Association of Legal Intervention Injuries With Race and Ethnicity Among Patients Treated in Emergency Departments in California

IMPORTANCE Increased public concern regarding police use of force has coincided with a dearth of available data to uncover the magnitude and trends in injuries, particularly across race or ethnicity. OBJECTIVE To examine trends in injury rates, severity, and disparities across black individuals, white individuals, Hispanic individuals, and Asian/Pacific Islander individuals. DESIGN, SETTING, AND PARTICIPANTS In this retrospective, cross-sectional study, data collected on every hospital visit in California from January 1, 2005, to September 30, 2015, were used to model trends in rates of legal intervention injuries (n = 92 386) per capita and per arrest for men aged 14 to 64 years, by race or ethnicity. The study also examined descriptive statistics on injury dispositions to assess changes in severity. Analyses were conducted between December 2017 and June 2018. MAIN OUTCOMES AND MEASURES All visits with an external cause of injury code of E970 to E977 were classified as legal intervention injuries. This range of codes includes injuries inflicted by the police or other law-enforcing agents in the course of arresting or attempting to arrest lawbreakers, suppressing disturbances, maintaining order, and other legal action. RESULTS The study identified a total of 92 386 hospital visits that were the result of legal intervention among males aged 14 to 64 years. Black individuals were at the highest risk of legal intervention injury per capita in 2005 (for black vs white individuals, rate ratio, 2.90; 95% CI, 2.74-3.06), and remained so across the study period. Although rates among Asian/Pacific Islander individuals remained stable, rates in all other groups increased from 2005 to 2009 and then declined from 2009 to 2015, nearly returning to 2005 levels. During the period of increasing rates, the black to white disparity widened by 3% annually (rate ratio, 1.03; 95% CI, 1.01-1.05), then narrowed as rates declined. In contrast, rates of injury per arrest have increased over the past decade, although rates were broadly similar across race or ethnicity. The proportion of injuries involving firearms (ie, shootings by police) declined from 7.0% in 2005 and 2006 to 3.7% in 2014 and 2015. CONCLUSIONS AND RELEVANCE States with central repositories for hospital visits offer data sources to illuminate the public health problem of legal intervention injuries, and warrant greater attention to ensure consistent coding for complete capture. JAMA Network Open. 2018;1(5):e182150. doi:10.1001/jamanetworkopen.2018.2150


Introduction
Public concern about police use of force has intensified in recent years following a series of highly publicized shootings of unarmed black men. [1][2][3][4][5][6][7] However, efforts to assess disparities in the use of force and changes over time have faced a basic obstacle: the lack of reliable data regarding the prevalence of law enforcement-related injuries. 8

+ Invited Commentary
Author affiliations and article information are listed at the end of this article.
Numerous governmental databases aspire to measure deaths due to police action, but all have significant and well-known methodological problems. For example, the Federal Bureau of Investigation's Supplementary Homicide Reports and the Bureau of Justice Statistics' Arrest-Related Deaths Program are both voluntary reporting systems with uneven compliance; the National Vital Statistics System, developed by the National Center for Health Statistics, relies on reports by death certifiers, who often fail to identify police involvement in homicides. [10][11][12] Victim surveys by the US Department of Justice have additional shortcomings, such as the lack of information on the location of interactions, inaccuracies associated with self-reporting, the exclusion of incarcerated individuals, and infrequent implementation (once every 3 years). 13 With no comprehensive national data on fatal injuries inflicted by police, The Guardian developed The Counted, 14 which uses reporting and crowdsourced information to build a database of deaths inflicted by police or other law enforcement agencies. However, nonfatal injuries are excluded. Some public health researchers have looked to alternative measures of police-related injuries to address the gap. One source has been national samples of emergency department (ED) visits, which include codes for legal intervention. Among these are the Centers for Disease Control and Prevention's Web-Based Injury Statistics Query and Reporting System, the National Inpatient Sample, and the Nationwide Emergency Department Sample. The latter 2 are part of the Healthcare Cost and Utilization Project family of databases. 15,16 Several recent studies have used these data sources to examine police use of force injuries, [17][18][19] although none of these studies have examined trends in injuries by race or ethnicity.
To address these shortcomings, we used a new data source-information collected on every outpatient ED visit and inpatient admission in California during an 11-year period (2005-2014).
Because California has a central repository for these data and requires all hospitals to collect patient race or ethnicity, the analysis offers insights into trends among white individuals, black individuals, Hispanic individuals, and Asian/Pacific Islander individuals not possible with other data sets. The data provide a unique opportunity for longitudinal analysis of all legal intervention injuries passing through any hospital, including the primary reason for the visit, any additional diagnoses recorded during the visits, the injury mechanism, and patient demographic characteristics.

Methods
Nonpublic (restricted use) patient discharge data and ED data were obtained from the California Office of Statewide Hospital Planning and Development. All licensed hospitals in the state except those that are federally operated are required to submit information on all inpatient and ED visits. We

Measures
To identify injuries that were the result of law enforcement contact, we used external cause of injury codes (E-codes) described in the ICD-9-CM system. 20 E-codes are intended to capture how an injury occurred and the intent. For each case, the hospital discharge data contain a primary E-code and up to 4 additional E-codes. All cases with an E-code of E970 to E977 were classified as legal intervention injuries. This range of codes includes injuries inflicted by the police or other law-enforcing agents in the course of suppressing disturbances, maintaining order, arresting or attempting to arrest offenders, or other legal action.

Statistical Analysis
We present statewide summaries of demographic characteristics of patients with legal intervention injuries, and the mechanisms, dispositions, primary diagnoses, and co-occurring diagnoses for those visits. We then model trends in monthly rates of legal intervention injuries by race or ethnicity (non-Hispanic white individuals, black individuals, Hispanic individuals, and Asian/Pacific Islander individuals), per population and per arrest. Poisson models were specified with robust standard errors and offsets for race-specific population and arrest counts to generate rates per population and per arrest, respectively. Calendar month dummies controlled for seasonal trends, and secular trends were modeled using linear splines for month-year to allow for changes in trend. Spline knots were placed at 2009 for the injuries per population model and 2013 for the injuries per arrest model, selected based on visual inspections of graphs of monthly rates to identify the points at which trends shifted. Secular trend variables were interacted with race or ethnicity to allow trends to differ by group. Linear combinations of model coefficients for trends in monthly rates were used to generate estimates of annual changes in rates and rate ratios (RRs) for given time periods, and 95% confidence intervals were calculated.

Characteristics of ED Patients With Legal Intervention Injuries
Over the 11-year study period, we identified a total of 92 386 hospital visits that were the result of legal intervention among males aged 14 to 64 years ( Table 1). Compared with their representation in the population, injuries were disproportionately high among younger age groups, black individuals, and Hispanic individuals. Men aged 25 to 34 represented 31.8% of injuries (n = 29 404), but 21.1% of

Characteristics of Legal Intervention Injuries
Injury severity appears to have changed over time. Table 2  The data also indicate an increase in the incidence of a co-occurring behavioral health diagnosis for patients with legal intervention injuries. Legal intervention ED visits with any diagnosis indicating a mental health condition increased from 12 574 (8.6%) to 37 037 (24.6%). Visits with an alcohol-or substance-related disorder increased as well, from 2232 (15.6%) to 3435 (22.2%). Similar trends were seen in co-occurring diagnoses for patients with injuries from assaults not associated with legal intervention.

Trends in Legal Intervention Injury Rates and Racial or Ethnic Disparities
Black individuals were at the greatest risk of legal intervention injury on a population level. In 2005, black individuals experienced 159 injuries per 100 000 population compared with 55 among white     In 2005, legal intervention injuries were highest among white individuals, at 470 per 100 000 arrest (Figure 2 and Table 3 Over the 11-year study period, legal intervention injuries per arrest increased more than legal intervention injuries per capita (Figure 1 and Figure 2). During this time period, the rate of legal intervention injuries per 100 000 population grew modestly (Figure 1). At the same time, the arrest rate per 100 000 declined substantially. Specifically, arrests of white individuals declined by 19%, black individuals by 28%, Hispanic individuals by 37%, and Asian/Pacific Islander individuals by 33%.

JAMA Network Open | Public Health
An increase in injuries per population despite declining arrest rates necessarily leads to a significant increase in the likelihood of injury per arrest, which is reflected in the data.

Discussion
Our findings suggest disparities in the rate of injury from legal intervention for different racial groups but do not identify the cause. One would expect that a group would have more injuries if that group was subject to more police interventions. Thus, we looked at legal intervention injuries per arrest for each group. Arrests do not account for all law enforcement actions (such as stop and frisk interactions short of arrest). As arrests decrease, stops without arrest could theoretically increase.
That said, police use of force typically corresponds with stops that resulted in an arrest, 23 so injuries per arrest can provide a rough proxy for the relevant legal interventions.

Rate of Legal Intervention Injuries per Population and per Arrest
Our  nature of police-public contacts or types of arrests. Additional research is warranted to explore the dynamics of the arrest disparities further.

Severity of Legal Intervention Injuries Over the Past Decade
The proportion of injuries involving firearms or resulting in death has declined. Results may reflect evolving police agency procedures requiring an ED visit following use of force incidents, or jail policies requiring that individuals be examined prior to booking. Increasing use of tasers might have contributed to the corresponding decrease in firearm-related injuries. One converse result is notable: despite a decreased proportion of visits resulting in death, the share of visits that resulted in a hospital admission has been increasing.
The uptick in both mental health conditions and substance use disorders among patients with legal intervention injuries warrants further research. This trend may be a result of increases in hospital testing for these conditions, rather than a change in the population with legal intervention injuries, especially considering the parallel increases in these diagnoses for patients who have been assaulted. Alternatively, it could reflect changes in the populations police are interacting with in communities, following criminal justice reforms that shifted individuals from state prisons and paroles to county jails and probation. 28 In either case, the data suggest there is a clear and potentially growing need for police agencies to continue to develop strategies for de-escalating interactions and proactively addressing mental health needs by collaborating with county mental health professionals. 29 Policies and practices in policing are evolving, particularly as public concern about police use of force has increased in recent years. Our findings reveal variations in legal intervention injury rates per population across race or ethnicity, but far less variation in risk of injury on a per arrest basis. The results also suggest that injuries may be decreasing in severity. These findings highlight the value of using hospital data to assess the rate of nonfatal injuries, injuries that can have a profound impact on community health and affect the public's trust in law enforcement. There is a need to promote the complete and accurate documentation of legal intervention injuries in EDs not only in California, but in other states as well.

Limitations
The data set used offers a comprehensive look at ED visits in California, but faces certain limits. First, the results may have been affected by changes in hospital coding practices over time. Increased attention to use of force nationwide may have contributed to greater care among hospital staff to indicate law enforcement involvement in visits for legal intervention injury, inflating upward trends in injuries. Overall, injuries may be underestimated since legal intervention is not a valid cause code for all types of injuries that may occur in the course of police interaction, such as bites by police dogs or falls.
Second, we are unable to determine who, among those injured by police, seeks care in hospitals or whether this varies by race or ethnicity. Of those who do seek care, there could be differences across race or ethnicity in reporting to the clinician that police inflicted the injury, which is a prerequisite for coding legal intervention.
With regard to race or ethnicity, there may also be discrepancies in self-reported race or ethnicity used in population denominators and race or ethnicity selected in hospital visit records or arrests. Research comparing medical administrative data and self-reported survey responses has indicated alignment between those identifying as black or white, but underrepresentation in administrative data among those identifying as Hispanic or Asian/Pacific Islander. 30 This would suggest the possibility that rates of injury per population could be underestimated in these 2 groups.
Racial or ethnic categories are further limited by the absence of multiracial options.
While any instance of injury should be of concern, there is no way to determine from our data whether instances of force used were considered excessive force. Making such a determination involves a somewhat subjective judgment. Fatal injuries are likely to be underrepresented in the data