Estimating Community Disruption from Nighttime Gunshots in 6 U.S. Cities, 2015 to 2021

Abstract

Background

Gunshots affect those directly involved in an incident and those in the surrounding community. The community-level impact of nighttime gunshots, which may be particularly disruptive to the sleep of nearby community members, is unknown.

Objective

Our aim is to estimate the number of people potentially affected by nighttime gunshots and the relationship between nighttime gunshots and median household income in the USA.

Design

We collected publicly available data on the timing and location of gunshots in six U.S. cities (Baltimore, MD; Boston, MA; Washington, D.C.; New York, NY; Philadelphia, PA; and Portland, OR) from 2015 to 2021. We then analyzed the data by computing rate ratios (RRs) to compare the frequency of gunshots during nighttime hours (6:00 pm to 5:59 am) versus daytime hours (6:00 am to 5:59 pm). Additionally, we used geospatial mapping to create choropleth maps to visualize the variation in nighttime gunshot density across cities. We estimated, using city-wide population, person-nights potentially impacted by the sound of gunshots within areas of 0.2- (low) and 0.5-mile (high) radius. Finally, for five of six cities where data on median household income were available by census tract, we built nonlinear regression models to estimate the relationship between the number of nighttime gunshots and median household income.

Key Results

We analyzed 72,236 gunshots. Gunshots were more common during the nighttime than daytime (overall RR = 2.5). Analyses demonstrated that the low estimates for the mean annual number of person-nights impacted by nighttime gunshots were 0.4 million in Baltimore and Portland, 1.3 million in Philadelphia, 1.6 million in Boston, 2.9 million in New York City, and 5.9 million in Washington. The number of nighttime gunshots was inversely related to median household income.

Conclusions

Nighttime gunshots are prevalent, particularly in low-income neighborhoods, and may have under-recognized effects on the surrounding community.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11606-024-08707-9.

Introduction

In 2021, approximately 48,000 people in the United States (U.S.) died from firearm-related injuries, which are now the leading cause of death among U.S. children and adolescents.^1,2 The U.S. rates of firearm homicide and suicide are 25 and 8 times higher, respectively, than those of other high-income countries.³ While the number of people killed every year by guns is well-documented, the sequelae of gun violence is profound and far-reaching and far less discussed, ranging from adverse health outcomes for family members of those who die or are injured⁴ to trauma associated with chronic threats to safety.⁵ One example of the underexplored sources of trauma is the sound from the bullet and associated emergency response (e.g., police, ambulance), which is troubling regardless the time of day; nighttime gunshots have the potential to serve as an added stressor by potentially disrupting sleep and circadian rhythms among those residing in nearby communities. Characterizing the prevalence of nighttime gunshots, and those residing in the nearby area within hearing distance, for whom sleep may be disrupted, could advance understanding of the scale of community disruption due to gun violence and inform trauma-informed clinical care of those impacted.

The noise emitted by a firearm is substantial, ranging from 156 to 170 decibels,⁶ and may be heard at more than a 1-mile radius.⁷ The subsequent sirens, lights, and other noises that result from the emergency response also potentially contribute to community-level disruption, by awakening people and making it difficult to fall back asleep. The light exposure when people are awake at night disrupts circadian rhythms.⁸ Even though a substantial portion of violent crimes take place at night,^9–11 little research has attempted to quantify the prevalence of nighttime gunshots and the attendant toll in terms of the emergency response on subsequent sleep and circadian disruption of members of the surrounding community.

The impacts of exposure to gun violence are profound, particularly for youth and adolescents.¹² Youth exposed to gun violence have a high risk of demonstrating trauma symptoms.¹³ Exposure to firearm-related violence has long-term implications, including increased risk of hypertension more than 10 years later;¹⁴ exposure to violence is associated with a 1.5 times greater risk of coronary heart disease in Black individuals, but not in White individuals.¹⁵ Blood pressure non-dipping, a known cardiovascular risk factor in which blood pressure does not have the normal nighttime decrease,¹⁶ is more common in those with exposure to violent crimes.¹⁷ Moreover, gunshot exposure disproportionately affects low income and minority race/ethnic communities. Black youth are 4 times more likely to observe firearm-related violence than White youth.^13,18,19 Per capita proportion of gunshot wounds is correlated with measures of community deprivation (e.g., poverty, unemployment),²⁰ and U.S. counties with the highest poverty concentrations have increased rates of firearm-related deaths.²¹

Nighttime gun violence exposure has the additional potential to disrupt sleep and circadian rhythms, which are essential for physical and mental health, daytime cognition, and emotional regulation.²² Insufficient sleep and/or circadian rhythm disruption impairs next day performance and/or learning,^23–25 and increases the risk for high blood pressure, heart attacks, stroke obesity, insulin resistance, and diabetes.^26–31 Of particular relevance, disrupted sleep (including insomnia) involving significant nocturnal wakefulness is associated with cognitive and behavioral dysregulation (e.g., delusional thinking and suicidal ideation) and increased risk of some psychiatric conditions.^22,32 Insufficient sleep in adolescents increases the risk of aggressive behaviors, such as getting in a fight, and potentially high-risk behaviors, such as carrying a gun.³³ Adolescents who self-reported sleepiness at 15 years of age were 4.5 times more likely to have a criminal conviction by age 29 compared with children without sleepiness.³⁴

Insufficient sleep and sleep disturbances are more common among race/ethnic minorities,³⁵ and insufficient and/or disturbed sleep is a mediator of the relationship between environmental factors (e.g., pollution, housing, or food insecurity) and disparities in health outcomes between race/ethnic groups.³⁶ Nighttime gunshots may be a specific environmental factor contributing to poor sleep and subsequent health disparities in underserved communities. We address this gap in the literature by examining the proportion of nighttime versus daytime gunshots in major U.S. cities where publicly available data tracking gunshot timing and location were available and quantified the potential number of residents near the nighttime gunshot for whom sleep and circadian rhythms were therefore potentially disrupted. We also explored the relationship between nighttime gunshots and median household income to understand if historically underserved communities shoulder a disproportionate burden of nighttime gun violence than do more economically privileged areas.

Method

Study Overview

Using U.S. Census Data,³⁷ we identified the 30 most populated cities in the U.S. We then queried each city’s government websites for publicly available data documenting date, time, and location information (either latitude/longitude or police precinct) of gunshots. Of these 30 cities, six—Baltimore, Maryland; Boston, Massachusetts; Washington, District of Columbia (D.C.); New York, New York; Philadelphia, Pennsylvania; and Portland, Oregon—had such publicly available datasets (Supplement A).

The method of data collection in each city varied and could include reporting of gunshots identified either through emergency (e.g., 9–1-1) calls, observations from law enforcement officers, or use of gunshot detection surveillance technology system utilized in many U.S. cities (ShotSpotter®, details in Supplement B),³⁸ or a combination of these sources. The method of data collection and reporting varied in each city and is detailed further, by city, below. In this analysis, we utilized all reported gunshot event data in each city. Given our focus on community effects of the gunshot events, we removed duplicates from the datasets: duplicate gun-related events were defined as gunshots taking place in the same location (latitude/longitude or, in the case of Boston, police precinct) within 60 s of one another. We have previously reported using these data sets in which duplicate gun-related events were included (so that the total number of gunshot events was analyzed).³⁹ The years with the greatest amount of overlap in available data for purposes of this analysis were the years 2015 to 2021, except for Portland, which made data available starting in 2019.

Gunshot Data by City

The dataset from Baltimore included events in a crime database that did not specify whether there were or were not victims. The dataset included reporting of date, time, and latitude and longitude coordinates of gunshots. The Baltimore dataset included 4894 total gunshots, 835 of which were duplicated and removed, leaving 4059 events for the present analysis. Two datasets from Boston were obtained, one that included gunshots that resulted in victims and one that included gunshot events that did not result in victims. Each dataset contained date, time, and police jurisdiction of gunshots. Across both datasets, 8326 gunshots were included, 327 of which were duplicated and removed, leaving 7999 events for this analysis. The Washington D.C. dataset included the date, time, and latitude and longitude coordinates of ShotSpotter®-detected gunshots, including 41,805 events, 1345 of which were duplicated and removed, leaving 40,460 events for this analysis. The New York City dataset included only gunshot events that resulted in at least one victim and included the date, time, and latitude and longitude coordinates for shootings. The New York dataset included 9496 gunshots, 1949 of which were duplicated and removed, leaving 7547 events for this analysis. The Philadelphia dataset included only incidents with victims and contained the date, time, and latitude and longitude coordinates for gunshots. The Philadelphia dataset included 11,428 gunshots, 927 of which were duplicated and removed, leaving 9521 events for this analysis. The Portland dataset included only incidents with victims and made available date, 2-h time window (e.g., midnight–1:59 am), and latitude and longitude coordinates for shootings. The Portland dataset included 2650 events with no duplicates. The combined dataset across all six cities included 72,236 gunshots. As noted, because data collection in each city varied substantially, this study is not intended to explore inter-city comparisons, but rather focuses on within-city analyses. Supplement C describes the gunshot, location, and timing data obtained from each city in greater detail.

Gunshot deaths for each city were obtained from the U.S. Centers for Disease Control and Prevention.⁴⁰

Demographic Data (Population and Median Household Income)

Demographic data (population and median household income) by census tract were obtained for Baltimore; Washington, D.C.; New York City; Philadelphia; and Portland from the U.S. Census Bureau and American Community Survey (ACS).⁴¹ The number of census tracts total 199 in Baltimore; 206 in Washington, D.C.; 2327 in New York City; 408 in Philadelphia; and 197 in Portland. The ACS datasets for the years 2015–2021 were used in the case of Baltimore, Washington D.C., New York City, and Philadelphia. The ACS datasets for the years 2019–2021 were used in the case of Portland. Demographic data by police precinct were obtained from the City of Boston. Since Boston does not report the precise location (i.e., longitude/latitude) of gunshots, we did not include Boston in the analysis exploring nighttime gunshots by median household income.

For all cities, demographic and gunshot data were matched by year, so that the gunshot data were matched with the corresponding demographic data available in the same year as the recorded gunshot. Details describing the source and components of the demographic data used in this analysis can be found in Supplement D.

Data Linking to Area for Choropleth Map Creation

In the case of Washington D.C., Baltimore, New York City, Philadelphia, and Portland, the variable used to match population data with gunshot data was census tracts. We report analyses by recognizable neighborhoods within each city (e.g., Manhattan in New York, NY) for ease of interpretation, which are comprised of census tracts. For Boston, population and gunshot data were provided by police precinct, which generally match recognizable neighborhoods. Neighborhoods within each city range in size from 0.01 square miles (Healy Heights, OR) to 17.01 square miles (Northeast Philadelphia, PA).

Statistical Analysis

We classified gunshots as nighttime (6:00 p.m. until 5:59 a.m.) or daytime (6:00 a.m. until 5:59 p.m.).

Mean annual firearm-related deaths by city were calculated by taking the average gun deaths reported each year for years 2015 through 2020 to match the availability of gunshot data. Firearm-related deaths for 2021 were not available at the time of writing. We calculated the population-adjusted rate of gunshots per 100,000 in each city.

We calculated rate ratios (RRs) to compare the mean annual nighttime to daytime gunshots by city. We calculated the number of gunshots observed in each 2-h time interval and normalized these estimates by dividing them by the total gunshots in each city.

To analyze the impact of nighttime gunshots, we defined “person-nights” as the number of individuals estimated to be within earshot of the nighttime gunshot. The number of persons within earshot was estimated from the city population over the study interval using data from the U.S. Census (in the case of Baltimore; Washington, D.C.; New York City; Philadelphia; and Portland) or data from the City of Boston (Boston). We generated a low and high estimate of the person-night impact by including the estimated number of people within a circular perimeter of 0.2-mile (low estimate) and 0.5-mile (high estimate) radius based on the city density in which the gunshot occurred. The estimates of impacted person-nights were computed by multiplying yearly average nighttime gunshots, city population density, and the computed affected area for low and high estimates.

We created Choropleth maps using heatmap techniques to display the average annual person-nights impact by neighborhood within each city. Each map was divided into their respective neighborhoods as described above. Estimates within neighborhoods were grouped into quantiles, each containing an equal number of observations.

Finally, we used nonlinear regression models of the relationship between the number of nighttime gunshots (dependent variable) and median household income (independent variable) in all cities except Boston, which did not have the appropriate data. For Baltimore; Washington, D.C; New York City; Philadelphia; and Portland, we summed the number of gunshots by census tract and then merged gunshot data and median household income data to perform the regression analysis. We assume the independence of observations in the construction of the model. We observed that the distribution of gunshot data was right-skewed. We conducted diagnostics and selected a logarithmic transformation of the variable detailing the number of gunshots for the regression model while keeping median household income data in its original scale. This selection has the least residual standard error and highest adjusted-R squared values. For instance, the regression equation for each city will look as follows:

$$\text{ln}\left(\mathrm{nighttime}\mathrm{gunshots}\right)={\beta }_0+{\beta }_1\times \mathrm{median}\mathrm{income}$$

We report the transformed coefficients. The regression formula implies that one unit increase in median income is associated with a change ($e^{{\beta }_1}-1)\times 100\%$ in number of nighttime gunshots. All analyses were conducted in R (R Core Team, www.R-project.org).

Results

We analyzed 72,236 gunshots across these 6 cities that occurred between 2015 and 2021 (Table 1). Across all cities, nighttime gunshots were more common than daytime gunshots (RR = 2.5). City-specific RRs were as follows: Baltimore, RR = 2.0; Boston, RR = 4.1; Washington D.C., RR = 2.6; New York City, RR = 2.8; Philadelphia, RR = 1.7; Portland, RR = 2.7. The population adjusted rate of gun deaths ranged from 3.2 deaths per 100,000 persons in New York City to 44.2 deaths per 100,000 persons in Baltimore  (Table 1). Nighttime gunshots were concentrated on Saturday and Sunday nights in all cities (Fig. 1).

Table 1.

The Number of Daytime Gunshots (6:00 a.m.–5:59 p.m.) and Nighttime (6:00 p.m.–5:59 a.m.), the Rate Ratio of Nighttime to Daytime Shots, and the Mean Annual Gun Deaths by City

	Daytime and nighttime gunshots*				Mean annual gun deaths	Mean population (in thousands)	Rate (gun deaths per 100,000 persons)
	Day	Night	Total	Night/day rate
City	N (%)	N (%)	N	Ratio
Baltimore	1340 (33%)	2719 (67%)	4059	2.0	266	601	44.2
Boston	1564 (20%)	6435 (80%)	7999	4.1	48	679	7.1
Washington, D.C	11,316 (28%)	29,144 (72%)	40,460	2.6	127	694	18.3
New York City	2001 (27%)	5546 (73%)	7547	2.8	271	8479	3.2
Philadelphia	3504 (37%)	6017(63%)	9521	1.7	366	1583	23.1
Portland	722 (27%)	1928 (73%)	2650	2.7	82	649	12.6

Gunshot and population data for all cities except Portland are for 2015–2021. Data from Portland were only available 2019–2021. Available gunshot data varied by data; graphs should not be used for inter-city comparisons, but rather to focus on within-city analyses

The total row for columns detailing gunshots includes a sum total of the day, night, and total gunshots in the sample. These totals are used to calculate the overall night/day rate ratio of daytime/nighttime gunshots

Gun death data was obtained from the Centers for Disease Control and Prevention(citation #40) and includes the following counties in each city: Baltimore: Baltimore City; Boston: Suffolk County; Washington, D.C.: Washington, D.C; New York City: New York County, Bronx County, Kings County, Queens County, Richmond County; Philadelphia: Philadelphia County; Portland: Multnomah County. Mean annual gun deaths per city were computed as the average of annual gun deaths for the years 2015-2020 in all cities. The mean annual gun deaths and mean population (in thousands) across the same interval (2015-2020) were used to calculate the rate of gun deaths per 100,000 persons in each city

Figure 1.

Heat map depicting gunshots by time and day of week by city (n = 72,236 gunshots across all six cities). Notes: Darker red color indicates more gunshots. Gunshot and population data for all cities except Portland are for 2015–2021. Data from Portland were only available 2019–2021. Available gunshot data varied by data; graphs should not be used for inter-city comparisons, but rather to focus on within-city analyses.

Based on low (0.2-mile radius) and high (0.5-mile radius) estimates of distance from which a gunshot could be heard, estimated annual person-nights affected were between 0.4 and 2.2 million in Baltimore; 1.6 and 10.1 million in Boston; 5.9 and 36.9 million in Washington; between 2.9 to 18.1 million in New York City; 1.3 and 8.1 million in Philadelphia; and 0.4 and 2.5 million in Portland (Table 2).

Table 2.

Characteristics of Nighttime Gunshots by city Population Density and Estimated Low (0.2-mile Radius) and High (0.5-mile Radius) Person-Nights per year Impacted by Nighttime Gunshots

	Number of gunshots reported		Population		Estimated mean annual person-nights impacted by gunshots
	Total gunshots 2015–2021*	Mean annual nighttime gunshots	Population (100,000 people)	Density (1000 people per square mile)	Low impact estimate	High impact estimate
Baltimore	2719	388	586	7.2	352,809	2,205,059
Boston	6435	919	676	14.0	1,616,096	10,100,602
Washington, D.C	29,144	4163	690	11.3	5,908,844	36,930,274
New York City	5546	792	8804	29.1	2,895,301	18,095,630
Philadelphia	6017	860	1604	12.0	1,292,527	8,078,291
Portland	1928	643	653	4.9	395,121	2,469,504

Low impact estimates computed by including the population within a 0.2-mile radius from each gunshot; high impact estimates included the population within a 0.5-mile radius

Gunshot and population data for all cities except Portland are for 2015–2021. Data from Portland were only available 2019–2021. Available gunshot data varied by data; data should not be used for inter-city comparisons, but rather to focus on within-city analyses

The highest estimated prevalence of person-nights impacted by nighttime gunshots by neighborhood and city included the following: West Baltimore (Baltimore; n = 330,000, Supplement E, Panel A); Roxbury (Boston; n = 2.9 million, Supplement E, Panel B); Eighth Ward (Washington D.C., n = 15.6 million, Fig. 2, Supplement E); Bronx (New York City; n = 2.8 million, Fig. 2, Supplement E); East Philadelphia (Philadelphia; n = 939,000, Fig. 2, Supplement E); and Hazelwood (Portland; n = 219,000, Fig. 2, Supplement E). The lowest estimated prevalence of person-nights impacted by nighttime gunshots by neighborhood and city included: Southeast Baltimore and Northern Baltimore (Baltimore; n = 165,000 each, Fig. 2, Supplement E); Charlestown (Boston; n = 88,000, Fig. 2, Supplement E); Second Ward (Washington D.C.; n = 355,000, Fig. 2, Supplement E); Manhattan and one area of Queens (New York City each; n = 0, Fig. 2, Supplement E); in Philadelphia were in Northwest and in Northeast Philadelphia (n = 19,000 each, Fig. 2, Supplement E); and in Portland were in multiple districts (n = 0, Fig. 2, Supplement E).

Figure 2.

Choropleth maps depicting mean annual person-nights (in thousands) impacted by nighttime gunshots by neighborhood (panels A, C, D, E, F) and neighborhood/police precinct (panel B) in each city. Notes: The choropleth maps display high impact estimates (0.5 miles radius) of nighttime gunshots per year in terms of average annual person-nights impacted, in thousands of person nights. Estimates within neighborhoods were grouped into quantiles, each containing an equal number of observations. Darker colors indicate higher values. Gunshot and population data for all cities except Portland are for 2015–2021. Data from Portland were only available 2019–2021. Available gunshot data varied by data; graphs should not be used for inter-city comparisons, but rather to focus on within-city analyses. Please refer to Supplement E for the neighborhoods that correspond to the numbers on each map.

There was an inverse relationship between the log-transformed number of nighttime gunshots and median household income by census tracts within each city (Fig. 3). The regression coefficients for change in income per city were as follows: Baltimore (− 0.02, 95% CI − 0.03, − 0.02); Washington, D.C. (− 0.03, 95% CI − 0.03, − 0.02); New York City (− 0.01, 95% CI − 0.02, − 0.01); Philadelphia (− 0.03, 95% CI − 0.4, − 0.03); Portland (− 0.03, 95% CI − 0.04, − 0.03). For instance, the coefficient (− 0.02) for Baltimore City may be interpreted as follows:

$$100·\left(e^{{\beta }_1},-,1\right)=100·(e^{-0.02}-1)=-2.37$$

Figure 3.

Number of nighttime gunshots versus median household income by city. Notes: Data from each census tract are plotted using red dots. The regression line is depicted in blue and 95% confidence interval for the regression line is shown in gray. We used nonlinear regression models of the relationship between log-transformed number of nighttime gunshots and median household income per census tract in each city. Gunshot and income data for all cities except Portland are for 2015–2021. Data from Portland were only available 2019–2021. Available gunshot data varied by data; graphs should not be used for inter-city comparisons, but rather to focus on within-city analyses.

Therefore, a one unit increase in median household income ($1000) would be associated with a 2.37% decline in nighttime gunshots.

Discussion

This cross-sectional study of the timing and location of gunshots in six U.S. cities demonstrates that gunshots were approximately two times more likely at night as compared to during the day. Using geospatial analysis, our study combined population data with geolocation data detailing the latitude and longitude of nighttime gunshots to estimates the person-night impact of nighttime gunshots, or in other words, an estimate of the number of people for whom a night may be disrupted due to gunshots. Our findings suggest that between 12.5 million and 77.9 million person-nights might be disrupted each year in just these six cities due to nighttime gunshots. Finally, non-linear regression revealed an inverse association between median household income and nighttime gunshots. For example, in Baltimore, each additional unit of median household income ($1000) was associated with a ~ 2% reduction in nighttime gunshots.

Our geospatial analysis demonstrates that nighttime gunshots were more common in economically vulnerable areas in each city. Our findings contribute to the strong evidence that gun violence disproportionately impacts low income and racial/ethnic minority communities,^13,18,19 Our study quantifies the prevalence of nighttime gunshots and its estimated impact on person-nights in these communities. Nighttime gunshots could be expected to increase the risk of sleep disruption. One or two nights of sleep deprivation can lead to short-term impacts, such as performance decrements⁴² and impaired immune response,⁴³ and chronic sleep deprivation can have deleterious impacts on learning and performance,²⁴ blood pressure,^26,29,44 and risk for chronic conditions, including obesity³⁰ and diabetes,³¹ and risk for a heart attack²⁸ or stroke.²⁷ Nighttime gunshots could also be expected to increase the risk of circadian rhythm disruption, or the interruption of cyclical (in a 24-h pattern) patterns of sleep, eating, and activity and their associated biological processes (e.g., hormone secretion). When experienced in the long term, circadian rhythm disruption is associated with hypertension, depression, kidney disease, and Parkinson’s disease.⁴⁵ Our work suggests that nighttime violence is prevalent in certain neighborhoods and is a potential chronic sleep disruptor. Future quantitative and qualitative research is needed to characterize the extent to which sleep itself is disrupted by nighttime gunshots.

Given that nighttime gunshots are more common in communities characterized by low median household income, and racial/ethnic minority individuals are more likely to reside in low income areas,⁴⁶ nighttime gun violence—and the resulting sleep and circadian rhythm disruptions and their consequences—could be an accelerator of health disparities. Future research is needed to test the hypothesis that sleep and circadian rhythms are disrupted by nighttime gunshots. Our study takes an important first step by documenting the higher relative frequency of nighttime gunshots and their possible community impact in terms of expected nighttime disruption of many individuals not directly involved in the gun violence. If nighttime gunshots do disrupt sleep and circadian rhythms among those residing nearby, then nighttime gunshots would represent an added social determinant of health, which, in vulnerable communities, would be an added burden on top of the well-documented determinants, including discrimination, historical disinvestment, police violence, poverty, residing in food swamps and deserts, and reduced access to healthcare that result in health disparities.^15,47

Consistent with the Society for General Internal Medicine (SGIM) position statement on an internist’s role in the social determinants of health, there are potential clinical implications of the presented findings and clear areas where future research is needed to support clinicians caring for patients affected by this violence.⁴⁸ Our findings illuminate a potential large barrier to sleep, particularly in underserved communities. The provision of trauma informed clinical care, then, would necessitate an understanding of this stress and the development of interventions to improve sleep in populations most affected. Providers could gain such an understanding of neighborhood-level stressors, such as nighttime gunshots, through communication that allows time for open-ended conversations in the clinical encounter about community and environmental barriers to sleep, health, and well-being.⁴⁹

In cities across the U.S., the number of gun-related deaths is often used to describe the scope of firearm-related violence; our analysis suggests suggest that the number of gun deaths alone substantially undercounts the true toll and community impact of this violence. For example, Boston is often considered one of the safest cities in the safest states in the U.S. given its relatively low number of annual firearm-related deaths (an average of 48 deaths per year in the city between 2015 and 2020).⁵⁰ However, this work estimates that orders of magnitude more people (at least 1.6 million person-nights affected annually in Boston) are within hearing range of nighttime gunshots. Therefore, the physical injuries associated with firearm violence may represent only a fraction of the impact of gun violence within communities. The potential sleep loss and circadian and other disturbances from nighttime gunshots has far-reaching implications for detrimental health outcomes in vulnerable communities.

It is notable that only 6 of the 30 most populated cities in the U.S. publicly reported sufficient data describing gunshot location and timing data to be included in this analysis, and the data provided by those 6 cities was heterogeneous. Improving the public reporting of the timing of all known gunshots would enable enhanced focus on this issue and promote research and inform policy makers who are in positions to enact policies to mitigate gun violence. The SGIM’s 2020 national research strategy recommendations to reduce firearm-related injury and death specifically called for facilitating access to improved, comprehensive data sources that would facilitate the development of reliable evidence. ⁵¹

Our study has several limitations. First, the person-night calculations do not indicate if sleep or circadian rhythms were disrupted, only that the gunshot took place during nighttime hours. Second, the findings are limited to the 6 cities that report the data required for the present analysis. It is unknown whether these findings are generalizable to other cities. Third, given the nature of these publicly available datasets, we are unable to verify whether these represent complete, precise reports of all gunshot events. Fourth, gunshot data for Washington, D.C. (entirely) and Boston (in part) were facilitated through use of ShotSpotter®, a for-profit company. In many instances, ShotSpotter® does not provide coverage for an entire city. Instead, sensors are placed selectively throughout a city, and could be disproportionately concentrated in low income areas. Moreover, it is possible that ShotSpotter® overestimates the gunshots observed in the communities where they are installed. If this is the case, inaccurately registering a gunshot still results in the dispatching of emergency response, which extracts a toll on city resources and presents further concerns regarding the noise disruption and associated sleep/circadian impact. That independent assessments of ShotSpotter® technology’s false positive and false negative rates are not, to our knowledge, publicly available calls attention to the incorporation of proprietary information into cities’ public safety and emergency response infrastructure. Fifth, while our geospatial analytical approach has strengths, the person-nights impacted by gun violence determined here based on averages of city-wide density are imperfect estimates of those impacted by nighttime gun violence as residents are not uniformly distributed across each city. Sixth, each city reports different metrics; therefore, these data should not be used to make direct comparisons between cities; trends within cities are the focus of our analyses. Finally, we did not have access to exact estimates of the distance with which a gunshot could be heard in an urban environment. The number of tall buildings versus open green space in the area surrounding each shot would alter the actual area of person-nights impacted by each gunshot. For instance, it is possible that in a region densely populated with tall buildings and busy traffic, the sound of a gunshot might be heard over a smaller area than if the gunshot occurred in a place near an open green space. Our use of two possible radii offers low and high estimates to reflect this uncertainty.

Future research also may consider the incorporation of quantitative sleep data collection measures, such as wrist-worn actigraphy, to capture the actual sleep and circadian rhythm disruption of those residing in areas that experience high burdens of nighttime violence. This would facilitate more direct quantification of the impact of nighttime gun violence on sleep, circadian rhythms, and other physiology. Future research may also collect qualitative data to explore how nighttime gunshots disrupt sleep and circadian rhythms and affect physical and mental health among residents of all ages. Coupling such analyses with longitudinal follow-up to explore the downstream impact of nighttime gun violence and sleep and circadian rhythms disruption, particularly in the areas experiencing high levels of nighttime gun violence on physical, mental, and emotional health and well-being of community residents and children’s learning in school, are important areas for future work.

Conclusion

Nighttime gunshots are prevalent and the resulting noise presents a possible risk for sleep and circadian rhythm disturbance among community members, not just the people directly involved in the violent act. Future research is needed to causally link nighttime gun violence to sleep interruption and resulting adverse health impacts. Nighttime gunshots may be an additional environmental hindrance to sleep, health, and well-being, particularly in economically vulnerable neighborhoods. A greater understanding of the ubiquity of nighttime gun violence in underserved communities, presented here, may inspire future research and practical efforts to forge inter-disciplinary care teams to support communities impacted by these events.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contribution:

All authors had full access to the publicly available data used in the present analysis and take full responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and Design

Robbins, Affouf, Masiakos, Griggs, Klerman, and Sacks.

Acquisition, Analysis, or Interpretation of the Data

Robbins, Affouf, Masiakos, Griggs, Klerman, and Sacks.

Drafting of the Manuscript

Robbins, Affouf, Masiakos, Iyer, Griggs, Klerman, and Sacks.

Critical Revision of the Manuscript for Important Intellectual Content

Robbins, Affouf, Masiakos, Iyer, Griggs, Klerman, and Sacks.

Statistical Analyses

Affouf.

Funding

This study was supported by the MGH Center for Gun Violence Prevention. Robbins has received grant support from the NIH/NHLBI (K01HL150339). Klerman has received grant support from the NIH (R01NS099055, U01NS114001, U54AG062322, R21DA052861, R21DA052861, R01NS114526-02S1, R01-HD107064), DoD (W81XWH201076), and Leducq Foundation for Cardiovascular Research.

Data Availability:

All data analyzed in the present study were obtained from publicly available sources, listed in Supplement D.

Declarations:

Conflict of Interest:

Robbins reports receiving consulting income from byNacht GmbH, Savoir Beds Ltd., Oura Health Oy, Castle Hot Springs, Sonesta International Hotels Corporation. Robbins has received funding from Bryte Labs. Klerman reports consulting income from the American Academy of Sleep Medicine Foundation, Circadian Therapeutics, National Sleep Foundation, Sleep Research Society Foundation, and Yale University Press; has received travel support from the European Biological Rhythms Society, EPFL Pavilion, the Santa Fe Institute, Sleep Research Society, and the World Sleep Society; and is an unpaid member of the scientific advisory board of Chronsulting. Klerman’s partner is founder, chief scientific officer of Chronsulting.