Understanding Police Use of Force via Hospital Administrative Data: Prospects and Problems

Mooney and colleagues¹ put forth an innovative examination of legal intervention injuries (ie, police use of force) over an 11-year period in California. The central innovation of this research is the use of administrative data mandatorily reported by license hospitals on all hospital visits made between 2005 and 2015. Such data are a resourceful means of examining police use of force, given that police departments often do not collect such data. From hospital visit data, legal intervention injuries were identified using administrative injury codes. Mooney and colleagues examined trends in injury rates, severity of injuries, and calculated per capita, as well as per arrest rates of legal intervention injuries for males aged 14 to 64 years by race or ethnicity. The analyses found distinct time trends in injury rates and injury severity, both of which generally increased until 2009 and then declined thereafter. Throughout the period of interest, black individuals had the highest per capita rates of legal intervention injuries followed by Hispanic individuals, white individuals, and Asian/Pacific Islander individuals, respectively. However, when each racial or ethnic group’s number of legal intervention injuries were standardized by taking into consideration race differences in arrests, racial or ethnic disparities in legal intervention injuries largely vanished (eg, in 2005, the 95% CI for black to white arrest rate ratio was 0.91-1.00). The authors conclude that state health care administrative records offer a rich data source that warrants further analysis.

I agree with the conclusion by Mooney and colleagues that health care administrative data offer a valuable source of data on legal intervention injuries. As the current research demonstrates, the scarcity of available data on police use of force can be partially remedied by obtaining and analyzing hospital administrative data. Not only are these data available across many jurisdictions (ie, cities and counties), they are also reported in a standardized format, which makes these data capable of supporting research on the prevalence, trends, and racial disparities in legal intervention injuries across an entire state.

Yet, this big data approach to understanding police use of force has limitations that need to be addressed. One limitation that is acknowledged by the authors but apparently unable to be addressed by the current research is the validity of the legal intervention variable. It is unknown how frequently legal intervention injuries lead to hospital visits, how frequently patients revealed to hospital personnel that they were harmed by law enforcement officers, and how frequently hospital personnel correctly coded injuries caused by law enforcement officers when identified by patients. Another limitation with this approach is that by relying on hospital admissions and emergency department visits, only the most severe injuries are captured. There is a continuum of police use of force, in order of likelihood of injury: soft empty hand control (eg, grabs and holds), hard empty hand control (eg, punches and kicks), less lethal methods (eg, baton, pepper spray, and taser), and lethal force (eg, firearm). The most commonly used police use of force is soft empty hand control,² yet such uses of force are unlikely to result in a hospital visit. Thus, the use of hospital records undoubtedly leads to serious undercounting of police use of force.

The racial or ethnic disparity analyses need to be interpreted with caution for several reasons. First, by focusing on the upper end of the use of force continuum, this and other research examining hospital visits may bias racial or ethnic comparisons as there is evidence police disproportionately use soft empty hand control against minorities but racial disparities are not evident in more severe forms of use of force,³ with the exception of taser use, which is more commonly used to subdue black individuals than white individuals.^2,4 Second, there appears to be race or ethnic differences in hospital visits, especially among black men who have had previous contact with the criminal justice system. As an example of this phenomenon, Goffman⁵ reported that black men in Philadelphia, Pennsylvania, actively avoided hospitals because police officers often identified individuals with warrants, suspicion for crimes, and probation and/or parole violations by visiting hospitals and emergency departments. Third, while the finding that race or ethnic disparities in use of force diminish after taking race or ethnic differences in arrest into account, arrests are susceptible to the same racial or ethnic biases as police use of force. My work finds that racial disparities in drug arrests, one of the most common arrest categories, cannot be explained away by race differences in drug offending, nondrug offending, and community characteristics.⁶ Likewise, the bulk of the research finds that black suspects generally are more likely to be arrested than white suspects.⁷

My final issue with the use of hospital administrative data to address the questions posed by Mooney et al¹ is that such data are largely context free. It is well established that police use of force varies by neighborhood and police agency characteristics.⁴ Use of force is more likely and race disparities are smaller in high-crime areas, and use of force is less likely in police agencies with more stringent policies regulating force. Unfortunately, hospital records by themselves are not well suited toward capturing and examining these kinds of important contextual variations in the prevalence, trend, and racial disparity in legal intervention injuries. A potential way forward would be to geocode hospital visits and match these visits to community and/or police agency characteristics.

Until police agencies are required to collect use of force data in a standardized form, the use of hospital administrative data appears to be the best available option in many research situations.

Published: September 14, 2018. doi:10.1001/jamanetworkopen.2018.2231

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2018 Mitchell O. JAMA Network Open.

Corresponding Author: Ojmarrh Mitchell, PhD, Department of Criminology, University of South Florida, 4202 E Fowler Ave, SOC 107, Tampa, FL 33620 (omitchell@usf.edu).

Conflict of Interest Disclosures: None reported.