Nathens AB, Jurkovich GJ, Cummings P, Rivara FP, Maier RV. The Effect of Organized Systems of Trauma Care on Motor Vehicle Crash Mortality. JAMA. 2000;283(15):1990-1994. doi:10.1001/jama.283.15.1990
Author Affiliations: Department of Surgery, Harborview Medical Center and the Harborview Injury Prevention and Research Center, University of Washington, Seattle.
Context Despite calls for wider national implementation of an integrated approach
to trauma care, the effectiveness of this approach at a regional or state
level remains unproven.
Objective To determine whether implementation of an organized system of trauma
care reduces mortality due to motor vehicle crashes.
Design Cross-sectional time-series analysis of crash mortality data collected
for 1979 through 1995 from the Fatality Analysis Reporting System.
Setting All 50 US states and the District of Columbia.
Subjects All front-seat passenger vehicle occupants aged 15 to 74 years.
Main Outcome Measures Rates of death due to motor vehicle crashes compared before and after
implementation of an organized trauma care system. Estimates are based on
within-state comparisons adjusted for national trends in crash mortality.
Results Ten years following initial trauma system implementation, mortality
due to traffic crashes began to decline; about 15 years following trauma system
implementation, mortality was reduced by 8% (95% confidence interval [CI],
3%-12%) after adjusting for secular trends in crash mortality, age, and the
introduction of traffic safety laws. Implementation of primary enforcement
of restraint laws and laws deterring drunk driving resulted in reductions
in crash mortality of 13% (95% CI, 11%-16%) and 5% (95% CI, 3%-7%), respectively,
while relaxation of state speed limits increased mortality by 7% (95% CI,
Conclusions Our data indicate that implementation of an organized system of trauma
care reduces crash mortality. The effect does not appear for 10 years, a finding
consistent with the maturation and development of trauma triage protocols,
interhospital transfer agreements, organization of trauma centers, and ongoing
In 1998, 41,480 persons died in motor vehicle crashes (MVCs) in the
United States.1 While traffic crashes continue
to be the single most important mechanism of injury death, mortality due to
traffic crashes declined from 22.7 per 100,000 person-years in 1979 to 15.9
per 100,000 person-years in 1995.2 Many factors
contributed to this decrease in mortality, including improvement in road and
automotive design and the legislation of restraint laws and laws restricting
drinking and driving.3 The possible contribution
of organized systems of trauma care to the decrease in mortality due to crashes
has not been evaluated.
Regional or state systems of trauma care were developed to ensure that
patients with injuries severe enough to require multispecialty care were identified
and transported to centers appropriate for the level of care required. As
early as 1976, the American College of Surgeons developed criteria for the
designation of trauma centers and the establishment of regional trauma systems
in an attempt to improve trauma care delivery.4
By 1992, only 5 states had fulfilled all the necessary criteria for a state
trauma system, while most had no organized system of trauma care.5 The limited implementation of comprehensive trauma
systems may be due to insufficient evidence of their effectiveness.
Deaths due to MVCs are a useful measure of trauma system effectiveness.
For example, injuries occurring as a result of crashes tend to challenge the
resources and medical personnel of the institutions caring for such patients
due to their complexity and multisystem involvement. Additionally, MVCs frequently
occur at sites remote from definitive medical intervention and thus challenge
prehospital transportation, triage, and interhospital transfer mechanisms.
In this context, we set out to evaluate the effectiveness of organized systems
of trauma care in reducing crash mortality by performing a national study
of crash deaths, comparing state crash mortality rates before and after implementation
of an organized system of trauma care while controlling for secular trends
in crash mortality.
The primary data source for information regarding regional or state
trauma systems was the 1993 Inventory of Trauma Systems,6
which provides data on the structural characteristics of trauma systems, both
statewide and regionally based, that were operational in 1993. All trauma
systems in the inventory have an enabling state statute, regulation, or executive
order granting administering organizations the legal authority to develop
and enforce trauma system policy. At a minimum, this legal authority involves
both the authority to designate trauma centers and the authority to establish
triage procedures that allow prehospital personnel to bypass nearer facilities.
To update these data, we designed a survey that we sent to all state emergency
medical service directors to determine the presence or absence of a trauma
system in their state, the nature of the supporting legislation, and the presence
and timing of specific components of the trauma system. Additionally, state
statutes were reviewed to assess whether there existed any trauma system legislation
at the state level that was not apparent either in the Inventory of Trauma
Systems or in the results of our survey. We evaluated systems at the state,
rather than the regional level, as in most cases, authority for the system
rested with state-level organizations and in those states with regionally
based systems, the majority of the state population resided within these regions.
We considered the year of first trauma center designation as the start
of trauma system implementation within each state. Full implementation of
an organized system of trauma care may take several years to develop.6 Further, the system's effect may not be evident immediately,
given that referral practices and triage policies, even if supported by legislation,
take time to change.7 We therefore assessed
how crash-related mortality within each state changed over time after the
start of trauma system implementation.
Data regarding deaths due to MVCs were obtained from the Fatality Analysis
Reporting System, which collects data on all fatal MVCs occurring on public
roads in the United States in which the death occurred within 30 days of the
crash.8 The analysis was limited to deaths
of all front-seat occupants of passenger vehicles aged 15 to 74 years occurring
between January 1, 1979, and December 31, 1995. Children were excluded because
regionalization of pediatric and adult trauma care may not necessarily develop
in parallel, and data on regionalization of pediatric trauma care are not
readily available. Fatalities were abstracted by year, age, and state of death.
The US Census Bureau provided data on intercensal population estimates by
year, state, and age.9 Our analyses were based
on crash mortality rates in the form of deaths per 100,000 person-years, although
an analysis based on deaths per vehicle miles driven yielded similar estimates
of the effect of trauma systems on crash mortality.
Our estimates of the effect of trauma system implementation on crash
deaths might be in error if we failed to account for other policies that reduce
crash mortality. To assess the possible confounding effects of other laws,
we selected 3 types of statutes that, according to previous studies, affect
deaths due to traffic crashes: (1) primary restraint laws that allow officers
to stop vehicles and issue citations if the occupants are not wearing seat
belts10; (2) secondary restraint laws that
allow officers to issue citations for failure to use restraints if the vehicle
was stopped for some other reason10; (3) administrative
license revocation laws, which allow for prompt license suspension if a driver
is cited for drunken driving or refuses an alcohol test11;
and (4) laws that allowed for highway speed limits greater than 55 mph.12 Data regarding these laws were obtained from the
National Safety Council13 and the National
Highway Traffic Safety Administration.14
We first carried out a stratified analysis comparing crash mortality
rates before and after system implementation in states that had implemented
a system during the study period. These state-specific mortality rate ratios
(MRRs) were combined using Mantel-Haenszel methods.15
We then used a regression approach to compare, within each state, crash mortality
rates in time periods with an organized system of trauma care to mortality
in time periods without such a system, while adjusting for secular trends
in crash mortality in all other states and the potential confounding variables
described above. The resulting MRR is a pooled estimate of the within-state
ratio of mortality rates at any given point in time after system implementation
to the mortality rate that would have been expected in those same states had
a trauma system not been implemented.
We first used a Poisson regression model to compare the mortality rate
in time periods that had an organized system of trauma care to time periods
without a system within the same state.16,17
To account for trends in crash mortality rates, we included all states in
the analysis and modeled time in years using the fractional polynomial method
described by Royston and Altman.18 Briefly,
combinations of polynomial terms in the set (−2, −1, −0.5,
0, 0.5, 1, 2, 3) were considered, first individually and then in increasingly
complex combinations. We used the likelihood ratio test to compare models
and systematically selected the best fitting model with the smallest number
of terms. Ultimately 6 terms were used to model trends in mortality and 1
term was used to model time since system implementation. Results were expressed
as the MRR for each year after system implementation, compared with expected
mortality if there were no system. Pointwise 95% confidence intervals (CIs)
To allow for possible overdispersion that might violate the restrictive
assumptions of the Poisson distribution, final results were estimated using
negative binomial regression.16 We used variance
estimators that allowed for the possibility that observations within each
state might be correlated.19 Analyses were
conducted using Stata software (Stata Statistical Software: Release 6.0; College
Station, Tex; Stata Corporation, 1999).
Over the 17 years of this study, there were 439,195 deaths due to MVCs.
The crash mortality rate for front-seat occupants aged 15 to 74 years was
17.8 per 100,000 person-years in 1979; it decreased to 13.1 per 100,000 person-years
in 1995. Adjusting for changes in the age distribution of the population,
crash mortality decreased by 22% (95% CI, 21%-23%) over this time interval.
From 1979 through 1995, 22 states were identified as having both designated
trauma centers and triage policies enforced by legislation (Table 1). In 10 states, the principal organizing and regulating
authority was at the state level. Two states had regional systems, wherein
a county or series of adjacent counties developed and enforced trauma system
policy, while 7 states had regional system development with state enforcement
of regulations. In 3 states, the organizational structure was delegated to
a private or quasi-private organization given authority by the state. Except
for the state of Maryland in which trauma center designation began in 1976,
all states began designation of trauma centers from 1979 through 1995.
States implementing a trauma system from 1979 through 1995 experienced
a drop in crash mortality from 16.2 to 11.6 deaths per 100,000 person-years.
In contrast, states without an organized system of trauma care experienced
a decrease in crash mortality from 20.4 per 100,000 person-years in 1979 to
15.4 per 100,000 person-years in 1995. The age-adjusted reduction in crash
mortality was 8% (95% CI, 5%-11%) greater over this time interval in states
with trauma systems compared to those without. The relative effectiveness
of traffic safety laws and trauma systems in contributing to this decline
are examined below under "Regression Analysis." When time periods with a trauma
system were compared with time periods without a system within the same state,
18 of 22 states had a statistically significant reduction in crash mortality.
Using stratified methods, the pooled MRR was 0.87 (95% CI, 0.86-0.88) in periods
with a trauma system compared with earlier periods (Table 1).
Initiation of an organized system of trauma care appeared to have no
beneficial effect on crash mortality with the first 10 years of implementation
(Figure 1). After 10 years, crash
mortality decreased compared with what would have been expected within the
same states had there been no established system of trauma care. After approximately
13 years, the decrease in mortality was statistically significant. By the
15th year after system implementation, we estimated that the MRR was 0.92
(95% CI, 0.88-0.97) after adjusting for the overall trend in mortality, changes
in the age distribution of the population, the introduction of restraint and
administrative license revocation laws, and changes in highway speed limits.
When primary or secondary restraint laws were in effect, mortality was
reduced compared with time periods without these laws; the reduction appeared
to be greater for primary restraint laws (Table 2). Administrative license revocation laws were also associated
with reduced mortality. Time periods with a highway speed limit greater than
55 mph had greater crash mortality compared with other time periods. Although
all of these laws were associated with mortality, they had little confounding
effect on our estimates of the effects of trauma systems on crash mortality.
When any or all of these statues were removed from the regression model, the
MRR estimates related to trauma systems were essentially unchanged.
Within a state, crash mortality rates vary from year to year. If a state
were to have a period of unusually high rates as part of this expected variation,
the state might be stimulated to propose and/or implement an organized system
of trauma care. Any subsequent mortality decline could be due partly to the
tendency of rates to regress to their mean value. To examine this possibility,
we calculated the crash mortality rate in the 3-year period before system
implementation and compared these rates with previous years in the same state
and adjusted for national mortality trends. Mortality incidence in the 3 years
prior to system implementation was little different from that in earlier years:
MRR, 1.02 (95% CI, 0.99-1.05).
From 1979 through 1995, 22 states either implemented or already had
organized systems of trauma care. About 10 years after system implementation,
it appears that mortality due to traffic crashes begins to decline; about
15 years following trauma system implementation, mortality was reduced by
8% (95% CI, 3%-12%). The finding that the effect of trauma systems on mortality
may evolve over time is plausible. Designation of the first trauma center,
which we chose as the starting date for each system, may have little immediate
impact on what actually happens to trauma patients. We believe that over time,
trauma center protocols mature, triage policies are implemented, referral
patterns and transfer policies change, and ultimately patient outcome improves.
Our estimates are based on the assumption that the time of first trauma
center designation represents the beginning of an organized system of care.
Although several alternative timepoints could have been chosen for the start
of trauma system implementation, the year of first trauma center designation
was selected for several reasons. First, it was a readily available and discrete
point in time that signified a change in the way trauma patients would be
treated in the region of interest. It was also the earliest identifiable such
point in most states and thus represented the initiation of some form of organized
trauma care. We acknowledge that several other components are critical to
the success of a trauma system, including the development of prehospital triage
criteria, interfacility transfer protocols, and quality assurance. Most of
the other trauma system components require the presence of a designated trauma
center, emphasizing the importance of this particular component to the overall
One limitation of our analysis is that we adjusted our estimates for
some laws related to traffic safety, but not all possibly relevant traffic
laws. We found, however, that adjustment for restraint laws, administrative
license revocation, and speed limit laws resulted in essentially no change
in our estimates. While the laws we examined were all independently associated
with traffic mortality, they had little association with the implementation
of trauma systems and were not important confounders of our estimates. We
suspect that most traffic safety laws are instituted with little regard to
the establishment of trauma systems and therefore are not likely to be important
sources of residual confounding of our estimates. Also, adjusting for secular
trends ensures that many other factors not directly measurable (eg, changes
in vehicle design, road design, and improvements in medical care) are incorporated
into our final estimates.
Our estimate of the reduction in mortality attributable to trauma system
implementation is likely to be conservative for several reasons. First, there
may have been misclassification of states with and without organized systems
of care. For example, our analysis was limited to trauma systems that were
backed by legislation in the form of a state statute, regulation, or executive
order. However, some regions or states that have relatively well-organized
systems without supporting legislation were not considered systems for the
purpose of this analysis. Additionally, we considered both regional and state-wide
systems of trauma care. In states with regional systems, it is unlikely that
the entire state population has access to an organized system of trauma care.
Finally, approximately half of all crash deaths occur at the scene and thus
no trauma system, irrespective of its breadth or quality, can affect the outcome
of many of these victims.20
The extent to which trauma system policy is actually enforced in states
with trauma systems is unclear. There are data suggesting that even in states
with mature trauma systems, compliance with triage criteria for major trauma
patients is, at best, approximately 85%.21
Data on compliance with triage criteria by state would provide information
on whether the process of care of these trauma patients differs following
system implementation and would strengthen the causal relationship between
system implementation and lower crash mortality. Unfortunately, these data
are not available in most states and are even less likely to be available
in states prior to system implementation.
Other assessments of the effectiveness of trauma systems have demonstrated
variable efficacy depending on the study design. For example, one study reported
a greater than 50% reduction in the proportion of deaths deemed potentially
preventable prior to and following regional system implementation.22 Although these data suggest a marked and rapid improvement
in outcome associated with the development of a regional trauma system, the
assessment of preventability requires judgments by a panel regarding the appropriateness
of care and salvageability. In the study mentioned above and other similar
studies, the preventable death fraction is calculated by excluding prehospital
deaths and patients dead on arrival, precluding any assessment of prehospital
care and/or resuscitation.
More recently, Mullins et al23 evaluated
the outcome of hospitalized patients after organization of trauma care in
Oregon over a 6-year period wherein mortality was compared in the preimplementation,
implementation, and postimplementation phases. Although there was no mortality
benefit in the implementation phase, a 35% reduction in mortality was reported
within 2 years after implementation, a finding confirming that the effects
of system implementation are not instantaneous.
Our study offers several strengths. By analyzing the nation as a whole,
we provide a reasonable estimate of the national impact of trauma system implementation.
Further, the wide geographic distribution of the trauma system states precludes
there being any specific state or regional distinctions that may affect the
results. Finally, we did not exclude prehospital deaths, thus providing a
more realistic estimate of the effect of organized trauma care on overall
Our finding of a noteworthy effect of primary seat belt laws on crash
mortality and little effect of secondary restraint laws is consistent with
the findings of a recent review of other studies of these laws.10
Our finding of a modest increase in crash mortality associated with raising
the 55 mph highway speed limit is consistent with previous studies of speed
limits and crash deaths in the United States.12,24
We included deaths on all roads, most of which are not affected by the speed
limit change, so it is not surprising that our estimate of the change in mortality
was small. The estimate of the effectiveness of administrative license revocation
laws on crash mortality is similar to that reported in a previous study of
legislation directed toward reducing the rates of alcohol-related crashes.11 These data support the relative effectiveness of
legislative initiatives focused on primary or secondary prevention strategies
to reduce crash mortality.
We conclude that implementation of an organized system of trauma care
results in a measurable decrease in crash mortality. The effect does not appear
for 10 years, a finding consistent with the progressive implementation of
organized systems of trauma care over time.