Video Abstract

Video Abstract

Close modal
BACKGROUND AND OBJECTIVES:

Road traffic accidents are a leading cause of child deaths in the United States. Although this has been examined at the national and state levels, there is more value in acquiring information at the county level to guide local policies. We aimed to estimate county-specific child mortality from road traffic accidents in the United States.

METHODS:

We queried the Fatality Analysis Reporting System database, 2010–2017, for road traffic accidents that resulted in a death within 30 days of the auto crash. We included all children <15 years old who were fatally injured. We estimated county-specific age- and sex-standardized mortality. We evaluated the impact of the availability of trauma centers and urban-rural classification of counties on mortality.

RESULTS:

We included 9271 child deaths. Among those, 45% died at the scene. The median age was 7 years. The overall mortality was 1.87 deaths per 100 000 children. County-specific mortality ranged between 0.25 and 21.91 deaths per 100 000 children. The availability of a trauma center in a county was associated with decreased mortality (adult trauma center [odds ratio (OR): 0.59; 95% credibility interval (CI), 0.52–0.66]; pediatric trauma center [OR: 0.56; 95% CI, 0.46–0.67]). Less urbanized counties were associated with higher mortality, compared with large central metropolitan counties (noncore counties [OR: 2.33; 95% CI, 1.85–2.91]).

CONCLUSIONS:

There are marked differences in child mortality from road traffic accidents among US counties. Our findings can guide targeted public health interventions in high-risk counties with excessive child mortality and limited access to trauma care.

What’s Known on This Subject:

Road traffic accidents are a leading cause of child deaths in the United States. This has been examined at the national and state levels, yet there may be more value in acquiring county-level information, where health care programs are designed and implemented.

There are marked differences in child mortality from road traffic accidents among US counties. Rural counties have a disproportionately high mortality, and availability of a trauma center in the county is associated with lower mortality.

Road traffic accidents are a leading cause of child deaths in the United States, accounting for as much as 1 in every 5 child deaths.1  Examinations of this disconcerting death toll have been focused on the role of pediatric trauma centers and access to pediatric trauma care.

There is a growing consensus today that injured children cared for at pediatric trauma centers have better outcomes in comparison with children managed at adult trauma centers or nontrauma designated hospitals.24  These differences in outcomes are more pronounced among younger children and those who suffer severe injuries.3  Notably, only 11% of injured children are treated at pediatric trauma centers.5,6

Appropriate access to pediatric trauma care is suboptimal, particularly for children residing in rural areas in the United States. For those children, 77% lack access to a pediatric trauma center.5  Furthermore, access to trauma care among those children is hindered by delayed response times and lengthy transport times.7  In addition, children involved in motor vehicle crashes in rural areas tend to have more severe injuries.8  All of these factors contribute to the increased mortality in this population.

Child mortality from road traffic accidents has been examined previously at the national and state levels.911  There is more value, however, in acquiring information specific to counties where health care programs and policies are formulated and implemented. County-level data on child mortality from road injuries may be used to aid local policy makers and health care providers in identifying areas of excess mortality and allow dissemination of resources to those areas. In this study, we conducted a nationally comprehensive county-level examination of child mortality from road traffic accidents in the United States.

After obtaining approval from our institutional review board, we queried the Fatality Analysis Reporting System (FARS) for road traffic accidents between the years 2010 and 2017,12  In this study, road traffic accidents comprised motor vehicle crashes as well as auto crashes involving nonmotorists struck by a motor vehicle. FARS is a comprehensive nationwide census of fatal motor vehicle crashes, which is employed by the National Highway Traffic Safety Administration to report to Congress and the American public regarding fatalities from road traffic accidents. In FARS, qualifying auto crashes are defined as those that occur on a public road and result in a death within 30 days of the auto crash. Nonmotorists, such as pedestrians and cyclists, involved in a fatal motor vehicle accident are captured as well. In the database, the characteristics and environmental conditions at the time of the auto crash are detailed. Information about the location and time of the auto crash as well as notification and arrival times of emergency medical services (EMS) to both the scene and hospital are detailed. EMS response time (the time from notification of EMS to the time of their arrival to the scene of the crash) and transport time (the duration from the time of arrival of EMS to the scene to the time of their arrival to a hospital) were captured. The database also includes information about the characteristics of the involved vehicles and the drivers of these vehicles. Specifically, we abstracted data on age, sex, race, and ethnicity. We also collated accident-specific data to estimate county-level percentage of fatal auto crash related to intoxicated drivers, average speed limit at the site of fatal auto crashes, and restraint use. We included the records of fatalities among children <15 years of age to focus primarily on this age group and diminish the dilutional effects of adolescent deaths >15 years of age.

We identified trauma centers in the United States using state health department Web sites and data available on the American College of Surgeons Web site.13  We gathered information related to the trauma designation level and the zip code locations of these trauma centers. We included pediatric and adult level-1 and -2–designated trauma centers. Among pediatric trauma centers, we did not distinguish between freestanding and adult trauma centers with pediatric trauma designation.

We assigned urban-rural designations to counties using the 2013 National Center for Health Statistics’ Urban–Rural Classification Scheme for Counties.14  Specifically, we defined 6 levels of urbanization. From the most urban to the most rural, these levels are large central metropolitan, large fringe metropolitan, medium metropolitan, small metropolitan, micropolitan, and noncore. The former 4 are metropolitan counties, and the latter 2 are nonmetropolitan or rural counties. These levels are based largely on population size. A detailed definition of each level can be found in Supplemental Table 1.

We supplemented our database with data from the US Census Bureau.15  Specifically, we collected county-level estimates of population number by age groups, sex ratio, education level, median household income, and unemployment rates.

We used small-area estimation methods to estimate county-specific child mortality from road traffic accidents. These methods address the small number of deaths in each county and ameliorate extreme variations in mortality associated with less populated counties. Specifically, we employed a Bayesian mixed-effects Poisson regression model of the following form:

In this model, di and di are the number of deaths, the underlying mortality, and the population <15 years of age in county i, respectively, according to the 2010 census. β represents fixed county-specific covariate effects, and γ represents random county effects. Fixed covariates, X, included the percentage of boys in county i, the percentage of children <5 years of age in county i, 6 categories of urbanization, the availability of a pediatric or an adult trauma center in county i, and interaction effects between the urban-rural categories and availability of a trauma center. Other fixed-effects county-specific covariates that were evaluated in the model comprised county-level education attainment level, median household income, unemployment rate, percentages of fatal auto crash related to intoxicated drivers, speed limit at the site of the crash, and restraint use. We compared several model iterations, including a conditional autoregressive spatial model, using deviance information criterion, Gelman-Rubin convergence diagnostic, and autocorrelation plots. We report county-specific age- and sex-standardized mortality per 100 000 children. We present the median estimated mortality and 95% credibility intervals (CIs). The models were fit by using the rjags and coda packages of R (version 3.5.1) software.1618  We used the ggplot2 and usmap packages to create the county maps.19,20  We also used Stata, version 14, (Stata Corp, College Station, TX) for data set management.21

We enumerated 9271 road traffic child deaths over the 8 years studied. The median age was 7 years (interquartile range [IQR], 3–11). Fifty-six percent of the children were boys. Approximately one-half were white (49%), 23% were Black, 23% were Hispanic, and 5% were Asian American or Pacific Islander. The majority (73%) of the deaths were in motorists (passengers of a motor vehicle), and the rest were in nonmotorists who were fatally injured by a motor vehicle crash. Notably, 45% of all deaths occurred before arriving at a hospital. Those deaths overwhelmingly comprised children who died at the scene (within 30 minutes of the crash); prehospital deaths were overwhelmingly deaths at the scene (97%); 3% died during transport.

The overall estimated child mortality from road traffic accidents in the United States over the study period was 1.87 (95% CI, 1.51–2.32) deaths per 100 000 children. The county-specific age- and sex-standardized mortality ranged between 0.25 and 21.91 deaths per 100 000 children (IQR, 2.10–4.19 deaths per 100 000 children). County-specific death rates are depicted geographically in Fig 1 and detailed for each county in Supplemental Table 2.

FIGURE 1

A, County-specific age- and sex-standardized child mortality from road traffic accidents. B, The county-specific number of children <15 years of age.

FIGURE 1

A, County-specific age- and sex-standardized child mortality from road traffic accidents. B, The county-specific number of children <15 years of age.

Close modal

We identified 552 hospitals with a level-1 or -2 adult trauma designation and 143 hospitals with a level-1 or -2 pediatric trauma designation. Among all 3142 counties in the United States, 89% did not have an adult or a pediatric trauma center located in the county. Adult and pediatric trauma centers were available in 333 (11%) and 109 (3%) counties, respectively. The availability of a trauma center was associated with lower child mortality. As depicted in Fig 2, compared with having no trauma center in a county, the availability of an adult trauma center was associated with a 41% lower mortality (odds ratio [OR]: 0.59; 95% CI, 0.52–0.66). Having a pediatric trauma center was associated with a 44% lower mortality (OR: 0.56; 95% CI, 0.46–0.67). In Fig 3, we show pediatric trauma centers located primarily in counties within the lowest decile of child mortality. Specifically, 20% of counties within the lowest decile of mortality have a pediatric trauma center, whereas only 1% of counties within the highest decile include a designated pediatric trauma center.

FIGURE 2

Association of the availability of a trauma center in a county and the urban-rural classification of a county with mortality.

FIGURE 2

Association of the availability of a trauma center in a county and the urban-rural classification of a county with mortality.

Close modal
FIGURE 3

Geographic distribution of pediatric trauma centers overlaying counties with the highest and lowest deciles of child mortality.

FIGURE 3

Geographic distribution of pediatric trauma centers overlaying counties with the highest and lowest deciles of child mortality.

Close modal

The majority of counties were designated nonmetropolitan; 641 (20%) were micropolitan, and 1335 (42%) were noncore. The metropolitan counties comprised 68 (2%) large central metropolitan counties, 368 (12%) large fringe metropolitan counties, 372 (12%) medium metropolitan counties, and 358 (11%) small metropolitan counties. Large metropolitan counties had significantly higher education attainment (attainment of bachelor’s degree or higher: 37%), median household income ($63 999), and restraint use (77%) compared to less urbanized counties (noncore counties: 18%,$45 570, 54%, respectively). With decreasing county urbanization category, average speed limits at the site of fatal crashes were incrementally increasing (large central counties: 41 mph; noncore counties: 55 mph), as was the percentage of drunk drivers (large central counties: 27%; noncore counties: 32%) involved in fatal crashes. County-level unemployment rate was not significantly associated with urban-rural category (large central counties: 7%; noncore counties: 7%). Less urbanized counties were associated with higher child mortality compared to large central metropolitan counties (micropolitan counties [OR: 1.60; 95% CI, 1.27–2.01]; noncore counties [OR: 2.33; 95% CI, 1.85–2.91]; Fig 2). Notably, the association of trauma center with mortality was not different within different urban-rural categories. In Fig 4, we depict the urban-rural class distribution of counties with the highest decile of mortality. Among those counties, 81% are noncore, the most rural category; none are large central metropolitan counties. A complete list of counties with the highest decile of child mortality and their urban-rural category can be found in Supplemental Table 3.

FIGURE 4

Urban-rural class distribution of counties in the highest decile of child mortality from road traffic accidents. Counties in the highest decile are depicted in red. The gray shading reflects the counties within the titled urban-rural category. A complete list of counties with the highest decile of child mortality and their urban-rural categories can be found in Supplemental Table 3. A, Large central metropolitan. B, Large fringe metropolitan. C, Medium metropolitan. D, Small metropolitan. E, Micropolitan. F, Noncore.

FIGURE 4

Urban-rural class distribution of counties in the highest decile of child mortality from road traffic accidents. Counties in the highest decile are depicted in red. The gray shading reflects the counties within the titled urban-rural category. A complete list of counties with the highest decile of child mortality and their urban-rural categories can be found in Supplemental Table 3. A, Large central metropolitan. B, Large fringe metropolitan. C, Medium metropolitan. D, Small metropolitan. E, Micropolitan. F, Noncore.

Close modal

The overall county-level median response time was 10 minutes (IQR, 7–13). The median response time was twofold higher in noncore counties compared with central metropolitan counties: 12 minutes (IQR, 9–15) vs 6 minutes (IQR, 5–7), respectively, P < .01. A notable trend of increasing response time with incrementally less urbanized counties was evident. A total of 6% of counties lacked response time data. Among those transferred to a hospital by EMS, the county-level median transfer time was 34 minutes (IQR, 27–44). The median transport time ranged from 24 minutes (IQR, 21–27) in large central metropolitan counties to 37 minutes (IQR, 27–48) in noncore counties, P < .01. A total of 17% of counties did not have transfer times available.

In this study, we conducted a nationally comprehensive examination of child mortality from road traffic accidents at the county level in the United States. We demonstrated marked variability in child mortality across counties. Rural counties had disproportionately high mortality rates, whereas counties with a trauma center were more likely to have low mortality rates.

Child mortality from road traffic accidents is the leading cause of child death in the United States.1  This mortality, however, is not distributed homogeneously across the nation. As our study revealed, child mortality varied dramatically among counties, even among neighboring counties. Counties in the Northeast, the West Coast, and around the Great Lakes had low mortalities. High mortality rates were notably more frequent among counties in the Midwest and South regions of the county. Counties with low mortality tended to be densely populated counties. Nonetheless, even among densely populated counties, mortality varied. For example, among counties with >100 000 children, there was a 13-fold increase in mortality between the county with the highest mortality (Polk County, FL, at 3.27 per 100 000 children) and lowest mortality (Fairfax County, VA, at 0.25 per 100 000 children).

The availability of either a pediatric or adult trauma center was associated with lower county-level mortality. Our finding augments a growing body of literature that suggests improved morbidity and mortality among injured children cared for at trauma centers compared with those managed at nontrauma hospitals.2,3  Among those studies, however, outcomes were improved in pediatric trauma centers compared with adult trauma centers especially among younger children and those with more severe injuries.3  In our analysis, there was a tendency for lower mortality among counties with a pediatric trauma center compared with counties with adult centers (and no pediatric centers), albeit the CIs largely overlapped.

Other factors have been implicated in high mortality rates among injured children in rural counties. Long transport times from rural areas, for example, have been demonstrated in previous studies.7  In our cohort, we noted a 6 minute delay in the median response time and another 13 minute delay in the median transport time between large central metropolitan counties and noncore counties. McCowan et al7  showed that helicopter EMS spent longer times at the scene in rural areas. Despite longer scene times, the authors noted that the overall transport times were equivalent between rural and urban areas. Notably, however, in their study, deaths that occurred before the arrival of helicopter EMS were not assessed. Other attributes of rural areas that have been associated with high mortality from road traffic accidents include the overall higher severity of injuries in rural areas, faster driving speeds, higher alcohol use, and lower restraint use.8,2729  These attributions to rural areas coincided with our findings of lower education, lower median household income, lower restraint use, higher speed limits, and higher rate of intoxicated drivers in rural counties.

Strikingly, almost one-half of all children in our cohort died at the scene. Prehospital deaths among road traffic accidents are scarcely reported in the literature. This is primarily because of the fact that the overwhelming majority of studies on this subject are hospital-based; hence, the study population is limited to individuals that make it to the hospital. Cooper et al30  and Vane et al31  highlighted the underreporting of prehospital deaths in hospital-based studies. In our population-based study, we overcome this limitation by examining the entire child population that is fatally impacted by road traffic accidents. In our cohort, 58% of children injured at noncore counties died at the scene; in contrast, 31% of child deaths occurred at the scene in large central metropolitan counties.

The findings of our study need to be interpreted in the light of the following limitations. First, because the overall mortality rate was low and the number of counties was large, mortality estimates were prone to extreme values, particularly in counties with a small number of children. We used small-area estimation methods to mitigate these effects and temper extreme variations in our mortality estimates. Second, although we adjusted for age and sex as well as the urban-rural classification of counties and availability of trauma centers, other predictors that may have a significant impact on mortality, particularly intangible ones such as the readiness of trauma systems or efficacy of prehospital triage protocols, were not addressed. Third, we estimated child mortality over an 8-year period. We did not examine temporal trends over this period. Such an analysis would have been untenable given the small number of fatalities in the majority of counties. Fourth, all counties with a pediatric trauma center also had an adult trauma center. Hence, our evaluation of the impact of the availability of a pediatric trauma center on county-level child mortality implicitly encompasses the role of adult trauma centers on mortality in counties with pediatric trauma centers. Fifth, a large proportion of records had missing response and transport times, necessitating much caution in interpreting these estimates. A fraction of the missing data pertains to auto crashes in which EMS was not notified or non-EMS personnel transported the child. Last, the FARS database captures deaths within 30 days of a crash. Children who died of their injures after this period were not included in the database.

Child mortality from road traffic accidents is a leading yet preventable cause of child deaths. In this population-based and nationally comprehensive county-level study, we identified high-risk counties with excessive child mortality and limited access to trauma care. Our results can help stakeholders and policymakers investigate high-risk clusters, design and implement policies at the local and national levels, and guide resources appropriately to areas in need.

Drs Mokdad and Qureshi conceptualized and designed the study, collected the data, conducted the initial analyses, drafted the initial manuscript, and reviewed and revised the manuscript; Drs Wolf, Pandya, and Ryan contributed substantially to the conception and design, collected data, and critically reviewed and critically revised the manuscript for intellectual content; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

FUNDING: No external funding.

CI

credibility interval

EMS

emergency medical services

FARS

Fatality Analysis Reporting System

IQR

interquartile range

OR

odds ratio

1
Institute for Health Metrics and Evaluation
.
GBD compare | Viz Hub
. Available at: http://viz.healthmetricsandevaluation.org/gbd-compare/. Accessed July 10, 2015
2
Dreyfus
J
,
Flood
A
,
Cutler
G
,
Ortega
H
,
Kreykes
N
,
Kharbanda
A
.
Comparison of pediatric motor vehicle collision injury outcomes at Level I trauma centers
.
J Pediatr Surg
.
2016
;
51
(
10
):
1693
1699
3
Sathya
C
,
Alali
AS
,
Wales
PW
, et al
.
Mortality among injured children treated at different trauma center types
.
JAMA Surg
.
2015
;
150
(
9
):
874
881
4
Webman
RB
,
Carter
EA
,
Mittal
S
, et al
.
Association between trauma center type and mortality among injured adolescent patients
.
JAMA Pediatr
.
2016
;
170
(
8
):
780
786
5
Carr
BG
,
Nance
ML
.
.
Curr Opin Pediatr
.
2010
;
22
(
3
):
326
331
6
Myers
SR
,
Branas
CC
,
French
B
,
Nance
ML
,
Carr
BG
.
A national analysis of pediatric trauma care utilization and outcomes in the United States
.
Pediatr Emerg Care
.
2019
;
35
(
1
):
1
7
7
McCowan
CL
,
Swanson
ER
,
Thomas
F
,
Handrahan
DL
.
Outcomes of pediatric trauma patients transported from rural and urban scenes
.
Air Med J
.
2008
;
27
(
2
):
78
83
8
Kim
K
,
Ozegovic
D
,
Voaklander
DC
.
Differences in incidence of injury between rural and urban children in Canada and the USA: a systematic review
.
Inj Prev
.
2012
;
18
(
4
):
264
271
9
National Trauma Data Bank
.
National Trauma Data Bank Pediatric Annual Report
.
Chicago, IL
:
American College of Surgeons
;
2016
10
Wolf
LL
,
Chowdhury
R
,
Tweed
J
, et al
.
Factors associated with pediatric mortality from motor vehicle crashes in the United States: a state-based analysis
.
J Pediatr
.
2017
;
187
:
295.e3
-
302.e3
11
Wolf
LL
,
Chowdhury
R
,
Tweed
J
, et al
.
State-level geographic variation in prompt access to care for children after motor vehicle crashes
.
J Surg Res
.
2017
;
217
:
75
83.e1
12
.
Fatality Analysis Reporting System (FARS)
. Available at: https://www.nhtsa.gov/research-data/fatality-analysis-reporting-system-fars. Accessed May 22, 2019
13
American College of Surgeons
.
Searching for verified trauma centers
. Available at: https://www.facs.org/search/trauma-centers. Accessed May 22, 2019
14
Ingram
DD
,
Franco
SJ
.
2013 NCHS Urban–Rural Classification Scheme for Counties
.
Washington, DC
:
National Center for Health Statistics
;
2014
15
US Census Bureau
. Annual county resident population estimates by age, sex, race, and hispanic origin: April 1, 2010 to July 1,
2017
. Available at: https://www.census.gov/newsroom/press-kits/2018/estimates-characteristics.html. Accessed January 2, 2020
16
R Core Team
.
R: A Language and Environment for Statistical Computing
.
Vienna, Austria
:
R Foundation for Statistical Computing
;
2015
17
Plummer
M
,
Best
N
,
Cowles
K
,
Vines
K
.
CODA: convergence diagnosis and output analysis for MCMC
.
R News
.
2006
;
6
(
1
):
7
11
18.
Plummer
M.
rjags: Bayesian graphical models using MCMC. Available at: https://CRAN.R-project.org/package=rjags. Accessed February 12, 2018
19
Wickham
H.
Ggplot2: Elegant Graphics for Data Analysis
, 2nd ed.
New York, NY
:
Springer-Verlag
;
2016
. Available at: https://ggplot2.tidyverse.org/. Accessed February 12, 2018
20.
Di Lorenzo
P.
usmap: US maps including Alaska and Hawaii. Available at: https://CRAN.R-project.org/package=usmap. Accessed February 12, 2018
21
Stata Statistical Software
.
Release 14 [Computer Program]
.
College Station, TX
:
StataCorp LP
;
2015
22
Svenson
JE
,
Spurlock
C
,
Nypaver
M
.
Factors associated with the higher traumatic death rate among rural children
.
Ann Emerg Med
.
1996
;
27
(
5
):
625
632
23
Ochoa
C
,
Chokshi
N
,
Upperman
JS
,
Jurkovich
GJ
,
Ford
HR
.
Prior studies comparing outcomes from trauma care at children’s hospitals versus adult hospitals
.
J Trauma
.
2007
;
63
(
6
,
suppl
):
S87
S91
;
discussion S92–S95
24
Densmore
JC
,
Lim
HJ
,
Oldham
KT
,
Guice
KS
.
Outcomes and delivery of care in pediatric injury
.
J Pediatr Surg
.
2006
;
41
(
1
):
92
98
;
discussion 92–98
25
Segui-Gomez
M
,
Chang
DC
,
Paidas
CN
,
Jurkovich
GJ
,
Mackenzie
EJ
,
Rivara
FP
.
Pediatric trauma care: an overview of pediatric trauma systems and their practices in 18 US states
.
J Pediatr Surg
.
2003
;
38
(
8
):
1162
1169
26
Vogel
LD
,
Vongsachang
H
,
Pirrotta
E
, et al
.
Variations in pediatric trauma transfer patterns in Northern California pediatric trauma centers (2001-2009)
.
.
2014
;
21
(
9
):
1023
1030
27
Muelleman
RL
,
Mueller
K
.
Fatal motor vehicle crashes: variations of crash characteristics within rural regions of different population densities
.
J Trauma
.
1996
;
41
(
2
):
315
320
28.
Australian Government Department of Infrastructure
, Transport, Regional Development and Communications. Roads. Available at: https://www.infrastructure.gov.au/roads/. Accessed January 3, 2019.
29
Agran
PF
,
Anderson
CL
,
Winn
DG
.
Factors associated with restraint use of children in fatal crashes
.
Pediatrics
.
1998
;
102
(
3
). Available at: www.pediatrics.org/cgi/content/full/102/3/e39
30
Cooper
A
,
Barlow
B
,
Davidson
L
,
Relethford
J
,
O’Meara
J
,
Mottley
L
.
Epidemiology of pediatric trauma: importance of population-based statistics
.
J Pediatr Surg
.
1992
;
27
(
2
):
149
153–154
31
Vane
DW
,
Shackford
SR
.
Epidemiology of rural traumatic death in children: a population-based study
.
J Trauma
.
1995
;
38
(
6
):
867
870

## Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.