PEDIATRICS Vol. 108 No. 2 August 2001, pp. 438-442

Predictors of Hazardous Child Seating Behavior in Fatal Motor Vehicle Crashes: 1990 to 1998

Eve Wittenberg, PhD, Sue J. Goldie, MD, MPH, and John D. Graham, PhD

From the Harvard Center for Risk Analysis, Harvard School of Public Health, Boston, Massachusetts.

	ABSTRACT

Top Abstract Methods Results Discussion Conclusion References

Objective.  Motor vehicle crashes are the leading cause of death in children ages 5 to 14. Children seated in the front seats of vehicles are at increased risk of death and injury in crashes, particularly in vehicles with passenger-side air bags. This study identifies factors associated with the seating of children in the front seats of vehicles involved in fatal crashes between 1990 and 1998.

Methods.  Using 1990 to 1998 data from the Fatal Analysis Reporting System, a US census of motor vehicle crashes involving a fatality, multivariable logistic regression was used to model the association between child seating behavior and vehicle, driver, and occupant characteristics.

Results.  The proportion of vehicles carrying children in the front declined from 42% to 31% over the 9-year period. Controlling for driver and vehicle characteristics, the risk of front-seating declined between 1990 and 1998, and this risk was smaller in vehicles carrying only younger children (6 years) than in those carrying older children. In the 3 years after the introduction of dual air bags into a significant proportion of the passenger fleet in late 1995, dual air bags were associated with fewer children being seated in the front seat. By the end of 1998, traveling in a vehicle with dual air bags and only children age 6 or younger was associated with a 95% lower chance of a child being seated in the front (odds ratio = 0.05; 95% confidence interval: 0.04-0.08). An important factor in safer seating position was the presence of multiple passengers, especially an older one, and children were at higher risk of front-seating when they traveled alone with the driver.

Conclusions.  The 1990s saw a decline in front-seating of children in vehicles involved in fatal crashes among all types of vehicles and drivers. Although this trend is encouraging, children ages 6 to 12 and children traveling alone with the driver remain at higher risk of being seated in the front. These traveling situations should be targeted for behavioral safety interventions to improve child motor vehicle safety.  Key words:  air bags, child seating, motor vehicles.

Motor vehicle crashes are the leading cause of death for children ages 5 to 14. In the United States, on average 7 children ages 0 to 14 are killed and 866 injured every day in motor vehicle crashes.¹ Child safety in motor vehicles is affected by many factors, including road conditions, vehicle design, driving behavior, restraint use, and seating position.^2-5 Seating position is a factor that takes no technology to correct and is primarily under the driver's control.

The first child death from a passenger air bag in 1993 occurred in a low-speed crash that otherwise would have been nonfatal to the child.⁶ Public alarm heightened as additional deaths and serious injuries occurred in subsequent similar accidents, triggering a variety of efforts to improve child seating behavior in motor vehicles. The Air Bag and Seat Belt Safety Campaign, a coalition of automobile manufacturers, physicians' organizations, government agencies, and parent groups, initiated an educational campaign in the mid-1990s to improve child safety through seating behavior and restraint use,⁷ and the US government mandated a safety sticker campaign in January 1996 to caution drivers against seating young children in front of a passenger air bag. Media attention followed these events with extensive coverage on television and in the press.^8,9

Research on seating position has shown the rear seat to be protective for children regardless of restraint use.^2-4,10 Observational studies and crash data show that approximately one-third of children are seated in the front seats of vehicles, or approximately one-third of vehicles carrying children have children seated in the front seats, including many vehicles equipped with passenger-side air bags.^10-15 Driver and passenger characteristics associated with child front-seating include single children traveling alone with the driver, no adult passengers in the vehicle, and older child age.^2-410-15 Although child safety seats generally are used for children under age 4, the majority of child safety seats are used incorrectly,¹⁶ proper restraint use for children aged 4 to 8 is rare,⁵ and a high proportion of older children ride unrestrained.¹⁷

The purpose of this study was to identify predictors of child seating position in fatal motor vehicle crashes in the United States from 1990 to 1998, the most comprehensive source of motor vehicle data including seating behavior. Our motivation was to provide information to guide efforts to improve child motor vehicle safety through behavior change.

	METHODS

Top Abstract Methods Results Discussion Conclusion References

Multivariable logistic regression was used to identify predictors of front-seating of children. The analysis used 9 years of data (1990-1998) from the Fatal Analysis Reporting System (FARS), a census of motor vehicle crashes in the United States involving at least 1 fatality that occurred within 30 days of the crash. FARS is the largest available US database on motor vehicle operation that includes occupant seating position and driver and vehicle characteristics. Other databases either are smaller, limiting the statistical power to identify predictors of seating behavior (eg, the National Automotive Sampling System/General Estimates System), or lack data elements critical to this analysis (eg, most observational data sets contain 1 time period, estimate child age, and lack vehicle air bag status). Inclusion criteria were vehicles containing at least 1 nondriving child (defined as under 13 years of age); passenger cars, sport utility vehicles (SUVs), and passenger vans (pickup trucks were excluded because they often have no rear seat); and available vehicle identification number (VIN). The VIN was used to determine the vehicle restraint system (ie, type of air bag).

Because decisions about child seating location depend on other occupants in the vehicle, the unit of analysis was each vehicle rather than each child.^2-410-15 Variables included in the analysis were vehicle type (passenger car, SUV, or van); model year of vehicle; state in which vehicle was registered; air bag status; number, age, and position of occupants; date and time of crash; and driver characteristics (age, sex, restraint use). Vehicles were coded as having a child in front if at least 1 child was seated in the front middle or front right seats. Date of crash was used to group vehicles into 3-month periods (calendar quarters), which was modeled as a linear trend variable. Indicator variables were created for driver belt use, sex, and vehicle type (van or SUV versus passenger car). The time of the crash was defined as daytime if between the hours of 6 AM and 10 PM and nighttime if between 10 PM and 6 AM (when children are more likely to be sleeping in vehicles). Driver and vehicle ages were continuous variables. The VIN was decoded using the VINDICATOR¹⁸ software package to create an indicator variable that identified vehicles with dual air bags.

Vehicles were categorized in 3 ways to reflect established trends in the frequency of child front seating.^2-410-15 Vehicles with a single child passenger were distinguished from those with >1 passenger, vehicles with only child passengers were distinguished from those with at least 1 adult or teenage passenger, and vehicles carrying exclusively younger children were distinguished from vehicles with at least 1 child over 6 years of age.

The dependent variable in the regression analysis was the presence of at least 1 child seated in the front right or middle positions of a vehicle. The model was built using a forward selection process,¹⁹ excluding any variables that did not meet a P  .05 criterion for inclusion. Pairwise interaction terms were chosen for inclusion based on relationships established in previous research and logical combinations and tested at the P  .05 level. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were calculated manually for all variables included in the final model, as were interacted odds ratios. Although FARS is a census, for this analysis the 9 years of data were considered a sample of vehicles involved in fatal crashes for all time periods, so confidence intervals were calculated for the infinite population of crashes that have occurred and will occur in the future. Goodness of fit of the final model was assessed using the residual score statistic (although it is noted that this test may overestimate the possibility of accepting the model fit because of the use of the forward selection process)²⁰; ² =1.60 with 3 degrees of freedom (P = .66), indicating that variables excluded from the model were not important in explaining variation.

Vehicles missing data on any of the variables included in the final model except restraint use were omitted from the analysis (2.3% of vehicles [654] were omitted; they did not significantly differ from rest). Because driver restraint use was missing for 22% of vehicles, it was coded as "unknown" if missing to preserve these vehicles in the model. This coding precluded the interpretation of the coefficient on driver restraint use. The effect of these missing data on parameter estimates was tested with interaction terms between driver restraint use and the independent variables. All analyses were conducted using SAS software version 6.12.²¹

	RESULTS

Top Abstract Methods Results Discussion Conclusion References

A total of 28 688 vehicles met the inclusion criteria for the time period studied. Overall, 39% of vehicles carried children in the front seat; this proportion decreased from 42% at the start of 1990 to 31% by the end of 1998. In 24% of the vehicles the driver and a single child were traveling alone; 56% of vehicles carried at least 1 teenage or adult passenger in addition to at least 1 child, and the rest carried only passengers under 13 years of age. In 47% of the vehicles the child passengers were all age 6 or younger. Twenty-eight percent of vehicles were >10 years old at the time of the crash (Table 1).

The proportion of children-carrying vehicles that were equipped with dual air bags increased from almost none in 1990-1993 to 29% by the end of 1998. The prevalence of children seated in the front varied with the presence of dual air bags and the age composition of children in the vehicle. Both with and without dual air bags, fewer vehicles carrying exclusively young children (age 6 or younger) had a child seated in the front than vehicles with at least 1 older child. And despite the rapid introduction of dual air bags into the passenger vehicle fleet from 1996 to 1998, the proportion of vehicles with dual air bags that carried at least 1 child seated in front did not increase during this period.

Driver and vehicle characteristics associated with front-seating of children included vehicle occupancy, including the number of passengers and ages of passengers and driver; date and time of the crash; vehicle type, age, and whether it had a passenger-side air bag; and driver sex and restraint use (Tables 2 and 3). Vehicles carrying only children age 6 and younger showed a different seating pattern than those with a mix of child ages or just older children. Controlling for other factors, the odds of a child being seated in the front declined over the time period, and this decline was larger both for vehicles carrying exclusively younger children and for those equipped with dual air bags (Table 4). Among vehicles without dual air bags, the odds of a child being seated in the front if all children were 6 years of age declined from 0.26 at the start of 1990 to 0.11 at the end of 1998 (OR = 0.26; 95% CI: 0.22-0.32; OR = 0.11; 95% CI: 0.08-0.15, respectively). If these vehicles were carrying at least 1 child >6 years, the decline in odds was from 0.99 to 0.76 over the same time period (OR = 0.99; 95% CI: 0.988-0.999; OR = 0.76; 95% CI: 0.64-0.89, respectively). Until early 1996, <10% of vehicles were equipped with dual air bags. In 1996 the odds of a child being seated in the front seat were 85% less for vehicles with dual air bags carrying only young children (OR = 0.15; 95% CI: 0.11-0.19). By the end of 1998, young children traveling in these types of vehicles were 95% less likely to be seated in the front seats (OR = 0.05; 95% CI: 0.04-0.08).

A child traveling in a vehicle with only the driver was more likely to be seated in the front seat than a child traveling in a vehicle with additional occupants besides the driver (OR for >2 occupants and all children 6 = 0.20; 95% CI: 0.16-0.24; OR for >2 occupants and at least 1 child >6 = 0.23; 95% CI: 0.20-0.27; Table 3). An adult or teenage passenger nearly always took precedence in the front seat over a child (OR = 0.09; 95% CI: 0.09-0.10; Table 2).

Children traveling in vans or SUVs were less likely to be seated in the front seats than children traveling in passenger cars (OR for vans and SUVs with all children 6 = 0.26; 95% CI: 0.21-0.32; OR for vans and SUVs with at least 1 child >6 = 0.78; 95% CI: 0.70-0.86; Table 3). One year of vehicle age added 3% chance of observing a child seated in the front seat (OR = 1.03; 95% CI: 1.02-1.04; Table 2). Children were less likely to be seated in the front seat in vehicles involved in crashes at night (OR = 0.86; 95% CI: 0.79-0.95; Table 2).

Men were less likely than women to seat children in the front seat (OR for male driver and all children 6 = 0.28; 95% CI: 0.23-0.34; OR for male driver and at least 1 child >6 = 0.83; 95% CI: 0.76-0.91; Table 3). Driver age had a small effect on seating behavior, with each decade of age adding 4% increased chance of a child traveling in the front (OR = 1.04 for 10 years of increased age; 95% CI: 1.01-1.07; Table 2).

	DISCUSSION

Top Abstract Methods Results Discussion Conclusion References

Controlling for other factors, our findings show that children were at less risk of being seated in front in 1998 than in 1990, and this was particularly true for vehicles carrying exclusively children under age 6 and vehicles equipped with dual air bags. Single children traveling alone with the driver were associated with an increased risk of front-seating, and the presence of a teenage or adult passenger in the vehicle dramatically reduced the risk of a child occupying the front seat. Our results imply that future safety efforts should be focused on children over age 6 and on driving situations in which the child is the only passenger.

Although this analysis identified factors associated with front-seating of children, it did not provide information about the reasons behind the observed behavior changes of the past decade. The decade has seen increased restraint use among children¹⁷ and increased attention to proper use of safety seats.¹⁶ The association between the presence of air bags in a vehicle and decreased chance of front-seating suggests that drivers have paid attention to the hazard air bags pose to children. However, front-seating of children has decreased among all vehicles and most rapidly in those carrying only young children, suggesting that the relative safety of the rear seat has been translated into seating behavior changes particularly among drivers of younger children.

Our finding that children traveling in older vehicles were at greater risk for front-seating paralleled others' finding that children traveling in older vehicles are less likely to be restrained.¹⁷ Of particular concern is that these older vehicles are equipped with fully powered, old-technology air bags, which are more dangerous to children than the newer, depowered air bags.³ Vehicle age may indicate vehicle models that lack rear seat belts, which could encourage front-seating for the use of belts or might indicate socioeconomic status (SES), which was not otherwise included in this dataset. If this is so, then people of lower SES may be more likely to seat children in the front. Understanding the association between SES and seating behavior is a priority for future research.

Although our findings are limited to fatal crashes because of our use of the FARS database, our results have implications for behavioral interventions for the general driving population. The use of FARS contains 2 possible biases. First, drivers involved in fatal accidents may be more risk-taking than other drivers, which might be associated with a greater likelihood of seating children in the front. Second, children seated in the front are at higher risk of fatal injury,^2,10 increasing their likelihood of being captured in FARS. Both trends would bias our estimates of front-seating upward, but it is unclear that they would misspecify predictors of child seating behavior. Our estimates of front-seating should thus be interpreted with caution because they may overestimate the true rate in the general driving population, but our conclusions on the predictors of seating behavior drawn from FARS data may be indicative of the general driving population. Nevertheless, our results mirror those found in the entire driving population, including vehicles involved in no crashes, less severe crashes, and fatal crashes,^4,12,14,15 suggesting that any bias present is small. Furthermore, although the vehicles included in FARS are distinctly different from vehicles overall, they are of particular concern when considering child safety precisely because of their involvement with fatalities. In addition, FARS may contain data errors that are particular to accident reporting, such as restraint use, because of the liability or insurance implications connected to reporting these data.

	CONCLUSION

Top Abstract Methods Results Discussion Conclusion References

Child motor vehicle seating behavior improved during the 1990s through decreased seating of children in the front seat. Nevertheless, a significant proportion of vehicles involved in fatal crashes still show children placed at greater safety risk by virtue of their seating choice.¹⁰ Opportunities to target behavioral interventions that can be used to improve the effectiveness and efficiency of safety resources have been identified in this article. Additional research is needed to track improvements in behavior over time and to monitor the impact of other safety efforts on behavior, including technological advances and legislative efforts. Pediatricians are in a position to influence the behavior of their patients and to contribute to motor vehicle safety improvements.

	ACKNOWLEDGMENT

This work was supported by the Harvard Center for Risk Analysis.

	FOOTNOTES

Received for publication Oct 9, 2000; accepted Mar 29, 2001.

Reprint requests to (E.W.) Dana Farber Cancer Institute, 44 Binney St, 454, S21-24, Boston, MA 02115. E-mail: eve_wittenberg{at}dfci.harvard.edu

	ABBREVIATIONS

FARS, Fatal Analysis Reporting System; SUV, sport utility vehicle; VIN, vehicle identification number; OR, odds ratio; CI, confidence interval; SES, socioeconomic status.

	REFERENCES

Top Abstract Methods Results Discussion Conclusion References

1. National Highway Traffic Safety Administration Traffic Safety Fact Sheet. Children, 1998. Washington, DC: NHTSA, Department of Transportation; 1998. DOT HS Publ. No. 808 951
2. Braver ER, Whitfield R, Ferguson SA Seating positions and children's risk of dying in motor vehicle crashes. Inj Prev 1998; 4:181-187 [Abstract/Free Full Text]
3. Graham JD, Goldie SJ, Segui-Gomez M, et al. Reducing risks to children in vehicles with passenger airbags. Pediatrics. 1998;102(1). URL: http://www.pediatrics.org/cgi/content/full/102/1/e3
4. 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 [Abstract/Free Full Text]
5. Winston FK, Durbin DR, Kallan MJ, Moll EK The danger of premature graduation to seat belts for young children. Pediatrics 2000; 105:1179-1183 [Abstract/Free Full Text]
6. National Highway Traffic Safety Administration. Special Crash Investigations Web site, June 2000 Report. Available at: http://www.nhtsa.dot.gov/people/ncsa/scireps.html. Accessed July 16, 2000
7. National Safety Council. Air Bag & Seat Belt Safety Campaign Web site. Available at: http://www.nsc.org/airbag.htm. Accessed July 16, 2000
8. Vanderbilt University. Television News Archive. Available at: http://tvnews.vanderbilt.edu/. Accessed July 16, 2000
9. Lexis-Nexis Academic Universe. News database. Accessed July 17, 2000
10. Glass R, Segui-Gomez M, Graham JD Child passenger safety: decisions about seating location, airbag exposure, and restraint use. Risk Anal 2000; 20:521-527 [CrossRef][Medline]
11. Glass R, Graham JD Kids at risk: where American children sit in passenger vehicles. J Saf Res 1999; 30:17-24
12. Wittenberg E, Nelson TF, Graham JD The effect of passenger airbags on child seating behavior in motor vehicles. Pediatrics 1999; 104:1247-1250 [Abstract/Free Full Text]
13. Edwards J, Sullivan K. Where Are All the Children Seated and When Are They Restrained? Warrendale, PA: Society of Automotive Engineers; 1997 SAE Technical Paper 971550
14. Segui-Gomez M, Glass R, Graham JD Where children sit in motor vehicles: a comparison of selected European and American cities. Inj Prev 1998; 4:98-102 [Abstract/Free Full Text]
15. Ferguson SA, Wells JK, Williams AF Child seating position and restraint use in three states. Inj Prev 2000; 6:24-28 [Abstract/Free Full Text]
16. Kohn M, Chausmer K, Flood MH Anticipatory guidance about child safety seat misuse. Lessons from safety seat "check-ups." Arch Pediatr Adolesc Med 2000; 154:606-609 [Abstract/Free Full Text]
17. Agran PF, Anderson CL, Winn DG. Factors associated with restraint use of children in fatal crashes. Pediatrics. 1998;102(3). URL: http://www.pediatrics.org/cgi/content/full/102/3/e39
18. Highway Loss Data Institute. VINDICATOR Software. Arlington, VA: Highway Loss Data Institute; 1997, 1999
19. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons; 1989
20. Stokes ME, Davis CS, Koch GG. Categorical Data Analysis Using the SAS System. Cary, NC: SAS Institute Inc; 1995
21. SAS Institute Inc. SAS, Version 6.12. Cary, NC: SAS Institute Inc; 1999