIONAL VIOLENT DEATH REPORTING SYSTEM

The NVDRS is a state-level surveillance system designed to collect details of circumstances under which violent deaths occur. Started in 2002, it gradually expanded to include all 50 states in late 2018. The NVDRS draws on a broad set of data sources, including death certificates, medical examiner reports, and law enforcement records, and it includes a wide array of variables such as precise location of an incident, types of weapons used, and the nature of the “victim-suspect” relationship. ¹⁷ Previous research suggests that underreporting of police killings is much lower in the NVDRS compared with the NVSS, ¹⁸ but to our knowledge, there have been no efforts to formally quantify NVDRS underreporting by comparing its counts to those in more comprehensive news media–based data sets. Moreover, the NVDRS—by design—excludes nonfirearm deaths that are ruled accidental, ¹⁷ which means that many deaths in custody with nonfirearm injury mechanisms are not captured. Finally, there is a long lag for data availability. The first year in which NVDRS data included all 50 states was in 2019, and the CDC anticipates that these data will not be available to researchers until late 2021.

RECOMMENDATIONS

We offer the following three recommendations to improve data collection on deaths in custody and strengthen efforts toward police accountability:

• 1.improve data collection and reporting practices,
• 2.establish mortality review committees for deaths in custody, and
• 3.reform death investigations.

Improve Data Collection and Reporting

The COVID-19 pandemic has shown that the CDC, state health departments, and local health departments can create online dashboards that provide the public with timely, disaggregated data on mortality. To our knowledge, the only analogous dashboards for deaths in custody are maintained, not by health departments, but by the state attorney general of California. ¹⁹ (Texas also maintained a similar website, but it was inconsistently available in 2020.) The California deaths-in-custody website provides individual-level data that include the responsible agency along with decedents’ demographics and cause and manner of death. Public health agencies at all levels can follow this reporting model. Additionally, state health departments can explore adding legal intervention to the state list of notifiable conditions in jurisdictions where those lists are permitted by law to include injuries, which may improve timeliness of reporting. ¹ Finally, the CDC should promote a “death in custody” checkbox on the standard US death certificate (the model on which each state bases its own death certificate), which is also a recommendation of the National Association of Medical Examiners, ²⁰ to allow for easier identification of these deaths in vital statistics mortality data. NVDRS data collection procedures can be revised to include records flagged as deaths in custody.

Review Committees for Deaths in Custody

Health departments across the United States currently use “mortality review committees” to assess preventable deaths with complex social and medical causes. Although maternal mortality review committees are the most prevalent and have the longest history, ²¹ the model has also been extended to address infant and child mortality and fatal drug overdose.^22,23 These committees meet regularly and include death investigators, forensic pathologists, clinicians of various specialties, public health officials, and members of relevant community organizations. The committee process entails reviewing all available information on the context of a death and then reaching agreement about the medical and social causes that led to the death. Committees release reports that address common, recurring themes from their mortality reviews along with recommendations to prevent further deaths. In New York City, for example, the Maternal Mortality Review Committee releases reports to the New York City Council. In some cases, the review process may also lead a participating death investigator to change the cause or manner of death determination. The review committee model should be extended to include deaths in police custody. These committees can explore not only the role of police practices that can lead to death, but also the systemic shortcomings of social services, mental health treatment, and societal responses to drug use that are often involved in deaths in custody.

Reform Death Investigations

CMEs have close working relationships with police, on whom they rely for access to crime scenes and evidence when investigating homicides, suicides, and drug overdoses. This relationship between police and death investigators can give rise to conflicts of interest in typical cases and even more so in particular California counties, where the sheriff-coroner position is combined into a single role. The National Association of Medical Examiners has called for medical examiners from an outside jurisdiction to investigate deaths in police custody. ¹⁹ Additionally, California, in particular, should continue its process of separating death investigations from sheriff’s office investigations. Death investigator independence is essential to determining whether police use of force contributed to a death in custody. Finally, the National Institutes of Health, the CDC, and other federal agencies should proactively fund additional forensics research on the physiological causes of deaths in custody as an alternative to research funded by interested parties such as Axon. This research can help to inform and improve cause and manner of death determinations.