Association of Neighborhood Gun Violence With Mental Health–Related Pediatric Emergency Department Utilization

Key Points

Question  Is neighborhood gun violence exposure associated with children’s mental health–related pediatric emergency department (ED) utilization?

Findings  In this cross-sectional study of 54 341 children, pediatric ED utilization was compared before and after episodes of neighborhood gun violence. Among children living within 2 to 3 blocks of a shooting, increased mental health–related ED utilization was found at 2 weeks, 1 month, and 2 months after the shooting.

Meaning  Exposure to gun violence is associated with an increase in children’s acute mental health symptoms, suggesting that trauma-informed care must be prioritized in communities with a high prevalence of violence exposure, and public policies that reduce neighborhood gun violence are urgently needed.

Importance  Many children and adolescents in the United States are exposed to neighborhood gun violence. Associations between violence exposure and children’s short-term mental health are not well understood.

Objective  To examine the association between neighborhood gun violence and subsequent mental health–related pediatric emergency department (ED) utilization.

Design, Setting, and Participants  This location-based cross-sectional study included 128 683 ED encounters for children aged 0 to 19 years living in 12 zip codes in Philadelphia, Pennsylvania, who presented to an urban academic pediatric ED from January 1, 2014, to December 31, 2018. Children were included if they (1) had 1 or more ED visits in the 60 days before or after a neighborhood shooting and (2) lived within a quarter-mile radius of the location where this shooting occurred. Analysis began August 2020 and ended May 2021.

Exposure  Neighborhood violence exposure, as measured by whether a patient resided near 1 or more episodes of police-reported gun violence.

Main Outcomes and Measures  ED encounters for a mental health–related chief complaint or primary diagnosis.

Results  A total of 2629 people were shot in the study area between 2014 and 2018, and 54 341 children living nearby had 1 or more ED visits within 60 days of a shooting. The majority of these children were Black (45 946 [84.5%]) and were insured by Medicaid (42 480 [78.1%]). After adjusting for age, sex, race and ethnicity, median household income by zip code, and insurance, children residing within one-eighth of a mile (2-3 blocks) of a shooting had greater odds of mental health–related ED presentations in the subsequent 14 days (adjusted odds ratio, 1.86 [95% CI, 1.20-2.88]), 30 days (adjusted odds ratio, 1.49 [95% CI, 1.11-2.03]), and 60 days (adjusted odds ratio, 1.35 [95% CI, 1.06-1.72]).

Conclusions and Relevance  Exposure to neighborhood gun violence is associated with an increase in children’s acute mental health symptoms. City health departments and pediatric health care systems should work together to provide community-based support for children and families exposed to violence and trauma-informed care for the subset of these children who subsequently present to the ED. Policies aimed at reducing children’s exposure to neighborhood gun violence and mitigating the mental symptoms associated with gun violence exposure must be a public health priority.

Many children living in the United States are exposed to gun violence in their neighborhoods. In a 2015 national survey, 1 in every 8 children aged 14 to 17 years reported having witnessed a shooting.¹ Black children are at particularly high risk of gun violence exposure; as many as 51% of Black children aged 10 to 17 years report having heard or witnessed a shooting in their neighborhood.^2-4

This persistent and pervasive violence exposure has been associated with poor mental health in children.^5-7 Prior studies have found that children with self-reported neighborhood gun violence exposure have increased rates of mental health diagnoses, including anxiety, depression, and posttraumatic stress disorder, and mental health symptoms, including withdrawn behavior, irritability, and disruptive behaviors in school.^8-17 A recent study found an increase in children’s self-reported anxiety and depression symptoms in the 12 months following gun homicides occurring within a mile of their home or school,¹⁸ while another study found an increase in youth antidepressant use in the 2 years after exposure to a school shooting.¹⁹

Much of the prior work examining associations between violence exposure and mental health has been limited by a reliance on self-reported data, leaving these studies vulnerable to recall bias. In addition, most previous studies have focused on the long-term mental health effects of violence exposure in the following months and years. Therefore, the immediate implications of neighborhood violence exposure for children’s mental health are less well understood.

Understanding children’s mental health symptoms in the days and weeks following an episode of neighborhood gun violence could enhance our understanding of the associations between violence exposure and short-term mental health and inform the design of interventions focused on providing mental health support following a shooting. However, no prior research has objectively evaluated the association between neighborhood gun violence and mental health symptoms in children, to our knowledge. Therefore, we conducted a location-based study of the places where shootings occurred and aimed to (1) examine the association between neighborhood gun violence and children’s acute mental health symptoms, as measured by mental health–related pediatric emergency department (ED) utilization in the 2 months after exposure to 1 or more shootings and (2) assess for variations in this association based on patients’ geographic and temporal proximity to a shooting.

We performed a location-based cross-sectional study examining the locations of all police-reported shootings taking place within 12 zip codes in Philadelphia, Pennsylvania, from January 1, 2014, to December 31, 2018. We selected zip codes surrounding the large tertiary care pediatric hospital from which ED utilization data were drawn because our study hospital houses the closest pediatric ED for children living in this area. The Children’s Hospital of Philadelphia institutional review board approved this study and waived the need for informed consent given the use of deidentified participant data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.

The primary exposure in this study was living within a quarter mile (4-5 blocks) of a neighborhood shooting. We used publicly available data from the Philadelphia Police Department’s repository of shootings on the city’s open-data website, OpenDataPhilly,²⁰ to identify the location of all gun violence incidents in our study geography and time period. Children were considered exposed if 1 or more shootings occurred within a quarter mile of their home address in the 60 days prior to their ED visit.

We sought to identify a comparison group of children whose neighborhood context would be similar to the exposed group. Therefore, the reference group consisted of children whose home address was also within a quarter mile of a shooting but who visited the ED in the 60 days before that shooting. Children were excluded from this reference group if they were exposed to any other shootings in the 60 days preceding their ED visit.

In our primary analysis, shooting exposure was coded as a binary variable, with 1 indicating exposure to any shooting in the 60 days preceding a visit and 0 indicating exposure to any shooting in the 60 days after a visit (without having had an exposure in the preceding 60 days). As a secondary analysis, we examined cumulative shooting exposure in the preceding 60 days as a categorical variable, distinguishing between children exposed to 1 shooting, 2 shootings, and 3 or more shootings in the 60 days preceding their ED visit.

The primary outcome of interest was mental health–related ED utilization, coded as a binary variable, with 1 indicating a mental health–related ED visit and a 0 indicating an ED visit that was not mental health related. ED visits were classified as mental health related if they had either (1) a mental health–related chief complaint or (2) a mental health–related primary diagnosis, as determined by International Classification of Diseases, ninth or tenth revision code (eTable 1 in the Supplement). ED chief complaints were selected by a triage nurse from a fixed set of options in a drop-down menu. ED primary diagnoses were assigned by the patient’s clinician at the time of ED discharge, hospital admission, or transfer. Chief complaint and primary diagnosis data were abstracted from the electronic health record.

Individual-level covariates used in this analysis included patient age, sex, race and ethnicity, zip code of residence, and receipt of public insurance (Medicaid or Children’s Health Insurance Program), all obtained from the electronic health record. Race and ethnicity data were entered into the electronic health record based on either caregiver self-report or assignment by a member of the ED registration team. Median household income data at the zip code level was obtained from the 2018 American Community Survey and included as an additional covariate.

Shooting location and patient residential address were geocoded using ArcGIS (ESRI) software. For each shooting location, we created 2 separate buffers at diameters of a quarter mile (4-5 blocks) and one-eighth of a mile (2-3 blocks) around the shooting, based on prior work evaluating associations between place-based interventions, crime, and health.²¹ ED encounters were spatially joined to shooting incidents if a patient’s residential address fell within the selected geographic buffer.

After spatially linking all ED visits and shooting incidents in the study area, we evaluated ED visits occurring within 6 time intervals before and after each event: 7, 14, 21, 30, 45, and 60 days (eFigure 1 in the Supplement).

We used descriptive statistics to summarize demographic characteristics for all people who were shot in the gun violence incidents included in our data set. We compared patient demographic characteristics for children with 1 or more ED visits in the 60 days before and after an episode of gun violence.

We then used multivariable logistic regression to examine associations between violence exposure and ED utilization for mental health–related conditions (eFigure 2 in the Supplement) and calculated odds ratios (ORs), adjusted ORs (aORs), and 95% CIs. All models included month and year dummy variables to adjust for seasonal variation and variation in mental health–related ED use over time. Models were clustered by patient medical record number to account for individual patients who had multiple visits during the study period.

To examine whether the association between neighborhood gun violence and mental health–related ED utilization varied based on geographic proximity to the shooting event, we conducted 2 sets of analyses, 1 for the larger set of children who lived within a quarter mile of an episode of neighborhood violence and a second for the smaller subset of children who lived within one-eighth of a mile of one of these episodes. To examine the association between repeated gun violence exposure and mental health symptoms, we conducted a secondary analysis with cumulative shooting exposure in the preceding 60 days included as a categorical exposure variable.

Location information was available for all shootings included in our data set, and home address information was available for all corresponding ED visits. Only 0.2% of ED visits were missing both a chief complaint and a primary diagnosis and were excluded. All covariates included in our models had less than 0.5% missing data, and we therefore used complete case analysis.

Analyses were conducted using Stata software version 15.1 (StataCorp). All P values were from 2-sided tests and were deemed statistically significant at P < .05.

Interaction tests were used to examine effect modification of the association between violence exposure and mental health–related ED utilization by distance, as well as by patient age, sex, and race.

We conducted several sensitivity analyses. First, to account for possible correlation in the ED utilization patterns of children living within each geographic buffer as a result of their residence in the same small area, we used conditional logistic regression models with grouping by shooting location. Second, we conducted a sensitivity analysis restricting our sample to the subset of patients who had visits before and after a shooting. Lastly, to assess the degree to which violence exposure may have been associated with nonmental health–related ED utilization in our study sample, we examined the association between gun violence exposure and ED utilization for respiratory and somatic symptoms. Analysis began August 2020 and ended in May 2021.

There were 2629 shooting incidents in our study area between January 1, 2014, and December 31, 2018 (Figure). Most shootings occurred outdoors (2401 [91.3%]), and most people who were shot were Black (2395 [91.0%]) and male (2436 [92.7%]) (Table 1).

Our study sample included 54 341 unique patients with a median (interquartile range) age of 5.5 (1.5-11.1) years. About half (27 972 [51.5%]) were male. The majority of these patients identified as Black or African American (45 947 [84.5%]). Most patients (42 480 [78.1%]) had Medicaid insurance.

Of 54 341 patients, 43 143 (97.4%) had 1 or more ED visits in the 60 days following a shooting, and 42 913 (79.0%) had 1 or more ED visits in the 60 days prior to a shooting. Patients with ED visits before and after shootings had similar demographic characteristics (Table 2).

The 54 341 patients included in our study sample had a total of 128 683 ED visits during the study period. Of these visits, 2428 (1.89%) were classified as mental health related. Of the 2629 shooting incidents in the data set, 814 (31.0%) had 1 or more corresponding mental health–related ED visits in the subsequent 60 days.

After adjusting for covariates, children residing within one-eighth of a mile of an episode of gun violence had greater odds of having a mental health–related ED visit (Table 3). This association was most marked in the 14 days following this shooting (aOR, 1.86 [95% CI, 1.20-2.88]). This association decreased in a time-dependent manner, including at 30 days (aOR, 1.49 [95% CI, 1.11-2.03]) and 60 days (aOR, 1.35 [95% CI, 1.06-1.72]). In an analysis of children living within a quarter mile of a shooting, we found a similar association between violence exposure and mental health–related ED utilization at 7 through 60 days. The association was stronger for children living closer to the shooting (14-day aOR, 1.64 [95% CI, 1.01-2.66]; P = .047 for interaction term). There was no modification of the association based on patient race (14-day aOR, 1.05 [95% CI, 0.54-2.03]; P = .88 for interaction term), age (14-day aOR, 0.85 [95% CI, 0.58-1.26]; P = .44 for interaction term), or sex (14-day aOR, 1.00 [95% CI, 0.68-1.47]; P = .99 for interaction term).

In our secondary analysis examining cumulative shooting exposure in the preceding 60 days, we found greater odds of mental health–related ED utilization among children exposed to 2 shootings (14-day aOR, 2.34 [95% CI, 1.44-3.82]) and children exposed to 3 or more shootings (14-day aOR, 1.92 [95% CI, 1.00-3.70]), suggesting that the association between neighborhood gun violence and mental health symptoms may be exacerbated by repeated violence exposure (Table 4).

Results were similar in our sensitivity analyses, including an analysis using conditional logistic regression, with matching based on shooting location, and an analysis restricting our sample to children with visits both before and after a shooting (eTables 2 and 3 in the Supplement). We found no increase in the odds of ED utilization for respiratory or somatic symptoms following violence exposure (eTable 4 in the Supplement).

In this location-based cross-sectional study, we found a significant increase in pediatric mental health–related ED utilization following incidents of neighborhood gun violence, most pronounced in the 2 weeks following the shooting, among children residing closest to where this violence occurred and among children exposed to multiple shootings.

Our results build on prior work demonstrating links between community violence exposure and mental health both by using an objective measure of gun violence exposure and by showing that mental health symptoms associated with this exposure can be seen within days.^4-13

While our findings are not sufficient to demonstrate a causal relationship between gun violence exposure and children’s mental health, there are several potential mechanisms that could underlie our observed association. First, violence exposure could lead to increased stress, causing harmful physiologic disturbances and promoting maladaptive coping behaviors.^22-29 Second, for children with previous mental health diagnoses, violence exposure could precipitate exacerbations of chronic conditions such as depression.^7,13,18 Third, these incidents could contribute to a perceived lack of safety and an increased level of anxiety for caregivers, lowering their threshold to bring children to the ED for mental health concerns they would otherwise have managed at home or with outpatient care.³⁰

We found that children exposed to multiple shootings in the 60 days preceding their ED visit had greater odds of subsequent mental health–related ED use. This result underscores the importance of addressing the cumulative mental health burden of gun violence exposure in the neighborhoods where shootings occur most frequently. In Philadelphia and other cities across the United States, gun violence disproportionately affects Black children and families, in large part owing to the legacy of historical and structural racism, including practices such as redlining and discriminatory bank lending and real-estate practices, which have led to persistent racial segregation and concentrated poverty.^31-33 Disproportionate violence exposure among Black children could perpetuate racial disparities in child health outcomes, and reducing this exposure should be a priority for health care systems working toward health equity.^34-36

Our findings underscore the need for effective public health interventions aimed at both reducing children’s exposure to gun violence and mitigating the mental health symptoms associated with this exposure.^37,38 Reducing the direct and indirect consequences of community violence will require a range of evidence-based violence prevention efforts, including safe storage and background check laws aimed at reducing firearm availability, place-based structural interventions such as vacant lot greening, interventions that support youth engagement through education, mentorship, and employment, and interventions that mitigate financial stress for all neighborhood residents.^21,39,40 Realignment of local government spending priorities could be an important tool in supporting children exposed to community violence, for example, by diverting funds previously allocated to law enforcement to instead provide additional support for mental health services and violence prevention.^41-43

Our findings highlight the importance of providing universal trauma-informed care in the ED because many children seen in the ED may have been exposed to neighborhood violence even if they do not report this exposure as a reason for their presentation.^44-46 In addition, health care systems should partner with community-based organizations to provide preventive and responsive support for children and families exposed to neighborhood gun violence. For example, there may be benefit to proactively reaching out to families of children known to a health care system who live close to a shooting in the days or weeks after that shooting to offer mental health resources and support in a trauma-sensitive manner.

Our study has several limitations. As a single-center study, our results may not be generalizable to other geographic settings. We may not have captured all ED utilization for children living in the included area because children may have sought care in other pediatric or adult EDs. Additionally, we were unable to capture changes in primary care or other urgent care utilization. In addition, we may not have detected all mental health–related ED utilization because some patients may have visited the ED for a primary physical health concern but had a concurrent mental health comorbidity not captured in their chief complaint or primary diagnosis. In addition, ED utilization rates for nonmental health–related stress-sensitive conditions, like asthma, may have been increased by violence exposure, dampening our observed associations. We would expect our focus on ED utilization, our use of a single-ED chief complaint and diagnosis, and our inability to account for changes in nonmental health–related ED use to bias our effect size to the null, so the association between neighborhood gun violence and mental health symptoms may be stronger than reflected in our data.

Our use of police-reported gun violence, while objective, may undercount exposure given that some gun violence may be unreported and does not adequately capture nuances of this exposure such as a child’s knowledge of the event and familiarity with the person who was shot, survivor, or perpetrator. Omitted variable bias may have affected our findings, given our inability to control for income at the household level, prior exposure to violence, and prior mental health symptoms or diagnoses. Our use of complete case analysis represents another limitation, although our amount of missing data was small (<0.5%).

In addition, the geographic buffers used in our analysis reflect physical distance but may not reflect how neighborhoods are perceived by their inhabitants. Lastly, race and ethnicity data used in this study were either self-reported or assigned by a member of the ED registration team, creating the potential for misclassification. Our study also has a number of strengths, including our large sample size, use of objective data to assess violence exposure and health care utilization, and ability to adjust for several potential confounders.

In this single-center observational study, recent exposure to neighborhood gun violence was associated with increased odds of mental health–related pediatric ED visits among children living within 4 to 5 blocks of the shooting. Our findings are particularly relevant given the recent surge in gun violence in cities across the United States during the COVID-19 pandemic.^47-49 Mitigating the health symptoms associated with gun violence exposure must be a priority for policy initiatives aimed at promoting health equity. Public health departments and pediatric health systems should partner with community-based organizations to provide both community-based support for all children exposed to violence and trauma-informed care for the subset of children who subsequently present to the ED.

Corresponding Author: Aditi Vasan, MD, MSHP, Department of Pediatrics, Children's Hospital of Philadelphia, Roberts Center for Pediatric Research, 2716 South St, Philadelphia, PA 19146 (vasana@chop.edu).

Accepted for Publication: June 22, 2021.

Published Online: September 20, 2021. doi:10.1001/jamapediatrics.2021.3512

Author Contributions: Dr Vasan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Vasan, Mitchell, Fein, Wiebe, South.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Vasan, Mitchell, Fein, Buckler.

Critical revision of the manuscript for important intellectual content: Mitchell, Fein, Buckler, Wiebe, South.

Statistical analysis: Vasan, Mitchell, Buckler, South.

Administrative, technical, or material support: Buckler, South.

Supervision: Fein, Wiebe.

Conflict of Interest Disclosures: Dr Vasan reported grants from the Academic Pediatric Association and the Agency for Healthcare Research and Quality outside the submitted work. No other disclosures were reported.