Air Bag and Seat Belt Efficacy Crandall et al. Mortality Reduction with Air Bag and Seat Belt Use in Head-on Passenger Car Collisions

Transcription

1 American Journal of Epidemiology Copyright 2001 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved Vol. 153, No. 3 Printed in U.S.A. Air Bag and Seat Belt Efficacy Crandall et al. Mortality Reduction with Air Bag and Seat Belt Use in Head-on Passenger Car Collisions Cameron S. Crandall, 1 Lenora M. Olson, 2 and David P. Sklar 1 To assess the efficacy of occupant protection systems, the authors measured the mortality reduction associated with air bag deployment and seat belt use for drivers involved in head-on passenger car collisions in the United States. They used a matched case-control design of all head-on collisions involving two passenger cars reported to the Fatality Analysis Reporting System in , and driver mortality differences between the paired crash vehicles for air bag deployment and seat belt use were measured with matched-pair odds ratios. Conditional logistic regression was used to adjust for multiple effects. There were 9,859 head-on collisions involving 19,718 passenger cars and drivers. Air bag deployment reduced mortality 63% (crude odds ratio (OR) = 0.37, 95% confidence interval (CI): 0.32, 0.42), while lap-shoulder belt use reduced mortality 72% (OR = 0.28, 95% CI: 0.25, 0.31). In a conditional logistic model that adjusted for vehicle (rollover, weight, age) and driver (age, sex) factors, air bags (OR = 0.71, 95% CI: 0.58, 0.87) and any combination of seat belts (OR = 0.25, 95% CI: 0.22, 0.29) were both associated with reduced mortality. Combined air bag and seat belt use reduced mortality by more than 80% (OR = 0.18, 95% CI: 0.13, 0.25). Thus, this study confirms the independent effect of air bags and seat belts in reducing mortality. Am J Epidemiol 2001;153: accidents, traffic; air bags; automobiles; case-control studies; logistic models; mortality; odds ratio; seat belts Recently, reports in both the popular and the scientific literature have questioned the safety and efficacy of air bags in passenger cars. Air bags were introduced in the 1960s and were created to reduce injury associated with frontal impacts. Since 1991, and in response to a US congressional mandate, air bags have been required on the driver side of most passenger cars and of light trucks and vans. As a result, air bags are now widely used, standard equipment on most passenger cars, light trucks, and minivans. From the time that air bags were introduced through early 2000, 152 deaths have been attributed to air bag deployment in low-severity crashes, and 58 decedents were drivers (1). These deaths have raised questions about air bag safety and efficacy. Despite concerns, the public perceives air bags as an effective and important automobile safety component (2). Previous reports have estimated the mortality reduction attributable to air bags for drivers (3, 4) and for front-seat Received for publication May 5, 2000, and accepted for publication October 20, Abbreviations: CI, confidence interval; FARS, Fatality Analysis Reporting System; OR, odds ratio. 1 Department of Emergency Medicine, Center for Injury Prevention Research and Education, University of New Mexico, Albuquerque, NM. 2 Department of Pediatrics, Intermountain Injury Control Research Center, University of Utah, Salt Lake City, UT. Reprint requests to Dr. Cameron S. Crandall, Department of Emergency Medicine, Ambulatory Care Center, 4-West, Albuquerque, NM ( ccrandall@salud.unm.edu). This work was presented in part at the Annual Meeting of the Society for Academic Emergency Medicine, Washington, DC, May 19 22, passengers (5). Estimates for drivers range from 24 to 28 percent (4, 5) and are 18 percent (5) for front-seat passengers. Although all studies have documented a benefit, measurement of the joint effect of air bags and seat belts after adjustment for confounding effects has remained difficult. Thus, we undertook a matched case-control study to measure the mortality reduction associated with air bag deployment and seat belt use for drivers involved in head-on passenger car collisions, after adjusting for potential confounders. We hypothesized that using either an air bag or a seat belt would significantly reduce mortality and that using seat belts in combination with air bags would provide a synergistic mortality reduction on the multiplicative scale greater than expected with either restraint system alone. MATERIALS AND METHODS We examined all head-on collisions involving two passenger cars reported to the Fatality Analysis Reporting System (FARS) during , inclusive, for differences in safety restraint use and mortality (6). The FARS system, managed by the National Highway Traffic Safety Administration, contains information on all known motor vehicle crashes in the United States occurring on public highways that involved at least one human death within 30 days of the event. The data set includes information on each vehicle involved in the crash and on each person in the crash, whether deceased, injured, or unharmed. The difference in mortality from a head-on collision was estimated from crashes that involved only two passenger cars. The drivers of these two vehicles constituted a matched 219

2 220 Crandall et al. pair. From the 218,971 total crashes recorded in the FARS database from 1992 to 1997, we identified pairs from the total number of 32,106 (14.7 percent) head-on collisions. We accepted the FARS definitions of passenger cars and head-on collisions. Passenger cars are defined as all standard automobiles (convertibles, sedans, station wagons, etc.) (FARS variable Body Type, values 1 13). This definition excludes pickup trucks, vans, sport utility vehicles, and limousines. A head-on collision refers to a crash in which the front end of one vehicle collides with the front end of another vehicle while the two vehicles are traveling in opposite directions (FARS variable Manner of Collision, value 4) (6). In the data set, drivers were identified by the FARS variable Person Type (value 1, Driver of a motor vehicle in transport). Only those drivers positioned in the front left seat were included (FARS variable Seating Position, value 11). We limited our analysis to crashes involving only two vehicles to simplify the complexity of the crash mechanics. Case and control statuses were defined by the injury severity outcome (FARS variable Injury Severity). Drivers who sustained fatal injuries (value 4) were classified as cases. Drivers in all other injury categories were classified as controls. Drivers whose injuries were of unknown severity were excluded (value 9). Matched pairs were linked by assigning a unique crash identifier (FARS variable ST_CASE) to create 9,859 pairs. Restraint use was measured by two using FARS variables (Air Bag Availability, Deployment and Restraint System Use). An air bag exposure was defined by deployment of the driver side air bag (value 3). Information on shoulder belt, lap belt, and lap-shoulder belt use was taken from the variable Restraint System Use. In the final regression analysis, any combination of the correct use of a lap or shoulder belt was reclassified as any seat belt use. Drivers who were coded as using their belts improperly (n 9) were recoded into the no restraint category. In three instances, the driver was recorded as wearing a motorcycle helmet. These three cases and their matched pairs were excluded from the analysis. Differences in categorical variables between fatal and nonfatal observations were tested against the chi-square distribution, and differences in continuous variables between cases and controls were tested against the t distribution. We hypothesized that vehicle rollover (FARS variable Rollover), vehicle weight (FARS variable Auto Weight), difference in vehicle age within pairs (in years), driver ejection (FARS variable Ejection), driver sex, and driver age (in years) would be potential confounders. Vehicle weights were reported in pounds and were transformed into Système International (SI) units (kilograms) for analysis and reporting. We considered both partial and complete ejection as evidence of ejection from a vehicle. To account for potential confounding related to improvements in crash design (crashworthiness) from one model year to another, we calculated the difference in vehicle age (in years) for the crash pair. The newer car was assigned a value of zero; the older car was assigned the number of years of difference between its own model year and the model year of the newer vehicle. In all analyses involving odds ratios, we used analytical methods that accounted for the matched design effect (7). We used a crude matched-pairs analysis for bivariate analyses involving categorical variables (7). In the analyses involving multiple effects, we used conditional logistic regression models to estimate odds ratios (7, 8). In the final regression model, vehicle weight and difference in vehicle age were entered as continuous variables. For presentation, the odds ratio for vehicle weight was reported for 100 kg increments. Driver age was entered into the model as an elder (aged 65 years or more) versus a younger driver. Finally, we constructed a statistical interaction term for air bag deployment and any seat belt use to test the hypothesis that having an air bag deploy and using a seat belt would provide additional benefit beyond that observed for either factor alone. In all analyses, we used a 5 percent, two-tailed, type I error rate to determine statistical significance, as evidenced by either a p value or a two-tailed 95 percent confidence interval. The University of New Mexico Health Sciences Center institutional review board approved the study design. RESULTS We identified 9,859 head-on crash pairs involving 19,718 passenger cars and drivers (table 1). Of the 19,718 drivers, 8,798 died. Not all crash events resulted in a driver fatality. In 11.3 percent of the crash pairs, both drivers died (n 1,116); in 66.7 percent of the pairs, one driver died and one survived (n 6,573); and in 22.1 percent of the pairs, neither driver died (n 2,177). When neither of the drivers died, either a passenger or a pedestrian died in the crash. Most drivers were male (62.1 percent). Of the vehicles involved in head-on crashes, 11.3 percent were equipped with an air bag. The majority of the air bags in vehicles so equipped had deployed (2,080 of 2,228 (93.4 percent)). More than half of all drivers (54.1 percent) used some form of a harness or seat belt restraint system. Among those drivers who wore seat belts, the lap and shoulder belt combination was most common (8,848 (83.0 percent)); a small number of drivers used a lap belt (253 (2.4 percent)) or shoulder belt (170 (1.6 percent)) only. Thirteen percent of belted drivers used an unspecified seat belt arrangement. Over a third of all drivers (37.7 percent) were recorded as using no restraint. Comparison of crash outcomes revealed significant differences for both drivers and vehicles (table 1). Drivers who survived were more likely to have had an air bag deploy (12.8 vs. 7.8 percent, p < ) or to have used any seat belt restraint (64.1 vs percent, p < ) and were 5.1 years younger on average than drivers who died (p < ). Drivers who were ejected from the vehicle or whose vehicle rolled over were more likely to die (p < for both). A lower vehicle weight was associated with an increased probability of a fatal outcome for the driver (fatal crash: mean vehicle weight, 1,225 kg; nonfatal crash: mean vehicle weight, 1,338 kg; p < ). In the crude matched-pair analysis, increasing vehicle mass was associated with decreased driver mortality compared with driver mortality in the lightest vehicles (table 2). As the vehicle age difference between the crash pairs increased, the

3 Air Bag and Seat Belt Efficacy 221 TABLE 1. Selected demographic characteristics of drivers involved in head-on collisions of passenger cars, United States, * Total Fatal crash Nonfatal crash No. % No. % No. % Crash and vehicle characteristics Vehicle rollover Driver ejected Vehicle weight (kg) <1,000 1,000 1,499 1,500 Unknown 2,263 12,523 3,689 1, ,426 5,659 1, ,864 2, Difference in vehicle age (years) Unknown Driver characteristics Male gender Age group (years) < Unknown Restraint system use Air bag Present Deployed Any seat belt restraint Lap and shoulder belt Lap belt only Shoulder belt only Used, type unknown No or improper use Unknown or data missing Total 10,426 3,432 2,358 3, ,244 5,535 4,407 3,392 3,832 2, ,228 2,080 10,660 8, ,389 7,442 1,616 19,718 * Source: Fatality Analysis Reporting System (6) ,354 1,586 1,146 1, ,408 2,169 1,848 1,476 1,736 1, ,655 3, , , ,072 1,846 1,212 1, ,836 3,366 2,559 1,916 2, ,485 1,394 7,005 5, , , mortality risk also increased for the driver of the older vehicle. Drivers whose vehicles rolled over (odds ratio (OR) 2.54, 95 percent confidence interval (CI): 2.07, 3.11) or who were ejected from their vehicles (OR 4.62, 95 percent CI: 3.75, 5.69) were more likely to die than drivers whose vehicles did not roll over or who were not ejected. Males were more likely to survive compared with females (OR 0.92, 95 percent CI: 0.85, 0.99). Compared with drivers less than 25 years of age, only elder drivers (aged 65 years or more) had an increased mortality rate (OR 3.4, 95 percent CI: 2.7, 4.3). The odds of dying in a head-on collision were reduced by approximately two-thirds when the driver was in a car whose air bag deployed compared with one whose air bag did not deploy or was not equipped with an air bag (OR 0.37, 95 percent CI: 0.32, 0.42). Wearing a lap belt only (OR 0.43), a combined lap-shoulder belt (OR 0.28), or an unspecified belt type (OR 0.29) all conferred a strong protective effect for the driver. Wearing only a shoulder belt, however, increased the odds of dying (OR 2.19) relative to wearing no belt. Despite this observation, and in part because of the very small number of pairs in this shoulderbelt-only group, we aggregated all types of seat belt use into an any belt use category for further analyses. After adjustment for vehicle rollover, vehicle weight, difference in vehicle age, driver age, and driver sex, both air

4 222 Crandall et al. TABLE 2. Crude and adjusted analyses of crash pairs of drivers involved in head-on collisions of passenger cars, United States, * Crude analysis Adjusted analysis Crash and vehicle characteristics Vehicle rollover Driver ejected Vehicle weight (kg) <1,000 1,000 1,499 1,500 Difference in vehicle age (years) Driver characteristics Male vs. female gender Age group (years) < Restraint system use Air bag deployed Any seat belt combination Lap belt only Shoulder belt only Lap and shoulder belt Used, type unknown Case exposure + Control exposure , ,299 1,440 No. of pairs bags and seat belts showed significant protective effects (table 2). Crash mortality was lower for males than for females (OR 0.79, 95 percent CI: 0.70, 0.88). Mortality among drivers whose air bags deployed was lower compared with that of drivers of cars without air bags or whose air bags did not deploy (OR 0.71, 95 percent CI: 0.58, 0.87). In addition, mortality was lower for drivers who used any combination of seat belts (OR 0.25, 95 percent CI: 0.22, 0.29). Mortality was substantially reduced when drivers used both an air bag and a seat belt (82 percent) (OR 0.18, 95 percent CI: 0.13, 0.25). Finally, to test the hypothesis that having an air bag deploy in combination with wearing a seat belt conferred a synergistic effect, we constructed a model with a statistical interaction term for their combined use. We failed to find evidence of an excess crash mortality reduction beyond the Case exposure Control exposure , ,007 1, , , Odds ratio expected multiplicative effect of using both restraint systems (air bag alone: OR 0.72; seat belt alone: OR 0.25; joint air bag and seat belt use with interaction: OR 0.18, p value for interaction term 0.91). DISCUSSION 95% confidence interval 2.07, , , , , , , , , , , , , , , , , , 0.36 Odds ratio % confidence interval 1.82, , , , , , , 0.29 * Source: Fatality Analysis Reporting System (6). Odds ratio estimates for the conditional logistic model were adjusted for all covariates for which estimates are provided. Odds ratio estimates for vehicle weight are per 100 kg, for difference in vehicle age are per year, and for age group are aged 65 years vs. others. Model based on 8,454 matched pairs. Refer to Materials and Methods in the text. We found a substantial reduction in mortality attributable to use of either air bags or seat belts. Driver air bags independently reduced head-on passenger car crash mortality by one fourth; wearing a seat belt reduced mortality by three fourths. In a head-on collision, both having an air bag deploy and wearing a seat belt were found to reduce the driver s risk of dying by over 80 percent. Our findings demonstrate a dramatic reduction in the risk of death from head-on collisions for drivers who use their seat belts and whose air bags deploys.

5 Air Bag and Seat Belt Efficacy 223 Evans (9, 10) is credited with first using a paired methodology to estimate the efficacy of seat belts in motor vehicle collisions. By comparing driver mortality with that of other occupant(s) in the same vehicle, he was able to estimate a mortality difference attributable to seat belt use. To our knowledge, ours is the first study to compare driver mortality from paired collisions and to measure the effect of air bags and seat belt systems on crash mortality. By using this strategy, we were able to precisely estimate the protective effect of both air bags and seat belts. By taking advantage of the inherent matched design of head-on crashes, we effectively matched on a number of potential confounding effects including, but not limited to, time of day; weather; lighting; roadway surface type, condition, and class (e.g., interstate, arterial); vehicle speed; and emergency medical response. Previous studies that have measured the mortality reduction due to air bags have estimated a reduction of from 24 to 28 percent in frontal crashes (3, 4). The crude odds ratio obtained from our study showed a substantially greater effect attributable to air bags (OR 0.37), and it remained significant after adjustment for concomitant seat belt use and other crash factors (OR 0.71). The greater observed effect may have been due to our analysis method. Previous studies have estimated the efficacy of air bags by comparing either frontal to nonfrontal crash mortality rates (3 5) or the observed fatality rates per 10,000 registered vehicles (4) with respect to the presence or absence of air bag equipment. Each method has yielded similar estimates and is subject to a limited ability to control for seat belt use and other confounders. Because air bags were designed to be most protective in frontal crashes and our study specifically looked at this type of crash, our results likely represent the best expected mortality reduction from air bags and seat belts. As a result, our results may be difficult to generalize to other crash types. We designed this study to measure the mortality reduction attributable to restraint use in head-on crashes of passenger cars. While air bags were designed to reduce the transfer of energy in a head-on collision, they may have some (albeit reduced) effect in other types of crashes, as the collision forces in most crashes are complex. We limited our analysis to passenger cars to minimize other potential confounding crash effects. For example, although we controlled for the weight difference between the two vehicles, we did not adjust for other factors such as vehicle height differences (11). For several reasons, we elected to measure exposure to an air bag as air bag deployed as opposed to air bag present, which differs from previous studies (3 5). First, deployment of the air bag and the resulting slowed deceleration of the occupant was the factor of interest. The mere presence of an air bag does not confer mortality reduction. Whereas drivers must fasten their seat belts to activate the protective effects, crash mechanics determine whether an air bag deploys. Second, we anticipate that should manufacturers adjust air bags to inflate with less force or should occupants install and use manual air bag cutoff switches, documenting air bag deployment as opposed to air bag presence will become very important to monitor the outcome of these manufacturing and policy changes. Finally, measuring actual air bag deployment will allow crash investigators to monitor and investigate cases of nondeployment in severe crashes to identify equipment failures and to improve safety designs. In our crude analysis, driver ejection from the vehicle was associated with a significantly greater odds of mortality. We elected not to enter driver ejection into the multivariate model because it lies between seat belt use and mortality in the causal chain. We are in the process of examining the role of seat belts, air bags, and driver ejection by using a similar analysis but with differential ejection between the crash pairs as the outcome of interest (as opposed to differential mortality). Air bag and seat belt use have been examined for their effect on the risk of ejection from the vehicle (12). Our initial findings from the crude analysis suggest that both seat belts (OR 0.03, 95 percent CI: 0.02, 0.05) and air bags (OR 0.27, 95 percent CI: 0.18, 0.41) reduce the risk of ejection. In the multivariate analysis, however, only seat belts remained significant. While the role of seat belt use in reducing ejection from the vehicle (and subsequent reduced morbidity and mortality) appears clear, the relation between air bag deployment and vehicle ejection deserves further study. Vehicle crashworthiness is another factor that may affect occupant safety. In our study, drivers of older vehicles had a greater mortality risk compared with drivers of newer vehicles, suggesting that newer vehicles are more crashworthy. We calculated the difference (in years) between the ages of the two vehicles involved in each collision and used this difference as a measure of vehicle crashworthiness. While simple in construct, crashworthiness is difficult to measure. Since the late 1970s, the federal government has tested passenger vehicles for crashworthiness with the New Care Assessment Program by crashing vehicles in simulated frontal collisions (13). The forces delivered to crash dummies in these standardized crashes, and assessment of the vehicle s external and internal structural integrity following such crashes, are used to measure crashworthiness. These measurements rely on sophisticated instruments and computer modeling techniques not available for everyday car crash investigations. Observational studies have used vehicle and model year as a proxy measurement for crashworthiness (13, 14). Use of vehicle age as a measure of crashworthiness may introduce other confounding and biases, as driver behavior (including alcohol consumption, seat belt use, and aggressive driving) may be related to the type and age of the vehicle driven. We observed a higher mortality rate for females involved in head-on crashes compared with males that persisted despite adjustment for other effects. We suspect that the physical size difference between men and women may explain this observation (15). Unfortunately, the FARS data set does not contain information on the height or weight of either drivers or passengers. Future studies should investigate this observation of gender difference. Both seat belt use and air bag exposure measurement are subject to reporting biases. Seat belt use is based on police data and is likely subject to more reporting bias than air bag use is. Passengers may incorrectly self-report seat belt use,

6 224 Crandall et al. or police officers may conclude that seat belts were either worn or not worn based on crash outcomes (e.g., mortality and occupant ejection) (16). Seat belts may be fastened or unfastened after the crash event. The net effect is probably an exaggeration of the protective effects of seat belts. Air bag deployment is an objective measurement that is likely recorded more accurately than seat belt use is. A deployed air bag is evident when a vehicle is inspected after the crash and cannot change to the undeployed state. Unfortunately, reports of adverse effects of air bags may have overshadowed reports of their efficacy. Since air bags were introduced, there have been numerous reports of nonfatal injuries from air bag inflation, including eye (17 19), face (20, 21), upper extremity (22), and thoracoabdominal injuries (23, 24). Reports of fatalities among children, smallstature adults, and out-of-position front-seat passengers whose air bags deployed in low-speed motor vehicle crashes are most disturbing, however (25). Despite the well-known adverse effects of seat belt use, including neck, chest, abdominal, spine, and fetal injuries (26), all 50 US states and the District of Columbia have passed child restraint laws, and many have passed primary seat belt laws. Most public health interventions have associated costs. Several widely accepted public health interventions, including vaccinations and childproof medicine bottles, have adverse effects. On balance, the public accepts these interventions because the benefits far outweigh the risks. Our results demonstrate the remarkable benefit of air bags and seat belts in reducing deaths in head-on passenger car collisions and may temper the reports in the literature of adverse effects. Despite recent concerns about air bag safety, our study demonstrates a substantial reduction in mortality associated with a head-on collision when an air bag deploys. The public health community should continue to urge manufacturers to improve air bags while promoting the benefits of using both an air bag and a seat belt to prevent death in head-on crashes. REFERENCES 1. Safety facts: airbag statistics. Arlington, VA: Insurance Institute for Highway Safety, Highway Loss Data Institute, November ( htm). 2. Graham JH, Thompson KM, Goldie SJ, et al. The costeffectiveness of air bags by seating position. JAMA 1997;278: Zador PL, Ciccone MA. Automobile driver fatalities in frontal impacts: air bags compared with manual belts. Am J Public Health 1993;83: Lund AK, Ferguson SA. Driver fatalities in cars with airbags. J Trauma 1995;38: Braver ER, Ferguson SA, Greene MA, et al. Reductions in deaths in frontal crashes among right front passengers in vehicles equipped with passenger air bags. JAMA 1997;278: National Highway Traffic Safety Administration. Fatality Analysis Reporting System, (Public access data set). Washington, DC: US Department of Transportation. 7. Schlesselman JJ. Case-control studies: design, conduct, analysis. New York, NY: Oxford University Press, SAS Institute, Inc. SAS/STAT software: the PHREG procedure, version Cary, NC: SAS Institute, Inc, (SAS technical report P-217). 9. Evans L. Double pair comparison a new method to determine how occupant characteristics affect fatality risk in traffic crashes. Accid Anal Prev 1986;18: Evans L. The effectiveness of safety belts in preventing fatalities. Accid Anal Prev 1986;18: Lund AK, Chapline JF. Potential strategies for improving crash compatibility in the U.S. vehicle fleet. Reprinted from: aggressivity and compatibility in automotive crashes (SP-1442). Warrendale, PA: Society of Automotive Engineers, Inc, (SAE technical paper series ). 12. Crandall CS, Sklar DP. Wear your seat belts! Driver ejection in motor vehicle crashes. (Abstract). Acad Emerg Med 2000;7: Roberston LS. Reducing death on the road: the effects of minimum safety standards, publicized crash tests, seat belts, and alcohol. Am J Public Health 1996;86: Evans L, Frick MC. Car mass and fatality risk: has the relationship changed? Am J Public Health 1994;84: Dischinger PC, Kerns TJ, Kufera JA. Lower extremity fractures in motor vehicle collisions: the role of driver gender and height. Accid Anal Prev 1995;27: Dean JM, Reading JC, Nechodom PJ. Overreporting and measured effectiveness of seat belts in motor vehicle crashes in Utah. Washington, DC: National Academy Press, (Transportation research record no. 1485, safety and human performance). 17. Duma SM, Kress TA, Porta DJ, et al. Air bag-induced eye injuries: a report of 25 cases. J Trauma 1996;41: Gault JA, Vichnin MC, Jaeger EA, et al. Ocular injuries associated with eyeglass wear and air bag inflation. J Trauma 1995; 38: Smally AJ, Binzer A, Dolin S, et al. Alkaline chemical keratitis: eye injury from air bags. Ann Emerg Med 1992;21: Foley S, Mallory SB. Air bag dermatitis. J Am Acad Dermatol 1995;33(5 pt 1): Rozner L. Air bag-bruised face. (Letter). Plast Reconstr Surg 1996;97: Huelke DF, Moore JL, Compton TW, et al. Upper extremity injuries related to air bag deployments. J Trauma 1995;38: Augenstein JS, Digges KH, Lombardo LV, et al. Occult abdominal injuries to air bag-protected crash victims: a challenge to trauma systems. J Trauma 1995;38: Dunn JA, Williams MG. Occult ascending aortic rupture in the presence of an air bag. Ann Thorac Surg 1996;62: McKay MP, Jolly T. A retrospective review of air bag deaths. Acad Emerg Med 1999;6: Hendey GW, Votey SR. Injuries in restrained motor vehicle accident victims. Ann Emerg Med 1994;24:77 83.