Characterizing resilient and at-risk neighborhoods with lower- and higher-than-expected firearm injuries and fatalities in Pittsburgh, Pennsylvania

Abstract

Introduction

Firearm injuries and fatalities adversely affect individual and community health and are important public health issues. Identifying factors that protect against firearm violence and related harms is needed. We aimed to identify neighborhoods with lower- and higher-than-expected firearm injuries and fatalities based on social vulnerability and their differing structural and social characteristics.

Methods

We used Gun Violence Archive data from the RISE Lab Firearm Injury Data Hub to estimate firearm injury and fatality counts between 2015–2020 in Pittsburgh, Pennsylvania. We used a negative binomial regression model to estimate the relationship between the CDC’s Social Vulnerability Index and firearm injuries and fatalities and the residual percentile method to identify neighborhoods with lower (resilient) and higher (at-risk) counts than expected based on social vulnerability. T-tests were used to compare resilient and at-risk neighborhoods for 158 US Census American Community Survey demographic, socioeconomic, housing, and transportation variables.

Results

Resilient neighborhoods (N=19, bottom quartile residuals) had lower rates of firearm injuries and fatalities (p=0.0002) compared to at-risk neighborhoods (N=19, top quartile residuals). Resilient and at-risk neighborhoods differed for 2 (percent male never married and female widowed) of the variables included in analyses.

Conclusions

Some neighborhoods, despite facing risk factors, experienced fewer firearm injuries and fatalities than would be expected based on neighborhood social vulnerability. Existing sources of secondary data on neighborhoods may fail to adequately capture potential factors that prevent against firearm violence and to measure resilience in thriving neighborhoods. Future community-engaged studies are needed to understand and measure neighborhood-level protective factors.

INTRODUCTION

Firearm injuries and fatalities are important public health issues in the United States, with neighborhood socioeconomic factors playing a key role in impacting the distribution of firearm violence across populations [1, 2]. Prior studies have shown the relationship between neighborhood disadvantage indices, such as the Social Vulnerability Index (SVI), and firearm violence [1, 2]. SVI, like other neighborhood disadvantage indices, uses variables from the United States (US) Census to capture various elements of disadvantage, with the goal of identifying areas that could benefit from resource allocation or intervention [3]. However, while measures of disadvantage are often closely associated with public health outcomes [3], studies have also identified areas with lower adverse outcomes than would be expected based on neighborhood disadvantage alone [4, 5]; this work has highlighted the importance of identifying unmeasured factors that may help mitigate the effect of adverse neighborhood conditions on firearm violence and related harms.

In this study, we aim to identify neighborhoods within Pittsburgh, Pennsylvania with higher (termed “at-risk”) and lower (termed “resilient”) counts of firearm injuries and fatalities than would be expected based on neighborhood social vulnerability. We chose to focus our study on Pittsburgh rather than nationally because our goal was to inform future research through The Pittsburgh Study, a community-engaged study on neighborhood thriving [6]. We also sought to compare neighborhood demographic, socioeconomic, housing, and transportation characteristics from the US Census between resilient and at-risk neighborhoods to determine if existing secondary data adequately capture potential factors that prevent against firearm violence. Identification of areas resilient to firearm violence can lay the foundation for future work that can inform interventions and community-led firearm violence reduction efforts in Pittsburgh [7, 8]. Findings also have the potential to shed light on the need to better understand local factors that prevent against firearm violence and related harms and promote neighborhood resilience.

METHODS

Point-level geocoded firearm injury and fatality data from the Gun Violence Archive (GVA) was downloaded from the RISE Lab Firearm Injury Data Hub [9, 10]. GVA is a nonprofit organization that provides publicly available data on firearm violence in the US, with their staff logging characteristics such as date, location, and the number of people injured or killed daily based on law enforcement and media resources [10, 11]. All incidents of firearm injuries and fatalities are included in the dataset which aligned with our focus on broad violence exposure rather than specific subsets of firearm violence. We then selected firearm injury and fatality points within Pittsburgh that occurred between 2015–2020 and aggregated them to the census tract level. Firearm injury and fatality counts were merged with the CDC’s census tract-level Social Vulnerability Index dataset [12], which includes SVI and the 15 variables from the US Census 2012–2016 American Community Survey (ACS) 5-year estimates that are used in the index. We utilized SVI as it is a common measure used in public health and has been shown to be a marker of community response to external stressors.

Because our goal was to inform research specific to neighborhood thriving in Pittsburgh, we aggregated census tracts to the neighborhood level, allowing us to consider culturally defined regions. Pittsburgh consists of geographically stable and culturally defined neighborhoods that comprise between 1 and 5 census tracts. We calculated neighborhood firearm violence and SVI by taking the sum of the firearm injuries and fatalities and the population weighted average of SVI and neighborhood characteristic variables. We excluded one neighborhood missing SVI from our analyses, leaving a final sample of 75 neighborhoods in Pittsburgh.

A negative binomial regression model was used to estimate the relationship between SVI and counts of firearm injuries and fatalities, with neighborhood population included as a covariate. Given the scope of our study and after examining separate models for firearm injuries and firearm fatalities, we chose to use the sum of both firearm injuries and fatalities as our outcome because of the small number of firearm fatalities and since there were not major differences between models. We then used the residual percentile method to identify neighborhoods with lower and higher counts of firearm injuries and fatalities than would be expected based on SVI (Figure 1) [4]. We considered neighborhoods with model residuals in the lowest quartile (lower counts than predicted) to be resilient and neighborhoods with model residuals in the highest quartile to be at-risk. T tests and Mann-Whitney U tests were used depending on the distribution of the data to compare firearm injury and fatality rates and the 15 neighborhood demographic, socioeconomic, housing, and transportation characteristics used in SVI between resilient and at-risk neighborhoods. In secondary analyses, we examined differences between resilient and at-risk neighborhoods for 158 US Census ACS variables. Two-tailed tests were used, and p<0.05 was considered significant. ArcGIS Pro 2.8.0 (Esri Inc) and SAS 9.4 (SAS Institute Inc) were used for all geospatial and statistical analyses. Institutional Review Board approval was not obtained as the data used are aggerated at the census tract and are publicly available.

Figure 1.

Identification of resilient and at-risk neighborhoods with lower- and higher-than-expected firearm injuries and fatalities based on neighborhood social vulnerability, Pittsburgh, Pennsylvania, 2015–2020.

RESULTS

On average, resilient neighborhoods (N=19, bottom quartile residuals) had 46.26 fewer and at-risk neighborhoods (N=19, top quartile residuals) had 54.58 greater firearm injuries and fatalities per 10,000 population from 2015–2020 than predicted based on neighborhood social vulnerability. Firearm injury (p=0.0005), fatality (p=0.0007), and the sum of injury and fatality (p=0.0002) rates per 10,000 population were significantly lower in resilient neighborhoods compared to at-risk neighborhoods (Table 1A). Resilient neighborhoods did not differ significantly from the at-risk neighborhoods for the US Census-based neighborhood demographic, socioeconomic, housing, and transportation characteristics included in the Social Vulnerability Index (Table 1B). In secondary analyses, we found that of the 158 census variables analyzed, only percent males never married (p=0.049) and percent females widowed (p=0.04) were significantly different between resilient and at-risk neighborhoods (Table 1C).

Table 1.

Comparison of firearm variables, Social Vulnerability Index variables, and significant US Census characteristics between resilient and at-risk neighborhoods, Pittsburgh, Pennsylvania, 2015–2020.

	Resilient Neighborhoods (N=19)c	At-risk Neighborhoods (N=19)c	p-value
A. Firearm Variables a
Firearm injury rate	25.41 (35.56)	82.30 (57.41)	0.0005
Firearm fatality rate	7.04 (8.95)	27.63 (26.48)	0.0007
Firearm injury and fatality rate	32.45 (43.25)	109.93 (76.88)	0.0002
B. Social Vulnerability Index Variables b
Social Vulnerability Index	0.68 (0.19)	0.65 (0.28)	0.91
Demographic characteristics
% Aged 65 years or older	14.26 (6.61)	14.25 (6.75)	0.72
% Aged 17 years or younger	17.62 (9.86)	20.24 (10.97)	0.44
% Racial and ethnic minority population	45.06 (28.51)	61.16 (39.11)	0.24
% Speaking English less than well	1.82 (2.32)	0.66 (0.77)	0.10
% Single-parent households	10.29 (9.97)	15.16 (10.03)	0.09
% Civilians with a disability	15.15 (7.19)	17.01 (7.15)	0.43
Socioeconomic characteristics
Per capita income	22,940.41 (11,139.62)	30,498.86 (25,266.70)	0.48
% Without a high school diploma	11.82 (5.47)	9.94 (6.65)	0.35
% Unemployed	11.03 (7.74)	13.56 (8.70)	0.34
% Below poverty	28.31 (18.26)	29.03 (15.05)	0.64
Housing and transportation characteristics
% Multi-unit structures	20.39 (20.03)	18.98 (24.86)	0.47
% Mobile Homes	0.61 (0.90)	0.21 (0.35)	0.10
% Crowded units	1.05 (1.11)	1.13 (1.30)	0.87
% Living in group quarters	11.08 (23.49)	6.29 (11.37)	0.33
% With no vehicle in household	31.84 (18.90)	31.73 (19.17)	1.00
C. Significant US Census Variables e, f
% Males never married	33.64 (11.76)	26.88 (10.19)	0.049
% Females widowed	9.54 (3.68)	12.72 (5.48)	0.043

^aRates are per 10,000 population for 2015–2020

^bBased on 2012–2016 American Community Survey 5-year estimates

^cValues presented as mean (standard deviation)

^dBoldface indicates statistical significance (p<0.05)

^eBased on 2014–2018 American Community Survey 5-year estimates

^f158 variables included in analyses

DISCUSSION

We found that neighborhoods resilient and at-risk for firearm injuries and fatalities differed for only 2 of the 158 US Census-based neighborhood characteristics included in our analyses. Our findings may suggest existing sources of secondary data on neighborhoods fail to adequately capture potential factors that prevent against firearm violence and related harms. By using a focused geographical approach, we identified unique local patterns that may confer risk or resilience. Our findings suggest that additional data on community assets related to firearm violence prevention is needed. For example, key themes for preventing firearm violence in communities can include role models, social cohesion, homeownership, and community-based organizations [7]. Similarly, looking beyond traditional US Census-based neighborhood measures may be crucial in understanding factors that promote neighborhood resilience. Shifting focus away from measures of neighborhood disadvantage toward strengths and assets-based approaches, may be important for driving community-based prevention efforts [8].

Future research could benefit from additional measures to better differentiate at-risk from resilient communities. Examining community-based organizations’ activities, reach, and programs is crucial [13–15]. Furthermore, measuring collective efficacy and social cohesion could clarify community-level protective factors against firearm violence [16–18]. While the Social Vulnerability Index (SVI) summarizes community vulnerabilities, it does not capture essential social dynamics like trust and mutual support [19]. Clarifying overlaps and distinctions between SVI and social measures can enhance targeted community interventions.

Limitations

An important limitation of our study is the use of GVA data, which has been shown to have gaps, particularly with firearm injuries [11, 20]. Moreover, we were unable to examine data on intent, age, sex, and other characteristics of individuals involved in each case. While there may be gaps due to data collection through public, law enforcement, and media resources, data were used due to the study’s exploratory nature and because no other publicly available data exists on firearm injury and fatality counts in Pittsburgh at the census tract level. Moreover, strengths of this dataset are that it does not rely on one source and is available at the point level, which allowed for more relevant analysis to the community being studied by using culturally defined neighborhoods in Pittsburgh. While our analyses focused exclusively on Pittsburgh, limiting generalizability to other geographies, we believe these findings lay the foundation for future community-engaged work to reduce firearm violence and related harms within Pittsburgh. Moreover, this geographically focused approach can be translated to different local contexts to identify areas of risk and resilience and perform targeted interventions. Ultimately, despite our study’s limitations, the lack of significant findings between resilient and at-risk neighborhoods with the US Census data highlights the need for additional data on factors that promote community resilience to firearm violence.

Public Health Implications.

Identifying neighborhoods as resilient for firearm injuries and fatalities beyond traditional measures has direct implications for public health interventions and resource allocation. This study demonstrates the inadequacy of existing secondary data in identifying neighborhood factors that measure resiliency and thriving and underscores the need for additional data on factors promoting community resilience. This opens avenues for community-engaged work in Pittsburgh to address firearm violence effectively. The identification of resilient neighborhoods highlights the need for a shift from a disadvantage focused approach to a strengths and assets-based perspective in public health strategies for firearm injury prevention.