Rivara FP, Koepsell TD, Grossman DC, Mock C. Effectiveness of Automatic Shoulder Belt Systems in Motor Vehicle Crashes. JAMA. 2000;283(21):2826-2828. doi:10.1001/jama.283.21.2826
Author Affiliations: Harborview Injury Prevention and Research Center (Drs Rivara, Koepsell, Grossman, and Mock), the Departments of Epidemiology (Drs Rivara, Koepsell, and Mock), Pediatrics (Drs Rivara and Grossman), Surgery (Dr Mock), and Health Services (Drs Koepsell and Grossman), University of Washington, and the Crash Injury Research and Engineering Network (Drs Rivara, Grossman, and Mock), Seattle.
Context Approximately 10 million cars with automatic shoulder belt systems are
currently in use in the United States. However, reports on the effectiveness
of such restraints have yielded conflicting results.
Objective To determine the effectiveness of automatic shoulder belt systems in
reducing the risk of injury and death among front-seat passenger vehicle occupants.
Design, Setting, and Subjects Analysis of data collected from the 1993-1996 National Highway Traffic
Safety Administration Crashworthiness Data System on front-seat occupants
involved in 25,811 tow-away crashes of passenger cars, light trucks, vans,
and sport utility vehicles.
Main Outcome Measures Death and serious injury to specific body areas by use of manual lap
and shoulder belts, automatic shoulder belts with manual lap belts, or automatic
shoulder belts without lap belts, compared with no restraint use.
Results Use of automatic shoulder belts without lap belts was associated with
a decrease in the risk of death vs no restraint use but was not statistically
significant for all crashes (odds ratio [OR], 0.66; 95% confidence interval
[CI], 0.42-1.06) or for frontal crashes (OR, 0.71; 95% CI, 0.38-1.35) after
adjustment for occupant age, sex, vehicle year, airbag deployment, estimated
change in vehicle speed during the crash, and principal direction of force.
This association was significantly weaker than the 86% lower risk observed
for use of automatic shoulder belts with lap belts (OR, 0.14; 95% CI, 0.07-0.26
vs no restraint; P<.05). Use of automatic shoulder
belts without lap belts was associated with an increased risk of serious chest
(OR, 2.66; 95% CI, 1.11-6.35) and abdominal (OR, 2.06; 95% CI, 1.004-4.22)
injuries for all crashes.
Conclusions These data indicate that improperly used automatic restraint systems
may be less effective than properly used systems and are associated with an
increased risk of serious chest and abdominal injuries. Given the continued
widespread use of these automatic systems, educational programs may be warranted.
Various strategies have been developed to improve motor vehicle occupant
protection, including an automatic shoulder belt, combined with a manual lap
belt or a knee bolster. Since 1987, more than 27 million cars have been manufactured
with automatic 2-point belt systems1 and more
than 10 million are still in use.2 Studies
have found that drivers do not use the manual lap belt 50% to 71% of the time.3- 5 Prior studies examining
the effectiveness of automatic systems have produced conflicting estimates.6,7 Numerous studies also report injuries
from use of the shoulder harness alone.8- 13
The purpose of this study was to examine the effectiveness of automatic
shoulder belts when used with and without a manual lap belt.
The 1993-1996 National Highway Traffic Safety Administration's Crashworthiness
Data System was the source of data, a national probability sample of light
passenger vehicles (passenger cars, light trucks, vans, and sport utility
vehicles) involved in police-reported tow-away crashes. About 5000 cases are
chosen annually in a 3-stage sampling system. It is the most comprehensive,
nationally representative system available and has among the most accurate
reporting of belt use, based on vehicle inspection, interviews, and police,
autopsy, and hospital records. Outboard front-seat occupants older than 15
years were included. Injury severity data were coded using the Abbreviated
The 4 categories of restraint use were unrestrained, manual lap and
shoulder belt, automatic 2-point shoulder belt with a manual lap belt or an
automatic 3-point lap-shoulder belt, and automatic 2-point shoulder belt used
without a lap belt. All other restraint systems (0.3% of known belt use) were
excluded. Belt-use information was missing on 22.5% of crashes. Compared with
occupants for whom this information was present, cases with missing information
were similar in age, estimated change in vehicle speed during the crash (delta-V),
principal direction of force, and case-fatality rate, but these cases were
slightly more likely to involve men who were in somewhat older cars. These
cases were excluded.
Information was missing on delta-V for 47.6% of the remaining cases
and on principal direction of force for 17.2%. Because these characteristics
appeared to be important potential confounders, multiple imputation15,16 was used by an adaptation of the
approximate Bayesian bootstrap method.15 Imputation
groups were formed by simultaneous stratification on survival, restraint use,
sampling weight, and other characteristics found to be strongly associated
with the variable to be imputed when its value was known. Ejection was an
additional variable used to impute principal direction of force. Injury severity
score and principle direction of force (after imputation) were additional
variables used to impute delta-V. Five complete data sets were generated by
this method; parallel analyses were then carried out on all 5 sets.
Outcome variables were survival to 30 days and presence of an Abbreviated
Injury Scale Score of 2 or more.14 Each outcome
was modeled as a function of restraint use and covariates using multiple logistic
regression, using survey data analysis commands in STATA statistical software
version 5.0 (STATA Corp, College Station, Tex). Results obtained from analysis
of each of the 5 complete data sets were combined using methods described
by Rubin and Schenker15 to obtain final point
estimates and confidence intervals (CIs) that account for both the multistage
sampling design and the imputation process.
Data were available for 25,811 crashes involving outboard front-seat
occupants, representing an estimated total of 11,963,796 such crashes nationally.
Among occupants, 21.9% were unrestrained, 60.3% were restrained with a manual
3-point system, 15.1% were restrained with an automatic shoulder belt and
a manual or automatic lap belt, and 2.7% were using an automatic shoulder
belt without the lap belt.
Occupant and crash characteristics of the 4 seat belt restraint groups
are displayed in Table 1. There
were 1226 fatal injuries in the sample, representing 71,387 fatalities in
the national set of crashes from which it was derived, for a rate of 6.0 deaths
per 1000 front seat occupants in crashes. After adjustment for important confounders,
automatic shoulder belts worn with a manual lap belt were associated with
at least as large a reduction in fatality risk as 3-point manual belts (Table 2). Automatic shoulder belts with
no lap belt appeared to be somewhat protective in all crashes (odds ratio
[OR], 0.66; 95% CI, 0.42-1.06) and frontal crashes (OR, 0.71; 95% CI, 0.38-1.35),
although the CI included 1. The adjusted OR for automatic shoulder belts without
use of a lap belt was significantly closer to 1 than that for automatic shoulder
belts with use of the lap belt (P<.05).
Users of automatic shoulder belts without lap belts had a risk of head
injury in all crashes and in frontal crashes, which was not significantly
different from that of unbelted occupants, and a 2.7-fold higher risk of chest
injury in all crashes and more than 7-fold higher risk of chest injury in
frontal crashes compared with no belt use (Table 3). Similar patterns were found in relation to risk of abdominal
injury, with the risk of injury 2-fold higher in all crashes and more than
3.7-fold higher in frontal crashes. There was no significant difference between
restraint systems in the risk of injuries to the spine or the extremities
(data not shown).
Using automatic belts without a manual lap belt, while probably better
than no restraint, appeared to convey less protection against death than did
using 3-point restraints. Use of an automatic shoulder belt by itself was
associated with a significant increase in the risk of serious thoracic and
Prior studies have focused on the impact of automatic restraints on
risk of fatal injury. Evans7 estimated that
improperly used systems decreased the risk of death by 29%, similar to our
observation of a 34% reduction for all crashes and 29% for frontal crashes.
This study has certain limitations. We examined only those automatic
systems using an automatic shoulder harness and a manual or automatic lap
belt. The CIs around many estimates were wide given the sample size available
in the Crashworthiness Data System data set. There is also the potential for
misclassification of seat belt use. The categorization of proper belt use
was made from a combination of interviews, police reports, medical records,
and field inspections. However, some misclassification may still have occurred.
Because data on delta-V and the principal direction of force were missing
for many crashes, we used multiple imputation to allow using these potential
confounders in multivariate analyses. Restricting analysis to cases without
missing data leads to valid inferences only if data are missing completely
at random.17 In contrast, multiple imputation
substitutes a weaker assumption that data are missing at random conditional
on the values of variables used to form the imputation groups. Nonetheless,
we cannot rule out residual bias due to associations between other unmeasured
aspects of crashes and missing data on delta-V or principal direction of force.
Finally, the number of crashes involving automatic systems was limited. Hence
for all crashes, the true effect of the restraint configuration on risk of
death could lie anywhere from a 58% reduction to a 6% increase and still be
reasonably compatible with the data.
Given the still widespread use of these automatic systems, educational
programs that encourage vehicle occupants to buckle up their lap belts may