Contextual Factors Influencing Firearm Deaths Occurring Among Children

Data were obtained from the NFR-CRS, which collects and aggregates data during retrospective review of pediatric deaths occurring within a state. The origin of the NFR-CRS and CDR, including strengths and limitations, has been described.⁹ Although all 50 states, the District of Columbia, Guam, and Puerto Rico conduct CDR, NFR-CRS utilization differs. Data are entered into the NFR-CRS, which contains over 2200 variables.¹¹ Review criteria, review process, and level of missing data vary by state. This analysis was limited to firearm-associated deaths occurring among children aged 0 to 10 for all available years (2004–2020) in 41 states with available data. For this analysis, a firearm-associated death was defined as any fatal firearm injury caused by a weapon using a powder discharge to fire a projectile (eg, handguns).

Individual-level data elements include child demographic factors (age, sex, race, ethnicity), child psychosocial factors, primary caregiver relationship, and caregiver psychosocial factors. For children where 2 primary caregivers were listed, the first primary caregiver was included.

As a social construct, abstracted race classifications of Alaskan Native, American Indian, Asian, Black, Native Hawaiian, Pacific Islander, and white and ethnic classification of Hispanic/Latino/a origin were collapsed into non-Hispanic white, non-Hispanic Black, Hispanic, and “additional races” for analysis given low base rates within other categories.

These included a history of child protective services (CPS) involvement, (ie, either sibling(s) removal from home or open CPS case) and personal history of child maltreatment (ie, physical, neglect, sexual, emotional or psychological).

Relationship of the primary caregiver included biologic childbearing parent, biologic nonchildbearing parent, adoptive parent, stepparent, foster parent, parent’s partner, grandparent, sibling, other relative, friend, and institutional staff. Biologic, step, adoptive, and foster parents were classified as “parent,” whereas all others were collapsed to “other adult.”

These included history of receipt of social services, intimate partner violence (IPV) victimization, substance use, and delinquency or criminal history. Receipt of social services was a dichotomous variable (“Yes or No”) incorporating services, including Women, Infants, and Children, home visiting, Temporary Assistance for Needy Families, Medicaid, food stamps, Supplemental Nutritional Assistance Program, or Electronic Benefit Transfer, Section 8 housing, Social Security Disability, and/or other. History of IPV victimization, substance use history, and criminal history were dichotomous variables in the dataset and based on CDR investigation of the period surrounding the child fatality.

Incident details included the location, urbanicity, supervision of the child during the incident, supervising person, and impairment of supervising person.

Not mutually exclusive, may include “child’s home”; relative’s home, friend’s home collapsed to “friend or relative’s home”; sidewalk, roadway, driveway, other parking area, state, county park, or other recreation area collapsed to “outdoors”; and licensed foster care home, relative foster care home, licensed group home, licensed childcare center, licensed childcare home, unlicensed childcare home, farm or ranch, school, Indian reservation or trust lands, military installation, jail or detention facility, hospital, and other collapsed to “other location.”

Categories included urban, suburban, rural, and frontier. Rural and frontier were collapsed into a single “rural” variable.

Categories included supervision present, no supervision - needed, and no supervision - not needed; the latter 2 were collapsed into “no supervision.”

The person responsible for supervision at the time of the incident, independent of whether they were present, variable included childbearing biological parent, nonchildbearing biological parent, adoptive parent, stepparent, foster parent, parent’s partner, grandparent, sibling, other relative, friend, acquaintance, hospital staff, institutional staff, babysitter, licensed childcare worker, and other. For analysis, supervising person was collapsed into “parent,” “other adult,” or “friend or sibling.”

Dichotomous variable assessed supervisor impairment from drugs, alcohol, distraction, absenteeism, illness, and/or disability at the time of the incident.

Firearm variable details included the firearm type, firearm owner, firearm storage, and firearm use(s) at the time of the incident.

Firearm type included handguns, shotguns, rifles, and 3D guns; handgun and 3D gun were collapsed to “handgun,” with shotgun and rifle collapsed to “long gun.”

Firearm owner included the child, caregiver, other family member, child’s significant other, friend or acquaintance, stranger, and other, with child’s significant other, stranger, and other collapsed to “other or stranger.” There were no instances where the child was the owner.

Firearm storage variables included firearm locked, unlocked, or unknown, and firearm loaded, unloaded, or unknown.

NFR-CRS firearm use(s) at the time of the incident (not mutually exclusive) included: self-injury, commission of crime, drug dealing or trading, drive-by shooting, random violence, child abuse, child was a bystander, argument, jealousy, IPV, hate crime, bullying, hunting, target shooting, playing with weapon, showing gun to others, Russian roulette, gang-related activity, self-defense, cleaning weapon, loading weapon, and other. Among response options representing less than 1% of the sample (eg, self-defense), researchers reviewed deaths for co-occurring use patterns, characterizing the death by those higher frequency options. Low frequency firearm uses without co-occurring higher frequency uses were recategorized and collapsed into “other.” For discrepancies or where “other” firearm use was indicated, the firearm use comments were reviewed to categorize the event. “Murder-suicide” was added as a firearm use based on the “other” comments.

Descriptive statistics were calculated for individual and contextual variables of child firearm deaths, noting missing or unknown data where applicable. Using R for analysis, a cluster analysis was performed examining contextual variables and nondemographic factors to identify unique subgroups, or clades, ending with defined variables or terminal leaves. Variable inputs for the cluster analysis included incident location, supervision, firearm type, and firearm use. There were 415 (26.2%) deaths with missing or unknown data among cluster analysis variables: 272 (17.2%) unknown firearm type, 168 (10.8%) unknown supervision during incident, 92 (5.8%) unknown firearm use, and 0 (0.0%) unknown location, with 117 (7.4%) decedents having multiple missing cluster analysis variables. Where cluster analysis variables were missing, we elected to perform case deletion. Variables were rotated and a matrix of Jaccard measures of similarity was computed to cluster variables with an average linkage function that was plotted in a dendrogram to identify the context of different types of scenarios during which firearm deaths occurred and, ideally, modifiable factors with opportunities for prevention. The Institutional Review Board at the University of Michigan determined the study to have an unregulated status.

Within the study period (2004–2020), 1582 firearm deaths among children (age 0–10) were documented within the NFR-CRS, of which 1167 (73.8%) had complete cluster analysis incident variables (ie, location, supervision, firearm type, firearm use) and were included for analysis. Descriptive characteristics of the child and caregiver are presented in Table 1. Most decedents were males with a mean age of 4.9 years (SD = 3.0), with CPS involvement in 106 (9.1%) of deaths and a history of prior child maltreatment in 216 (18.5%) children. The primary caregivers for most incidents were parents, with 218 (18.7%) caregivers reporting a history of IPV victimization. Of the 494 caregivers where receipt of social services was known, 255 (51.6%) had received assistance. Among caregivers where substance use status was known (n = 500), 166 (33.2%) had a history of substance use. Where known (n = 624), 138 (22.1%) caregivers were noted to have a criminal history.

Incident details of firearm deaths are presented in Table 2. The most common incident location was the child’s home (n = 805, 69.0%) in an urban location (n = 460, 39.4%) with supervision present (n = 868, 74.4%). Within the context of the fatal firearm incident, the person responsible for supervising the child during the incident was the parent in 891 (76.3%) deaths, another adult in 148 (12.7%), friend or sibling in 20 (1.7%), and unknown in 108 (9.3%). Where known (n = 696), 38.4% of supervising adults were noted to be impaired at the time of the firearm death. Firearms were stored loaded in 449 (92.4%) of the 486 deaths and unlocked in 611 (95.0%) of the 643 deaths where firearm storage methods were documented.

Firearm details are presented in Table 3. In over half the incidents (n = 621, 53.2%), the primary caregiver was the firearm owner. The most common firearm uses resulting in a child death included: crime, playing and showing a firearm to someone, and self-injury. Among child firearm deaths secondary to gang-related and/or drive-by or random shootings, the majority (n = 86-of-138, 62.3%) occurred in urban locations. Within child abuse and IPV firearm-related deaths, 22.3% (n = 42-of-188) of children had CPS involvement or a history of child maltreatment. Additionally, 28.1% (n = 31-of-110) of primary caregivers in murder-suicides reported IPV victimization. Firearms were stored unlocked in 84.2% (n = 319) and loaded in 79.2% (n = 300) of the 379 child deaths associated with playing and showing a firearm to someone.

Cluster analysis demonstrated several clades of co-occurring characteristics of firearm deaths among children, shown in Fig 1. Within 1 clade, terminal leaves include drive-by and random shootings, outdoor locations, and gang-related shootings. A second clustering clade highlights that child firearm deaths occurring by cleaning, hunting, or target shooting clustered with long-guns. Similarly, those deaths characterized as bystander child firearm deaths clustered with those occurring because of IPV, abuse, and/or a reported argument or jealousy. Another distinct cluster involved those within the child’s home while supervised by an adult, primarily involving a handgun. Some of these incidents were characterized as involving a crime. Within this overall clade, another clustering of firearm deaths occurring within the home involved those that were characterized as murder-suicide events. Finally, there was also a clustering of firearm deaths that occurred while the firearm was being shown or played with when not supervised; such deaths were often associated with occurrence in other homes (friend or family home).

The current analysis of almost 1200 fatal firearm deaths among children ages 0 to 10 across 41 states provides one of the first characterizations of the associated clustering of key contextual factors surrounding these tragic events. Understanding such contextual factors provides information on the pathways by which child firearm fatalities occur. Within this analysis, we identified 6 distinct clustering clades of firearm fatalities, including (1) community violence; (2) long-gun discharge while cleaning, hunting, or target shooting; (3) child abuse or IPV; (4) supervised handgun use; (5) murder-suicide; and (6) unsupervised playing or showing of firearms.

Clusters where the underlying intent was related to an unintentional cause (eg, playing with firearms; cleaning firearms, supervised activity at home) highlight the importance of a child’s developmental stage when considering firearm access, storage, and need for careful supervision. Unlike older adolescents and teenagers, children 10 years and younger have not yet acquired the developmental skills to evaluate and navigate dangerous situations^12,13 and/or appropriately enact safety measures that would limit the potential for injury.^14–16 Although educational interventions designed to teach children to stay away from, and not play with, firearms exist, when tested in behavioral simulations, findings suggest that these programs lack efficacy.^17–21 Thus, our results emphasize the need for enhancing firearm safety and reducing firearm access as a means to prevent these deaths.

Additionally, the high proportion of unsecure storage (ie, unlocked and loaded) in these clades of unintentional deaths underscores the need for a multipronged approach to increase locked and unloaded firearm storage in children’s homes. Data demonstrating that a significant proportion of these deaths occurred in the context of playing, cleaning, and/or accessing firearms within the household highlight the importance of locked storage to prevent access at this early developmental stage. This, combined with data demonstrating that children as young as 3 years have the manual dexterity and finger strength to depress a firearm trigger,²² raises concerns about unsupervised access to firearms within a child’s own household or in other environments where they spend time. Given that data shows that the risk for unintentional firearm fatalities is lower among households practicing secure storage,^23–25 the American Academy of Pediatrics recommends pediatricians provide locked firearm storage counseling to all parents as part of routine anticipatory guidance.²⁶ Further, consistent with recommendations from the ASK campaign – Asking Saves Kids – and findings that 15.9% of child deaths occurred in a friend or family home, such counseling should extend to recommending parents to inquire about firearm access in homes where their children spend time playing or visiting.²⁷ Policy-based approaches, such as child access prevention laws, also have demonstrated reductions in unintentional firearm fatalities among children when enacted with felony penalties.^28–31 Regardless, employing a universal approach, where household firearms are kept secure, remains the most effective intervention for reducing the potential for firearm fatalities among children.

Clusters of child firearm deaths occurring within the household, associated with interpersonal violence (eg, argument, jealousy) or murder-suicide, raise concerns regarding the role of IPV in precipitating firearm deaths among children. Nearly half of US women experience IPV during their lifetime, most often during the reproductive years.³² Firearms increase the lethality of IPV; partners with firearm access are 5 times more likely to kill their female intimate partner.³³ Although not the primary victim, children are frequently a collateral victim, with IPV accounting for up to a quarter of homicides occurring among children.^7,34–36 Additionally, most firearm-related murder-suicides involving children are precipitated by an IPV event.^37–41 Situations where CPS is actively involved, or there was previously identified IPV victimization, represent missed opportunities within the framework of our existing systems to prevent firearm injuries and deaths. CPS workers are a critical element in identifying IPV and/or providing education for families about benefits of locked firearm storage.

The United States Preventive Services Task Force recommends universal IPV screening in women of reproductive age,⁴² and others have expanded this universal screening to identify IPV in intimate relationships.^43,44 If a parent or caregiver discloses IPV to providers, the AAP suggests using healing-centered engagement by responding with validation and empathy while providing resources and referral to services, including legal options with firearm access prohibition (eg, domestic violence restraining orders [DVROs] and extreme risk protection orders).⁴⁵ Although no data exists on the impact of DVROs on child deaths, firearm restrictions in DVROs are associated with decreases in IPV homicides.^46–48 As a result, the effect of restricting firearm access in situations of IPV may also extend to decreasing pediatric firearm deaths occurring within IPV incidents.

Child firearm deaths occurring outside the child’s home, or the home of a friend or family, often occurred in urban settings within the context of community violence, where the child was the victim of community violence, including drive-by shootings. Examination of pediatric firearm injuries in urban trauma centers demonstrates that, in outdoor shootings, a third to half were drive-by shootings, often perpetrated by an unknown assailant.⁴⁹ Such firearm deaths disproportionately affect Black children in urban settings with high poverty rates and a legacy of redlining and economic disinvestment.^50–56 Such factors highlight the potential for prevention programs in urban settings to reduce firearm psychological trauma, injuries, and deaths on disadvantaged children. Individual-level hospital-based violence interventions (eg, SafERteens) employing behavioral therapy combined with referral to resources demonstrate efficacy in reducing both violent aggression and victimization.^57–60 Further, community-level greening and vacant lot remediation reduce firearm violence by creating busy streets where residents can safely interact and congregate. Such interventions, especially when combined with policy-based initiatives to address economic disinvestment,^61–63 enhance employment opportunities, and reduce illegal firearm access^64–67, may be an effective mechanism for reducing community violence and the potential for bystander child firearm deaths.

Limitations should be acknowledged.⁹ Although most states contribute data to the NFR-CRS, only 41 states permit access to their data for research, and states do not include every child death for review. Yet, the dataset captured 51.6% of the 3063 child firearm-related deaths in the United States from 2004 to 2020.² Second, several variables within the dataset have high rates of missing data, limiting their use and interpretability. Given the novelty of this contextual data to understanding these injuries, results remain an important contribution to advancing prevention efforts. Improvements to data collection during death investigations, especially regarding firearm storage, could bolster these missing data in the future. Third, the NFR-CRS data dictionary historically lacked standard definitions for several contextual variables, leading to the potential for CDR teams to inconsistently define various terms (eg, commission of a crime). Despite these limitations, the NFR-CRS provides national, broad contextual data on pediatric death, providing a unique opportunity to identify important contextual factors in child fatalities. Further, such analyses have also provided opportunities to highlight this variability and begin to define stricter criteria for future analyses.

Firearm deaths of children aged 0 to 10 have multiple causes that include identifiable opportunities for prevention. These causes include unlocked, accessible firearms that are unintentionally discharged, intimate partner violence-related deaths, and community violence, and at least half of the child deaths involved an unlocked firearm.

We thank the states that participate in the NFR-CRS and Carrie Musolf for her assistance in manuscript preparation.

CDR

child death review

CPS

Child Protective Services

IPV

intimate partner violence

NFR-CRS

National Fatality Review-Case Reporting System