Effects of Child Age and Body Size on Serious Injury From Passenger Air-Bag Presence in Motor Vehicle Crashes
http://www.100md.com
《小儿科》
Center for Policy and Research in Emergency Medicine, Department of Emergency Medicine, Oregon Health and Science University, Portland, Oregon
David Geffen School of Medicine, University of California, Los Angeles, California
Department of Emergency Medicine, Harbor-University of California, Los Angeles, Medical Center, Torrance, California
Los Angeles Biomedical Research Institute, Torrance, California
ABSTRACT
Background. Current recommendations regarding children traveling in passenger vehicles equipped with passenger air bags are based, in part, on evidence that the air-bag–related risk of injury and death is higher for children 12 years of age. However, the age or body size required to allow a child to be seated safely in front of a passenger air bag is unknown.
Objective. To evaluate specific cutoff points for age, height, and weight as effect modifiers of the association between the presence of a passenger air bag and serious injury among children involved in motor vehicle crashes (MVCs), while controlling for important crash factors.
Design. A national population-based cohort of children involved in MVCs and included in the National Automotive Sampling System (NASS) Crashworthiness Data System (CDS) database from 1995 to 2002 was studied. NASS CDS clusters, strata, and weights were included in all analyses.
Subjects. Children 0 to 18 years of age involved in MVCs and seated in the right front passenger seat.
Main Outcome Measure. Serious injury, defined as an Abbreviated Injury Scale score of 3 for any body region.
Results. A total of 3790 patients (1 month to 18 years of age) were represented in the NASS CDS database during the 8-year period. Sixty children (1.6%) were seriously injured (Abbreviated Injury Scale score of 3). Among age, height, and weight, age of 0 to 14 years (versus 15–18 years) was the only consistent effect modifier of the association between air-bag presence (or air-bag deployment) and serious injury, particularly for crashes with a moderate probability of injury. In analyses stratified according to age and adjusted for important crash factors, children 0 to 14 years of age involved in frontal collisions seemed to be at increased risk of serious injury from air-bag presence (odds ratio [OR]: 2.66; 95% confidence interval [CI]: 0.23–30.9) and deployment (OR: 6.13; 95% CI: 0.30–126), although these values did not reach statistical significance. Among children 15 to 18 years of age involved in frontal collisions, there was a protective effect on injury from both air-bag presence (OR: 0.19; 95% CI: 0.05–0.75) and deployment (OR: 0.31; 95% CI: 0.09–0.99). These findings persisted in analyses involving all collision types. We did not identify similar cutoff points for height or weight.
Conclusions. Children up to 14 years of age may be at risk for serious preventable injury when seated in front of a passenger air bag, and children 15 to 18 years of age seem to experience protective effects of air-bag presence and deployment. Age may be a better marker than height or weight for risk assessment regarding children and air bags.
Key Words: air bag motor vehicle crash children age height weight
Abbreviations: NASS, National Automotive Sampling System NHTSA, National Highway Traffic Safety Administration MVC, motor vehicle crash CDS, Crashworthiness Data System AIS, Abbreviated Injury Scale OR, odds ratio CI, confidence interval
In 1995, the National Highway Traffic Safety Administration (NHTSA) issued a press release with a strong warning that air bags may pose significant risks of injury and death for children.1 The warning was prompted by the results of an investigation of crashes in which deployment of the passenger-side air bag resulted in critical or fatal injuries to children seated in the right front seat.1 The following year, NHTSA published a final rule requiring warning labels on all new production passenger vehicles beginning in 1997.2 In addition to several well-supported recommendations regarding seat position and restraint use, the warning label stated, "Children 12 and under can be killed by the air bag."2 Similar labeling requirements were instituted for child safety seats.2
Evidence supporting the recommendations for age-appropriate restraints3–11 and rear seat locations10–13 for children traveling in motor vehicles has been well established. However, evidence for the age-based recommendations regarding air bags and children is lacking. Although the NHTSA guidelines for children and air bags have been circulated widely, the age (or body size) that best defines when a child's additional risk of injury or death from an air bag is replaced by benefit is unknown. Two initial studies suggested an association between air bags and childhood fatalities (defining children as 0–12 years and 0–9 years of age), but both analyses were limited by relatively small numbers of fatalities, yielding statistically inconclusive results.14,15 Subsequent studies defined children as 12 years of age and provided more conclusive evidence for the association between air bags and death among right front seat passengers within this age range.16–18 Kahane14 suggested that the increased mortality risk from air bags was most pronounced among right front seat passengers through age 10, becoming less pronounced from 11 to 14 years and turning to a net benefit for passengers 15 years of age, but the results did not reach statistical significance. A more recent study examining differences in air-bag effectiveness according to age and restraint use suggested a net increase in the risk of death among child passengers 12 years of age, but sample size limitations prevented definitive statements regarding the childhood age at which the net risk might change to no effect (or benefit) and the study did not assess markers of body size.18 Durbin et al19 demonstrated that the risk of injury (rather than death) among restrained children 3 to 15 years of age exposed to passenger air-bag deployment was twice that among front-seated children not exposed to an air bag, with the risk of injury being relatively constant among children 3 to 8, 9 to 12, and 13 to 15 years of age.
We hypothesized that specific cutoff points in age, height, and/or weight among children could be used to define when the risk of serious injury from the presence of a passenger air-bag changes from harmful to no effect (or beneficial), after adjustment for crash severity and other important crash factors. We tested this hypothesis with age, height, and weight as effect modifiers (interaction terms) of the association between the presence of a passenger air bag and serious injury among right front seat passengers 0 to 18 years of age involved in motor vehicle crashes (MVCs).
METHODS
Study Design and Setting
Subjects included in the National Automotive Sampling System (NASS) Crashworthiness Data System (CDS) database from 1995 through 2002 were used in the study. The NASS CDS is a probability-sampled, population-based, nationally representative cohort of persons involved in MVCs. Data in the NASS CDS database were collected through 3-stage sampling of MVCs from specific regions throughout the United States,20 to ensure generalizability of the data to the nation as a whole. The 8-year time period was selected because it encompasses the largest, most current block of data with comprehensive air-bag information in the NASS CDS database and because data collection changes for air bags were instituted in 1995. During this time period, increased production and eventual mandatory inclusion of both driver and passenger air bags in passenger vehicles occurred. By 1995, 98% of new cars sold were equipped with driver and passenger air bags.1 The study was approved by the institutional review boards of Oregon Health and Science University and the Los Angeles Biomedical Research Institute.
Subjects
A national, population-based sample of children 0 to 18 years of age who were seated in the right front seat and involved in MVCs with passenger vehicles or light trucks, as included in the NASS CDS database, was included in the analysis. The study included only children because previous studies have demonstrated a higher risk of injury or death among adults, compared with children, when controlling for other important crash characteristics.21,22 We defined "children" as persons 0 to 18 years of age because this age range allowed for adequate comparison samples for many age, height, and weight cutoff points while minimizing factors such as alcohol and drug use between groups and because US growth charts encompass this full age range.
Main Outcome Measure
Serious injury, the primary outcome, was defined as an Abbreviated Injury Scale (AIS) score of 3 for any body region. The AIS for a given injury ranges from 1 (minor) to 6 (nonsurvivable). A score of 3 represents a serious injury.23
Patient and Vehicle Characteristics
The exposure variable of interest was the presence of a passenger-side front air bag. Previous studies have defined air-bag exposure as air-bag presence rather than deployment, to account for variability in deployment and effects after activation.14–18 This definition has been suggested as a means of assessing simultaneously the effect of air-bag activation and the effect of the air bag after activation.24 There were only 13 cases (<1%) during this time period in which the air bag had been disconnected or removed from the vehicle. In these instances, the air-bag variable was coded as not present because of an inability to deploy.
In addition to air-bag information, variables were selected on the basis of known associations with injury and crash severity, as well as age and anthropometric measures for children. Thirteen variables were considered in the analysis: air-bag status (described above), age (years), height (centimeters), weight (kilograms), restraint use (manual lap or lap and shoulder belt, automatic belt system, or child safety seat with belt use), direction of impact (frontal, left lateral, or right lateral, with a reference group of rear, undercarriage, and top crashes), rollover with collision, entrapment, steering wheel deformity (in increments of 1 cm), vehicle model year (1998 versus earlier model years), vehicle weight,25 change in velocity, and passenger space intrusion (<15, 15–29, 30–45, 46–60, or >60 cm).26 We used change in velocity, passenger space intrusion, steering wheel deformity, rollover, and entrapment to adjust for crash severity, the main confounding factor in the analysis. Change in velocity, passenger space intrusion, vehicle weight, and steering wheel deformity were coded as continuous variables, whereas the remaining nonanthropometric covariates were coded as dichotomous categorical variables. In addition to the dichotomous interaction terms (described below), age, height, and weight were entered into the analysis as continuous variables, to test for a linear association with injury. Because some automotive manufacturers began implementing second-generation air bags in 1998,1 we included a dichotomous covariate for model year (1998 versus earlier model years) as a proxy for vehicles equipped with depowered air bags.
Age (range: 0–18 years), height (range: 30–203 cm), and weight (range: 2–150 kg) were coded as dichotomous categorical variables for use in interaction terms, with multiple different cutoff points. The dichotomous terms (ie, cutoff points) considered were as follows: 8 to 17 years for age, in increments of 1 year (eg, 0–8 vs 9–18 years); 110 to 180 cm for height, in increments of 10 cm (eg, 30–110 vs 111–203 cm); 30 to 75 kg for weight, in increments of 5 kg (eg, 2–30 vs 31–150 kg). Potential cutoff points were selected on the basis of results from previous studies14,18,19 and standard US male and female growth charts for children,27,28 as well as to represent a wide range of potential strata for age and body size. In total, 27 different potential cutoff points were tested.
Statistical Analyses
All analyses were restricted to children seated in the right front passenger seat to isolate the effects of air-bag presence and to control for the differential risk of injury according to seat position. Noncollision crashes and pure rollovers were excluded to allow for the appropriate use of changes in velocity to control for crash severity and because noncollision crashes are unlikely to result in air-bag deployment. After exclusion of noncollision crashes (n = 1019), pure rollovers (n = 313), pregnant female subjects (n = 35), and subjects with missing air-bag information (n = 5), the 1995–2002 NASS CDS database contained 3790 subjects 0 to 18 years of age. All children 0 to 18 years of age, regardless of body size, were included in the analysis. Although we note the total unadjusted number of children included in the analysis (n = 3790), we refer to the NASS CDS sample-adjusted number and proportion of occupants throughout the article, rather than the unadjusted numbers or national estimates.
To allow inclusion of all pediatric subjects contained in the NASS CDS database during this time period and to preserve the original weighting scheme of the NASS CDS database, we used multiple imputation to impute missing values.29–35 We imputed missing values by using parallel chains of multiple imputation, split on the presence or absence of a passenger air bag, to maximize the statistical efficiency of assessing interaction terms (age, height, and weight) with the air-bag variable.33 Variables for cluster, strata within clusters, and year were included as fixed effects in the imputation models to preserve the complex sampling design features of NASS CDS.34,36
Age-, height-, and weight-dependent cutoff points were investigated through interaction terms with the presence of a passenger air bag in multivariate logistic regression models.37,38 These models accounted for the complex sampling design of NASS CDS and included the additional variables noted above. We considered passenger air-bag presence the focal independent variable in the interaction term, with the age, height, or weight term as an effect modifier.37 Analyses were first performed on the subset of crashes that involved primary or secondary frontal impacts (defined as a primary or secondary direction of force in the 10 o'clock to 2 o'clock position of the vehicle), because these collisions are most likely to cause air-bag deployment and to confer the most potential benefit from air bags for adult passengers.1 Subsequent analyses were conducted with all crash types and included a covariate for direction of impact (ie, frontal, left lateral, or right lateral).
Results for the interaction terms are presented as P values because of the difficulty of interpreting the odds ratio (OR) of an interaction term. For terms that suggested the presence of an interaction (P < .10), we performed stratified analyses to assess the role of such effect modifiers and to quantify these associations in terms that would be clinically meaningful. In addition, we calculated the cross-derivative (cross-difference) of the interaction effect for interaction terms identified above and plotted this value against the probability of serious injury, as calculated with the multivariate model.39 As a qualitative check of the consistency of these results, all methods described above were repeated with the use of passenger air-bag deployment instead of passenger air-bag presence. To assess modification of the air-bag effect by restraint use and gender, we tested additional interaction terms with these covariates. In this article, we use the term risk synonymously with odds because the outcome (serious injury) was rare (ie, <2% of the sample). To explore the effects of multiple imputation and different potential underlying patterns of missing data on the results, we conducted sensitivity analyses of the imputation models with different patterns of missing data for variables with >5% missing data in a hypothetical data set with identical sample size and variables.
Database management was performed with SAS version 8.1 software (SAS Institute, Cary, NC). SAS-callable IVEware (survey methodology program; Survey Research Center, Institute for Social Research, University of Michigan, Ann Arbor, MI) was used for multiple imputation and multivariate analyses to account for the complex sampling design of the NASS CDS database and to ensure appropriate variance calculations.40 Stata software (version 8; Stata Corp, College Station, TX) was used to calculate the cross-derivative of the interaction effect.
RESULTS
We studied 3790 right front seat child occupants (age range: 1 month to 18 years), representing a national population of 2055390 children involved in MVCs during the 8-year period. A total of 2535 children (67%) were involved in primary or secondary frontal collisions. No children were excluded because of missing outcome (AIS score) information. Sixty children (1.6%) had at least 1 injury with an AIS score of 3. There were 10 fatalities (17% of seriously injured children) in the outcome group. Characteristics for right front seat occupants, according to air-bag presence versus absence, are listed in Table 1.
In multivariate regression models for children seated in the right front seat and involved in frontal collisions, age cutoff points of 14 years (interaction term: P = .054) and 15 years (interaction term: P = .074) were identified as potential effect modifiers of the association between air-bag presence and serious injury. When the same interaction terms were included in models assessing air-bag deployment (rather than air-bag presence), the results were more pronounced (interaction term for age 14 years: P = .019; interaction term for age 15 years: P = .024). These cutoff points persisted in models that included all types of collisions but had less statistical contribution. There were no dichotomous cutoff points in height or weight that modified the association between an air bag and serious injury, regardless of collision type and whether the air-bag variable was coded as presence or deployment (all P > .20). In addition, there was no linear relationship between height or weight and injury in any of the models (all P > .20).
In analyses stratified according to the age cutoff points identified above, an age split at 0 to 14 vs 15 to 18 years provided the most consistent risk strata in all models. Among children 0 to 14 years of age involved in frontal collisions, point estimates suggested an increased risk of serious injury from air-bag presence (OR: 2.66; 95% confidence interval [CI]: 0.23–30.9) and deployment (OR: 6.13; 95% CI: 0.30–126), although these values did not reach statistical significance. Among children 15 to 18 years of age involved in frontal collisions, there were protective effects on injury from both air-bag presence (OR: 0.19; 95% CI: 0.05–0.75) and deployment (OR: 0.31; 95% CI: 0.09–0.99). These findings persisted in stratified analyses involving all collision types (Table 2).
The interaction effect (cross-derivative) between air-bag presence and age of 0 to 14 years (versus 15–18 years) was positive for all probabilities of serious injury, although there was a more pronounced statistical contribution in crashes with a moderate probability of serious injury (Fig 1). For children with a very high or very low probability of injury, the interaction effect was negligible. We found similar but more pronounced results when the air-bag variable was defined as air-bag deployment (data not shown).
The use of restraints and gender did not modify the association between air-bag presence (or deployment) and injury (for interaction terms in the full and stratified models, all P > 0.30). There were no qualitative differences or indications of systematic bias in the results of sensitivity analyses with different patterns of missing data.
NASS CDS national estimates for the percentage of children seated in the right front seat demonstrated that the majority of children 13 to 14 years of age (67% in 2002) sit in the front passenger seat, regardless of the presence of an air bag. With the same national estimates, there has been a substantial decrease in the number of children 0 to 12 years of age seated in the right front seat since the introduction of air-bag warning labels in 1997 (15% in 2002). However, the number of children >12 years of age seated in the right front seat has not changed significantly since 1997, even when 13- to 14-year-old children are considered separately (data not shown).
DISCUSSION
These results suggest that, among occupants 0 to 18 years of age seated in the right front seat of a vehicle equipped with a passenger air bag, children 14 years of age have the greatest odds of serious injury, particularly in crashes with a moderate probability of injury. However, both air-bag presence and deployment seem to have a protective effect on injury among older children (15–18 years). This age cutoff point (ie, 0–14 years versus 15–18 years) is consistent with results from a previous study that assessed the age at which fatality risk from air bags changed from harmful to net beneficial14 and a study that assessed the risk of injury among 3- to 15-year-old restrained children.19 Although the effect of passenger air-bag presence on the risk of serious injury seemed to be influenced by age, we were unable to identify any effect modification with child height or weight.
Child developmental changes associated with puberty may provide one explanation for these findings. Although there is variation in the timing and duration of puberty among children, the onset of puberty typically occurs at 11 years of age among girls and 13 years among boys.41 There are increases not only in weight and height but also in body composition, including lean body mass, bone mineral content, and bone density (bone mass), during puberty.41,42 In addition, age has been shown to be a better determinant of pubertal development than anthropometric measurements.43 If changes in body composition and bone mass during early puberty play a role in susceptibility to injury from air bags in MVCs, it may not be surprising to demonstrate age as a better discriminator for risk of injury than height or weight. Despite the differences in onset of puberty between boys and girls, we did not find gender to be a significant effect modifier.
The age cutoff point identified in this analysis is older than that suggested by several previous studies15–18 and the current recommendations by NHTSA.2 Although the broad application and distribution of NHTSA guidelines regarding children traveling in vehicles with passenger air bags have been credited with reducing the number of air-bag-associated deaths among children,44 the results presented here suggest that there may be a significant proportion of at-risk children being missed by the current recommendations.
Restraint use has been suggested as a modifier of the association between air bags and death among children,16,18 with one study demonstrating that certain groups of younger restrained children (ie, 9–12-year-old children) actually have a survival benefit associated with the air bag.16 Our results failed to demonstrate such effect modification by restraint use. However, our finding that younger children were at risk of injury from an air bag despite restraint use is consistent with several other studies.17–19,45,46
There are several potential limitations in our analysis. The NASS CDS data set provided a relatively limited number of children who were seated in the right front seat and seriously injured in a vehicle equipped with a passenger air bag. Attempts to use interaction terms in multivariate models to demonstrate effect modification generally require large sample sizes to avoid type II error. It is possible that this sample and the number of subjects with serious injury were not large enough to demonstrate the effect modification by height and weight, and we cannot exclude the possibility that a larger sample of such children might suggest height or weight cutoff points for children and air-bag risk.
Our inability to find specific cutoff points in height or weight might have several additional explanations, ie, (1) multiple imputation was required for a portion of missing height and weight data, (2) values for height and weight in the NASS CDS data set might not have been accurate for all children, (3) the relationship between height or weight and injury might be more complex than can be tested with a dichotomous cutoff point, and (4) the risk strata for body size might be better estimated with a combination of height and weight values, which we did not assess. Of these possibilities, we do not suspect that the multiple imputation process adversely affected our ability to detect height or weight cutoff points because we performed sensitivity analyses for all variables with >5% missing data (including height and weight) with 3 different patterns of missing data (missing completely at random, missing at random, and missing in a nonrandom pattern). The nonrandom pattern of missingness simulated a scenario in which missing values for height and weight were associated with less serious crashes and less severe injuries. There was no indication of any qualitative change or systematic bias in the results secondary to multiple imputation, and our results were robust with all patterns of missing data.
There is also the possibility that the methods we used to identify cutoff points for age, height, and weight (ie, the use of dichotomous interaction terms) might not identify the ideal risk strata. Modeling age as a continuous (rather than dichotomous) variable in the interaction term produced similar point estimates for when the odds of serious injury began to change; however, small numbers of children in each age group reduced the precision of estimates and created difficulty in drawing firm conclusions.
We attempted to adjust for different types of air bags (eg, depowered air bags) by using vehicles with model year of 1998 versus earlier models as a proxy for vehicles equipped with depowered air bags. This variable represents not only the effect of newer-generation air bags in certain vehicles starting in 19981 but also that of other safety features included in newer-model vehicles. When we stratified the sample according to model year of 1998 versus older vehicles and assessed the same interaction terms for age, height, and weight, age of 14 years was the only cutoff point that persisted for both newer and older vehicles, but results were limited by the small number of newer-model vehicles.
CONCLUSIONS
An age of 0 to 14 years (versus 15–18 years) modified the effect of passenger air-bag presence on serious injury among children seated in the right front seat and involved in MVCs. Children 0 to 14 years of age seem to be at increased risk of injury from a passenger air bag, particularly in crashes with a moderate probability of injury, whereas the presence of an air bag has a protective effect on serious injury among children 15 to 18 years of age. We were unable to identify similar cutoff points for height or weight. The association with childhood developmental changes (measured by age) should be considered in the ongoing effort to reduce unnecessary air-bag–related injuries and deaths among children, because age may be a better marker than height or weight for risk assessment for children and air bags. On the basis of these results, children 0 to 14 years of age should not be seated in the right front passenger seat of vehicles equipped with a passenger air bag.
ACKNOWLEDGMENTS
This project was supported by Agency for Healthcare Research and Quality grant F32 HS00148 and a research training grant from the Society for Academic Emergency Medicine.
We thank K. John McConnell, PhD, for assistance and insight with the calculation of interaction effects.
FOOTNOTES
Accepted Sep 20, 2004.
No conflict of interest declared.
REFERENCES
National Highway Traffic Safety Administration. Fifth/Sixth Report to Congress: Effectiveness of Occupant Protection Systems and Their Use. Washington, DC: US Department of Transportation; 2001
National Highway Traffic Safety Administration. Federal Motor Vehicle Safety Standards: Occupant Crash Protection. 49 CFR Part 571 [Docket 74-14, Notice 103, Final Rule] RIN 2127-AG14. Washington, DC: US Department of Transportation; 1997
National Highway Traffic Safety Administration. Traffic Safety Facts: Occupant Protection. Report DOT HS 809 474. Washington, DC: US Department of Transportation; 2001
National Highway Traffic Safety Administration. Traffic Safety Facts: Children. Report DOT HS 809 471. Washington, DC: US Department of Transportation; 2001
National Highway Traffic Safety Administration. Research Note: Revised Estimates of Child Restraint Effectiveness. Washington, DC: US Department of Transportation; 1996
Osberg JS, Di Scala C. Morbidity among pediatric motor vehicle crash victims: the effectiveness of seat belts. Am J Public Health. 1992;82 :422 –425
Decker MD, Dewey MJ, Hutcheson RH, Schaffner W. The use and efficacy of child restraint devices. JAMA. 1984;252 :2571 –2575
Johnston C, Rivara FP, Soderberg R. Children in car crashes: analysis of data for injury and use of restraints. Pediatrics. 1994;93 :960 –965
Narayan KM, Ruta D, Beattie T. Seat restraint use, previous driving history, and non-fatal injury: quantifying the risks. Arch Dis Child. 1997;77 :335 –338
Berg MD, Cook L, Corneli HM, Vernon DD, Dean JM. Effect of seating position and restraint use on injuries to children in motor vehicle crashes. Pediatrics. 2000;105 :831 –835
Petridou E, Skalkidou A, Lescohier I, Trichopoulos D. Car restraints and seating position for prevention of motor vehicle injuries in Greece. Arch Dis Child. 1998;78 :335 –339
Braver ER, Whitfield R, Ferguson SA. Seating positions and children's risk of dying in motor vehicle crashes. Inj Prev. 1998;4 :181 –187
Braver ER, Whitfield R, Ferguson SA. Risk of death among child passengers in front and rear seating positions. In: Child Occupant Protection 2nd Symposium. Report SAE 973298. Warrendale, PA: Society of Automotive Engineers; 1997:25 –34
Kahane CJ. Fatality Reduction by Air Bags: Analyses of Accident Data Through Early 1996. Report DOT HS 808 470. Washington, DC: US Department of Transportation; 1996
Braver ER, Ferguson SA, Greene MA, Lund AK. Reductions in deaths in frontal crashes among right front passengers in vehicles equipped with passenger air bags. JAMA. 1997;278 :1437 –1439
Glass RJ, Segui-Gomez M, Graham JD. Child passenger safety: decisions about seat location, air bag exposure, and restraint use. Risk Anal. 2000;20 :521 –527
Cummings P, Koepsell TD, Rivara FP, McKnight B, Mack C. Air bags and passenger fatality according to passenger age and restraint use. Epidemiology. 2002;13 :525 –532
Durbin DR, Kallan M, Elliot M, Cornejo RA, Arbogast KB, Winston FK. Risk of injury to restrained children from passenger air bags. Traffic Inj Prev. 2003;4 :58 –63
National Highway Traffic Safety Administration, National Center for Statistics and Analysis. National Automotive Sampling System (NASS) Crashworthiness Data System Analytic User's Manual, 2001 File. Washington, DC: US Department of Transportation; 2001
Corneli HM, Cook LJ, Dean JM. Adults and children in severe motor vehicle crashes: a matched-pairs study. Ann Emerg Med. 2000;36 :340 –345
Robertson LS. Reduced fatalities related to rear seat shoulder belts. Inj Prev. 1999;5 :62 –64
Association for the Advancement of Automotive Medicine. The Abbreviated Injury Scale, 1990 Revision, Update 98. Barrington, IL: Association for the Advancement of Automotive Medicine; 2001
Cummings P, Weiss NS. Mortality reduction with air bag and seat belt use in head-on passenger car collisions [letter]. Am J Epidemiol. 2001;154 :387 –389
Kahane CJ. Vehicle Weight, Fatality Risk and Crash Compatibility of Model Year 1991–99 Passenger Cars and Light Trucks. Report DOT HS 809 662. Washington, DC: US Department of Transportation; 2003
National Highway Traffic Safety Administration, National Center for Statistics and Analysis, Crash Investigation Division. 1988–1996 NASS CDS Variable-Attribute Structure Manual. Washington, DC: US Department of Transportation; 1998
Little RJA, Rubin DB. Statistical Analysis With Missing Data. New York, NY: John Wiley & Sons; 1987
Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med. 1991;10 :585 –598
Barnard J, Meng XL. Application of multiple imputation in medical studies: from AIDS to NHANES. Stat Methods Med Res. 1999;8 :17 –36
Schafer JL. Analysis of Incomplete Multivariate Data. New York, NY: Chapman & Hall; 1997
Allison PD. Missing Data. Sage University Papers Series: Quantitative Applications in the Social Sciences, Publication 07-136. Thousand Oaks, CA: Sage; 2001
Lee ES, Forthofer RN, Lorimor RJ. Analyzing Complex Survey Data. Sage University Papers Series: Quantitative Applications in the Social Sciences, Publication 07-071. Thousand Oaks, CA: Sage; 1989
Raghunathan TE, Lepkowski J, Van Hoewyk J, Solenberger PW. A multivariate technique for multiply imputing missing values using a sequence of regression models. Surv Methodol. 2001;27 :85 –95
Reiter JP, Raghunathan TE. Multiple Imputation for Missing Data in Surveys With Complex Sample Designs. Ann Arbor, MI: Institute for Social Research, University of Michigan; 2002
Jaccard J. Interaction Effects in Logistic Regression. Sage University Papers Series: Quantitative Applications in the Social Sciences, Publication 07-135. Thousand Oaks, CA: Sage; 2001
Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd ed. New York, NY: John Wiley & Sons; 2000
Ai C, Norton ED. Interaction terms in logit and probit models. Econ Lett. 2003;80 :123 –129
Kish L, Frankel M. Inference from complex samples (with discussion). J R Stat Soc. 1974;36 :1 –37
Rogol AD, Roemmich JN, Clark PA. Growth at puberty. J Adolesc Health. 2002;31 :192 –200
Saggese G, Baroncelli GI, Bertelloni S. Puberty and bone development. Best Pract Res Clin Endocrinol Metab. 2002;16 :53 –64
Damsgaard R, Bencke J, Matthiesen G, Petersen JH, Muller J. Body proportions, body composition and pubertal development of children n competitive sports. Scand J Med Sci Sports. 2001;11 :54 –60
National Safety Council. Crisis to Progress: Five Years of Air Bag Safety in America. Itasca, IL: National Safety Council; 2001
Grisoni ER, Pillai SB, Volsko TA, et al. Pediatric airbag injuries: the Ohio experience. J Pediatr Surg. 2000;35 :160 –163
Winston FK, Arbogast KB, Lee LA, Menon RA. Computer crash simulations in the development of child occupant safety policies. Arch Pediatr Adolesc Med. 2000;154 :276 –280(Craig D. Newgard, MD, MPH)
David Geffen School of Medicine, University of California, Los Angeles, California
Department of Emergency Medicine, Harbor-University of California, Los Angeles, Medical Center, Torrance, California
Los Angeles Biomedical Research Institute, Torrance, California
ABSTRACT
Background. Current recommendations regarding children traveling in passenger vehicles equipped with passenger air bags are based, in part, on evidence that the air-bag–related risk of injury and death is higher for children 12 years of age. However, the age or body size required to allow a child to be seated safely in front of a passenger air bag is unknown.
Objective. To evaluate specific cutoff points for age, height, and weight as effect modifiers of the association between the presence of a passenger air bag and serious injury among children involved in motor vehicle crashes (MVCs), while controlling for important crash factors.
Design. A national population-based cohort of children involved in MVCs and included in the National Automotive Sampling System (NASS) Crashworthiness Data System (CDS) database from 1995 to 2002 was studied. NASS CDS clusters, strata, and weights were included in all analyses.
Subjects. Children 0 to 18 years of age involved in MVCs and seated in the right front passenger seat.
Main Outcome Measure. Serious injury, defined as an Abbreviated Injury Scale score of 3 for any body region.
Results. A total of 3790 patients (1 month to 18 years of age) were represented in the NASS CDS database during the 8-year period. Sixty children (1.6%) were seriously injured (Abbreviated Injury Scale score of 3). Among age, height, and weight, age of 0 to 14 years (versus 15–18 years) was the only consistent effect modifier of the association between air-bag presence (or air-bag deployment) and serious injury, particularly for crashes with a moderate probability of injury. In analyses stratified according to age and adjusted for important crash factors, children 0 to 14 years of age involved in frontal collisions seemed to be at increased risk of serious injury from air-bag presence (odds ratio [OR]: 2.66; 95% confidence interval [CI]: 0.23–30.9) and deployment (OR: 6.13; 95% CI: 0.30–126), although these values did not reach statistical significance. Among children 15 to 18 years of age involved in frontal collisions, there was a protective effect on injury from both air-bag presence (OR: 0.19; 95% CI: 0.05–0.75) and deployment (OR: 0.31; 95% CI: 0.09–0.99). These findings persisted in analyses involving all collision types. We did not identify similar cutoff points for height or weight.
Conclusions. Children up to 14 years of age may be at risk for serious preventable injury when seated in front of a passenger air bag, and children 15 to 18 years of age seem to experience protective effects of air-bag presence and deployment. Age may be a better marker than height or weight for risk assessment regarding children and air bags.
Key Words: air bag motor vehicle crash children age height weight
Abbreviations: NASS, National Automotive Sampling System NHTSA, National Highway Traffic Safety Administration MVC, motor vehicle crash CDS, Crashworthiness Data System AIS, Abbreviated Injury Scale OR, odds ratio CI, confidence interval
In 1995, the National Highway Traffic Safety Administration (NHTSA) issued a press release with a strong warning that air bags may pose significant risks of injury and death for children.1 The warning was prompted by the results of an investigation of crashes in which deployment of the passenger-side air bag resulted in critical or fatal injuries to children seated in the right front seat.1 The following year, NHTSA published a final rule requiring warning labels on all new production passenger vehicles beginning in 1997.2 In addition to several well-supported recommendations regarding seat position and restraint use, the warning label stated, "Children 12 and under can be killed by the air bag."2 Similar labeling requirements were instituted for child safety seats.2
Evidence supporting the recommendations for age-appropriate restraints3–11 and rear seat locations10–13 for children traveling in motor vehicles has been well established. However, evidence for the age-based recommendations regarding air bags and children is lacking. Although the NHTSA guidelines for children and air bags have been circulated widely, the age (or body size) that best defines when a child's additional risk of injury or death from an air bag is replaced by benefit is unknown. Two initial studies suggested an association between air bags and childhood fatalities (defining children as 0–12 years and 0–9 years of age), but both analyses were limited by relatively small numbers of fatalities, yielding statistically inconclusive results.14,15 Subsequent studies defined children as 12 years of age and provided more conclusive evidence for the association between air bags and death among right front seat passengers within this age range.16–18 Kahane14 suggested that the increased mortality risk from air bags was most pronounced among right front seat passengers through age 10, becoming less pronounced from 11 to 14 years and turning to a net benefit for passengers 15 years of age, but the results did not reach statistical significance. A more recent study examining differences in air-bag effectiveness according to age and restraint use suggested a net increase in the risk of death among child passengers 12 years of age, but sample size limitations prevented definitive statements regarding the childhood age at which the net risk might change to no effect (or benefit) and the study did not assess markers of body size.18 Durbin et al19 demonstrated that the risk of injury (rather than death) among restrained children 3 to 15 years of age exposed to passenger air-bag deployment was twice that among front-seated children not exposed to an air bag, with the risk of injury being relatively constant among children 3 to 8, 9 to 12, and 13 to 15 years of age.
We hypothesized that specific cutoff points in age, height, and/or weight among children could be used to define when the risk of serious injury from the presence of a passenger air-bag changes from harmful to no effect (or beneficial), after adjustment for crash severity and other important crash factors. We tested this hypothesis with age, height, and weight as effect modifiers (interaction terms) of the association between the presence of a passenger air bag and serious injury among right front seat passengers 0 to 18 years of age involved in motor vehicle crashes (MVCs).
METHODS
Study Design and Setting
Subjects included in the National Automotive Sampling System (NASS) Crashworthiness Data System (CDS) database from 1995 through 2002 were used in the study. The NASS CDS is a probability-sampled, population-based, nationally representative cohort of persons involved in MVCs. Data in the NASS CDS database were collected through 3-stage sampling of MVCs from specific regions throughout the United States,20 to ensure generalizability of the data to the nation as a whole. The 8-year time period was selected because it encompasses the largest, most current block of data with comprehensive air-bag information in the NASS CDS database and because data collection changes for air bags were instituted in 1995. During this time period, increased production and eventual mandatory inclusion of both driver and passenger air bags in passenger vehicles occurred. By 1995, 98% of new cars sold were equipped with driver and passenger air bags.1 The study was approved by the institutional review boards of Oregon Health and Science University and the Los Angeles Biomedical Research Institute.
Subjects
A national, population-based sample of children 0 to 18 years of age who were seated in the right front seat and involved in MVCs with passenger vehicles or light trucks, as included in the NASS CDS database, was included in the analysis. The study included only children because previous studies have demonstrated a higher risk of injury or death among adults, compared with children, when controlling for other important crash characteristics.21,22 We defined "children" as persons 0 to 18 years of age because this age range allowed for adequate comparison samples for many age, height, and weight cutoff points while minimizing factors such as alcohol and drug use between groups and because US growth charts encompass this full age range.
Main Outcome Measure
Serious injury, the primary outcome, was defined as an Abbreviated Injury Scale (AIS) score of 3 for any body region. The AIS for a given injury ranges from 1 (minor) to 6 (nonsurvivable). A score of 3 represents a serious injury.23
Patient and Vehicle Characteristics
The exposure variable of interest was the presence of a passenger-side front air bag. Previous studies have defined air-bag exposure as air-bag presence rather than deployment, to account for variability in deployment and effects after activation.14–18 This definition has been suggested as a means of assessing simultaneously the effect of air-bag activation and the effect of the air bag after activation.24 There were only 13 cases (<1%) during this time period in which the air bag had been disconnected or removed from the vehicle. In these instances, the air-bag variable was coded as not present because of an inability to deploy.
In addition to air-bag information, variables were selected on the basis of known associations with injury and crash severity, as well as age and anthropometric measures for children. Thirteen variables were considered in the analysis: air-bag status (described above), age (years), height (centimeters), weight (kilograms), restraint use (manual lap or lap and shoulder belt, automatic belt system, or child safety seat with belt use), direction of impact (frontal, left lateral, or right lateral, with a reference group of rear, undercarriage, and top crashes), rollover with collision, entrapment, steering wheel deformity (in increments of 1 cm), vehicle model year (1998 versus earlier model years), vehicle weight,25 change in velocity, and passenger space intrusion (<15, 15–29, 30–45, 46–60, or >60 cm).26 We used change in velocity, passenger space intrusion, steering wheel deformity, rollover, and entrapment to adjust for crash severity, the main confounding factor in the analysis. Change in velocity, passenger space intrusion, vehicle weight, and steering wheel deformity were coded as continuous variables, whereas the remaining nonanthropometric covariates were coded as dichotomous categorical variables. In addition to the dichotomous interaction terms (described below), age, height, and weight were entered into the analysis as continuous variables, to test for a linear association with injury. Because some automotive manufacturers began implementing second-generation air bags in 1998,1 we included a dichotomous covariate for model year (1998 versus earlier model years) as a proxy for vehicles equipped with depowered air bags.
Age (range: 0–18 years), height (range: 30–203 cm), and weight (range: 2–150 kg) were coded as dichotomous categorical variables for use in interaction terms, with multiple different cutoff points. The dichotomous terms (ie, cutoff points) considered were as follows: 8 to 17 years for age, in increments of 1 year (eg, 0–8 vs 9–18 years); 110 to 180 cm for height, in increments of 10 cm (eg, 30–110 vs 111–203 cm); 30 to 75 kg for weight, in increments of 5 kg (eg, 2–30 vs 31–150 kg). Potential cutoff points were selected on the basis of results from previous studies14,18,19 and standard US male and female growth charts for children,27,28 as well as to represent a wide range of potential strata for age and body size. In total, 27 different potential cutoff points were tested.
Statistical Analyses
All analyses were restricted to children seated in the right front passenger seat to isolate the effects of air-bag presence and to control for the differential risk of injury according to seat position. Noncollision crashes and pure rollovers were excluded to allow for the appropriate use of changes in velocity to control for crash severity and because noncollision crashes are unlikely to result in air-bag deployment. After exclusion of noncollision crashes (n = 1019), pure rollovers (n = 313), pregnant female subjects (n = 35), and subjects with missing air-bag information (n = 5), the 1995–2002 NASS CDS database contained 3790 subjects 0 to 18 years of age. All children 0 to 18 years of age, regardless of body size, were included in the analysis. Although we note the total unadjusted number of children included in the analysis (n = 3790), we refer to the NASS CDS sample-adjusted number and proportion of occupants throughout the article, rather than the unadjusted numbers or national estimates.
To allow inclusion of all pediatric subjects contained in the NASS CDS database during this time period and to preserve the original weighting scheme of the NASS CDS database, we used multiple imputation to impute missing values.29–35 We imputed missing values by using parallel chains of multiple imputation, split on the presence or absence of a passenger air bag, to maximize the statistical efficiency of assessing interaction terms (age, height, and weight) with the air-bag variable.33 Variables for cluster, strata within clusters, and year were included as fixed effects in the imputation models to preserve the complex sampling design features of NASS CDS.34,36
Age-, height-, and weight-dependent cutoff points were investigated through interaction terms with the presence of a passenger air bag in multivariate logistic regression models.37,38 These models accounted for the complex sampling design of NASS CDS and included the additional variables noted above. We considered passenger air-bag presence the focal independent variable in the interaction term, with the age, height, or weight term as an effect modifier.37 Analyses were first performed on the subset of crashes that involved primary or secondary frontal impacts (defined as a primary or secondary direction of force in the 10 o'clock to 2 o'clock position of the vehicle), because these collisions are most likely to cause air-bag deployment and to confer the most potential benefit from air bags for adult passengers.1 Subsequent analyses were conducted with all crash types and included a covariate for direction of impact (ie, frontal, left lateral, or right lateral).
Results for the interaction terms are presented as P values because of the difficulty of interpreting the odds ratio (OR) of an interaction term. For terms that suggested the presence of an interaction (P < .10), we performed stratified analyses to assess the role of such effect modifiers and to quantify these associations in terms that would be clinically meaningful. In addition, we calculated the cross-derivative (cross-difference) of the interaction effect for interaction terms identified above and plotted this value against the probability of serious injury, as calculated with the multivariate model.39 As a qualitative check of the consistency of these results, all methods described above were repeated with the use of passenger air-bag deployment instead of passenger air-bag presence. To assess modification of the air-bag effect by restraint use and gender, we tested additional interaction terms with these covariates. In this article, we use the term risk synonymously with odds because the outcome (serious injury) was rare (ie, <2% of the sample). To explore the effects of multiple imputation and different potential underlying patterns of missing data on the results, we conducted sensitivity analyses of the imputation models with different patterns of missing data for variables with >5% missing data in a hypothetical data set with identical sample size and variables.
Database management was performed with SAS version 8.1 software (SAS Institute, Cary, NC). SAS-callable IVEware (survey methodology program; Survey Research Center, Institute for Social Research, University of Michigan, Ann Arbor, MI) was used for multiple imputation and multivariate analyses to account for the complex sampling design of the NASS CDS database and to ensure appropriate variance calculations.40 Stata software (version 8; Stata Corp, College Station, TX) was used to calculate the cross-derivative of the interaction effect.
RESULTS
We studied 3790 right front seat child occupants (age range: 1 month to 18 years), representing a national population of 2055390 children involved in MVCs during the 8-year period. A total of 2535 children (67%) were involved in primary or secondary frontal collisions. No children were excluded because of missing outcome (AIS score) information. Sixty children (1.6%) had at least 1 injury with an AIS score of 3. There were 10 fatalities (17% of seriously injured children) in the outcome group. Characteristics for right front seat occupants, according to air-bag presence versus absence, are listed in Table 1.
In multivariate regression models for children seated in the right front seat and involved in frontal collisions, age cutoff points of 14 years (interaction term: P = .054) and 15 years (interaction term: P = .074) were identified as potential effect modifiers of the association between air-bag presence and serious injury. When the same interaction terms were included in models assessing air-bag deployment (rather than air-bag presence), the results were more pronounced (interaction term for age 14 years: P = .019; interaction term for age 15 years: P = .024). These cutoff points persisted in models that included all types of collisions but had less statistical contribution. There were no dichotomous cutoff points in height or weight that modified the association between an air bag and serious injury, regardless of collision type and whether the air-bag variable was coded as presence or deployment (all P > .20). In addition, there was no linear relationship between height or weight and injury in any of the models (all P > .20).
In analyses stratified according to the age cutoff points identified above, an age split at 0 to 14 vs 15 to 18 years provided the most consistent risk strata in all models. Among children 0 to 14 years of age involved in frontal collisions, point estimates suggested an increased risk of serious injury from air-bag presence (OR: 2.66; 95% confidence interval [CI]: 0.23–30.9) and deployment (OR: 6.13; 95% CI: 0.30–126), although these values did not reach statistical significance. Among children 15 to 18 years of age involved in frontal collisions, there were protective effects on injury from both air-bag presence (OR: 0.19; 95% CI: 0.05–0.75) and deployment (OR: 0.31; 95% CI: 0.09–0.99). These findings persisted in stratified analyses involving all collision types (Table 2).
The interaction effect (cross-derivative) between air-bag presence and age of 0 to 14 years (versus 15–18 years) was positive for all probabilities of serious injury, although there was a more pronounced statistical contribution in crashes with a moderate probability of serious injury (Fig 1). For children with a very high or very low probability of injury, the interaction effect was negligible. We found similar but more pronounced results when the air-bag variable was defined as air-bag deployment (data not shown).
The use of restraints and gender did not modify the association between air-bag presence (or deployment) and injury (for interaction terms in the full and stratified models, all P > 0.30). There were no qualitative differences or indications of systematic bias in the results of sensitivity analyses with different patterns of missing data.
NASS CDS national estimates for the percentage of children seated in the right front seat demonstrated that the majority of children 13 to 14 years of age (67% in 2002) sit in the front passenger seat, regardless of the presence of an air bag. With the same national estimates, there has been a substantial decrease in the number of children 0 to 12 years of age seated in the right front seat since the introduction of air-bag warning labels in 1997 (15% in 2002). However, the number of children >12 years of age seated in the right front seat has not changed significantly since 1997, even when 13- to 14-year-old children are considered separately (data not shown).
DISCUSSION
These results suggest that, among occupants 0 to 18 years of age seated in the right front seat of a vehicle equipped with a passenger air bag, children 14 years of age have the greatest odds of serious injury, particularly in crashes with a moderate probability of injury. However, both air-bag presence and deployment seem to have a protective effect on injury among older children (15–18 years). This age cutoff point (ie, 0–14 years versus 15–18 years) is consistent with results from a previous study that assessed the age at which fatality risk from air bags changed from harmful to net beneficial14 and a study that assessed the risk of injury among 3- to 15-year-old restrained children.19 Although the effect of passenger air-bag presence on the risk of serious injury seemed to be influenced by age, we were unable to identify any effect modification with child height or weight.
Child developmental changes associated with puberty may provide one explanation for these findings. Although there is variation in the timing and duration of puberty among children, the onset of puberty typically occurs at 11 years of age among girls and 13 years among boys.41 There are increases not only in weight and height but also in body composition, including lean body mass, bone mineral content, and bone density (bone mass), during puberty.41,42 In addition, age has been shown to be a better determinant of pubertal development than anthropometric measurements.43 If changes in body composition and bone mass during early puberty play a role in susceptibility to injury from air bags in MVCs, it may not be surprising to demonstrate age as a better discriminator for risk of injury than height or weight. Despite the differences in onset of puberty between boys and girls, we did not find gender to be a significant effect modifier.
The age cutoff point identified in this analysis is older than that suggested by several previous studies15–18 and the current recommendations by NHTSA.2 Although the broad application and distribution of NHTSA guidelines regarding children traveling in vehicles with passenger air bags have been credited with reducing the number of air-bag-associated deaths among children,44 the results presented here suggest that there may be a significant proportion of at-risk children being missed by the current recommendations.
Restraint use has been suggested as a modifier of the association between air bags and death among children,16,18 with one study demonstrating that certain groups of younger restrained children (ie, 9–12-year-old children) actually have a survival benefit associated with the air bag.16 Our results failed to demonstrate such effect modification by restraint use. However, our finding that younger children were at risk of injury from an air bag despite restraint use is consistent with several other studies.17–19,45,46
There are several potential limitations in our analysis. The NASS CDS data set provided a relatively limited number of children who were seated in the right front seat and seriously injured in a vehicle equipped with a passenger air bag. Attempts to use interaction terms in multivariate models to demonstrate effect modification generally require large sample sizes to avoid type II error. It is possible that this sample and the number of subjects with serious injury were not large enough to demonstrate the effect modification by height and weight, and we cannot exclude the possibility that a larger sample of such children might suggest height or weight cutoff points for children and air-bag risk.
Our inability to find specific cutoff points in height or weight might have several additional explanations, ie, (1) multiple imputation was required for a portion of missing height and weight data, (2) values for height and weight in the NASS CDS data set might not have been accurate for all children, (3) the relationship between height or weight and injury might be more complex than can be tested with a dichotomous cutoff point, and (4) the risk strata for body size might be better estimated with a combination of height and weight values, which we did not assess. Of these possibilities, we do not suspect that the multiple imputation process adversely affected our ability to detect height or weight cutoff points because we performed sensitivity analyses for all variables with >5% missing data (including height and weight) with 3 different patterns of missing data (missing completely at random, missing at random, and missing in a nonrandom pattern). The nonrandom pattern of missingness simulated a scenario in which missing values for height and weight were associated with less serious crashes and less severe injuries. There was no indication of any qualitative change or systematic bias in the results secondary to multiple imputation, and our results were robust with all patterns of missing data.
There is also the possibility that the methods we used to identify cutoff points for age, height, and weight (ie, the use of dichotomous interaction terms) might not identify the ideal risk strata. Modeling age as a continuous (rather than dichotomous) variable in the interaction term produced similar point estimates for when the odds of serious injury began to change; however, small numbers of children in each age group reduced the precision of estimates and created difficulty in drawing firm conclusions.
We attempted to adjust for different types of air bags (eg, depowered air bags) by using vehicles with model year of 1998 versus earlier models as a proxy for vehicles equipped with depowered air bags. This variable represents not only the effect of newer-generation air bags in certain vehicles starting in 19981 but also that of other safety features included in newer-model vehicles. When we stratified the sample according to model year of 1998 versus older vehicles and assessed the same interaction terms for age, height, and weight, age of 14 years was the only cutoff point that persisted for both newer and older vehicles, but results were limited by the small number of newer-model vehicles.
CONCLUSIONS
An age of 0 to 14 years (versus 15–18 years) modified the effect of passenger air-bag presence on serious injury among children seated in the right front seat and involved in MVCs. Children 0 to 14 years of age seem to be at increased risk of injury from a passenger air bag, particularly in crashes with a moderate probability of injury, whereas the presence of an air bag has a protective effect on serious injury among children 15 to 18 years of age. We were unable to identify similar cutoff points for height or weight. The association with childhood developmental changes (measured by age) should be considered in the ongoing effort to reduce unnecessary air-bag–related injuries and deaths among children, because age may be a better marker than height or weight for risk assessment for children and air bags. On the basis of these results, children 0 to 14 years of age should not be seated in the right front passenger seat of vehicles equipped with a passenger air bag.
ACKNOWLEDGMENTS
This project was supported by Agency for Healthcare Research and Quality grant F32 HS00148 and a research training grant from the Society for Academic Emergency Medicine.
We thank K. John McConnell, PhD, for assistance and insight with the calculation of interaction effects.
FOOTNOTES
Accepted Sep 20, 2004.
No conflict of interest declared.
REFERENCES
National Highway Traffic Safety Administration. Fifth/Sixth Report to Congress: Effectiveness of Occupant Protection Systems and Their Use. Washington, DC: US Department of Transportation; 2001
National Highway Traffic Safety Administration. Federal Motor Vehicle Safety Standards: Occupant Crash Protection. 49 CFR Part 571 [Docket 74-14, Notice 103, Final Rule] RIN 2127-AG14. Washington, DC: US Department of Transportation; 1997
National Highway Traffic Safety Administration. Traffic Safety Facts: Occupant Protection. Report DOT HS 809 474. Washington, DC: US Department of Transportation; 2001
National Highway Traffic Safety Administration. Traffic Safety Facts: Children. Report DOT HS 809 471. Washington, DC: US Department of Transportation; 2001
National Highway Traffic Safety Administration. Research Note: Revised Estimates of Child Restraint Effectiveness. Washington, DC: US Department of Transportation; 1996
Osberg JS, Di Scala C. Morbidity among pediatric motor vehicle crash victims: the effectiveness of seat belts. Am J Public Health. 1992;82 :422 –425
Decker MD, Dewey MJ, Hutcheson RH, Schaffner W. The use and efficacy of child restraint devices. JAMA. 1984;252 :2571 –2575
Johnston C, Rivara FP, Soderberg R. Children in car crashes: analysis of data for injury and use of restraints. Pediatrics. 1994;93 :960 –965
Narayan KM, Ruta D, Beattie T. Seat restraint use, previous driving history, and non-fatal injury: quantifying the risks. Arch Dis Child. 1997;77 :335 –338
Berg MD, Cook L, Corneli HM, Vernon DD, Dean JM. Effect of seating position and restraint use on injuries to children in motor vehicle crashes. Pediatrics. 2000;105 :831 –835
Petridou E, Skalkidou A, Lescohier I, Trichopoulos D. Car restraints and seating position for prevention of motor vehicle injuries in Greece. Arch Dis Child. 1998;78 :335 –339
Braver ER, Whitfield R, Ferguson SA. Seating positions and children's risk of dying in motor vehicle crashes. Inj Prev. 1998;4 :181 –187
Braver ER, Whitfield R, Ferguson SA. Risk of death among child passengers in front and rear seating positions. In: Child Occupant Protection 2nd Symposium. Report SAE 973298. Warrendale, PA: Society of Automotive Engineers; 1997:25 –34
Kahane CJ. Fatality Reduction by Air Bags: Analyses of Accident Data Through Early 1996. Report DOT HS 808 470. Washington, DC: US Department of Transportation; 1996
Braver ER, Ferguson SA, Greene MA, Lund AK. Reductions in deaths in frontal crashes among right front passengers in vehicles equipped with passenger air bags. JAMA. 1997;278 :1437 –1439
Glass RJ, Segui-Gomez M, Graham JD. Child passenger safety: decisions about seat location, air bag exposure, and restraint use. Risk Anal. 2000;20 :521 –527
Cummings P, Koepsell TD, Rivara FP, McKnight B, Mack C. Air bags and passenger fatality according to passenger age and restraint use. Epidemiology. 2002;13 :525 –532
Durbin DR, Kallan M, Elliot M, Cornejo RA, Arbogast KB, Winston FK. Risk of injury to restrained children from passenger air bags. Traffic Inj Prev. 2003;4 :58 –63
National Highway Traffic Safety Administration, National Center for Statistics and Analysis. National Automotive Sampling System (NASS) Crashworthiness Data System Analytic User's Manual, 2001 File. Washington, DC: US Department of Transportation; 2001
Corneli HM, Cook LJ, Dean JM. Adults and children in severe motor vehicle crashes: a matched-pairs study. Ann Emerg Med. 2000;36 :340 –345
Robertson LS. Reduced fatalities related to rear seat shoulder belts. Inj Prev. 1999;5 :62 –64
Association for the Advancement of Automotive Medicine. The Abbreviated Injury Scale, 1990 Revision, Update 98. Barrington, IL: Association for the Advancement of Automotive Medicine; 2001
Cummings P, Weiss NS. Mortality reduction with air bag and seat belt use in head-on passenger car collisions [letter]. Am J Epidemiol. 2001;154 :387 –389
Kahane CJ. Vehicle Weight, Fatality Risk and Crash Compatibility of Model Year 1991–99 Passenger Cars and Light Trucks. Report DOT HS 809 662. Washington, DC: US Department of Transportation; 2003
National Highway Traffic Safety Administration, National Center for Statistics and Analysis, Crash Investigation Division. 1988–1996 NASS CDS Variable-Attribute Structure Manual. Washington, DC: US Department of Transportation; 1998
Little RJA, Rubin DB. Statistical Analysis With Missing Data. New York, NY: John Wiley & Sons; 1987
Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview and some applications. Stat Med. 1991;10 :585 –598
Barnard J, Meng XL. Application of multiple imputation in medical studies: from AIDS to NHANES. Stat Methods Med Res. 1999;8 :17 –36
Schafer JL. Analysis of Incomplete Multivariate Data. New York, NY: Chapman & Hall; 1997
Allison PD. Missing Data. Sage University Papers Series: Quantitative Applications in the Social Sciences, Publication 07-136. Thousand Oaks, CA: Sage; 2001
Lee ES, Forthofer RN, Lorimor RJ. Analyzing Complex Survey Data. Sage University Papers Series: Quantitative Applications in the Social Sciences, Publication 07-071. Thousand Oaks, CA: Sage; 1989
Raghunathan TE, Lepkowski J, Van Hoewyk J, Solenberger PW. A multivariate technique for multiply imputing missing values using a sequence of regression models. Surv Methodol. 2001;27 :85 –95
Reiter JP, Raghunathan TE. Multiple Imputation for Missing Data in Surveys With Complex Sample Designs. Ann Arbor, MI: Institute for Social Research, University of Michigan; 2002
Jaccard J. Interaction Effects in Logistic Regression. Sage University Papers Series: Quantitative Applications in the Social Sciences, Publication 07-135. Thousand Oaks, CA: Sage; 2001
Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd ed. New York, NY: John Wiley & Sons; 2000
Ai C, Norton ED. Interaction terms in logit and probit models. Econ Lett. 2003;80 :123 –129
Kish L, Frankel M. Inference from complex samples (with discussion). J R Stat Soc. 1974;36 :1 –37
Rogol AD, Roemmich JN, Clark PA. Growth at puberty. J Adolesc Health. 2002;31 :192 –200
Saggese G, Baroncelli GI, Bertelloni S. Puberty and bone development. Best Pract Res Clin Endocrinol Metab. 2002;16 :53 –64
Damsgaard R, Bencke J, Matthiesen G, Petersen JH, Muller J. Body proportions, body composition and pubertal development of children n competitive sports. Scand J Med Sci Sports. 2001;11 :54 –60
National Safety Council. Crisis to Progress: Five Years of Air Bag Safety in America. Itasca, IL: National Safety Council; 2001
Grisoni ER, Pillai SB, Volsko TA, et al. Pediatric airbag injuries: the Ohio experience. J Pediatr Surg. 2000;35 :160 –163
Winston FK, Arbogast KB, Lee LA, Menon RA. Computer crash simulations in the development of child occupant safety policies. Arch Pediatr Adolesc Med. 2000;154 :276 –280(Craig D. Newgard, MD, MPH)