Results

Table 1 shows the characteristics of cases and controls and the univariate relations between potential risk factors and child RTA injury. Cases were significantly less likely to have been restrained (p < 0.001). Travel between noon and 6 am, journey length, the interviewee not being the driver of the car, driver having points on the licence, number of points on the licence, and driver being in full time employment were all significantly associated with child injury (p < 0.01 for each). Driver’s gender, marital status, age at completion of full time education, and type of housing were not associated with child injury.

Because of small sample sizes, data on cases and controls were pooled to assess associations of potential confounding variables with seat restraint, and these are presented in table 2. There were no significant interactions between being a case or control and any of the independent variables on the effect of seat restraint use. Therefore the pooled estimates are unlikely to have any inherent bias. Journey time, journey length, driver having points on the licence, and driver’s marital status and housing status were all associated with reported seat restraint use (p < 0.05 for each), but driver’s gender and employment status were not. Table 3 shows the results of multiple logistic regression analysis, where each odds ratio is controlled for all other factors in the table. Thus, controlled for potential confounding factors, a restrained child was 93% less likely to have suffered an injury than an unrestrained child. Journeys between 6 pm and 6 am were over 30 times more likely to be associated with child injury than journeys between 6 am and noon, journeys 10 miles or longer were nearly nine times more likely to be associated with child injury than journeys less than one mile in length. Child injury was also associated with the driver’s accident history—drivers with points on their licence were over five times more likely to have had an accident resulting in child injury than drivers without points, and the risk of child injury increased with the number of points on the license.

Table 4 shows the significant predictors identified from a stepwise logistic regression model to which the following variables were available: seat restraint, parent’s sex, marital status, employment, past history of accidents, points on licence, housing, education, journey time, and journey length. Seat restraint entered the model first as the strongest predictor of child injury. The driver’s accident history during the past five years and journey time were also significant predictors of child injury.

Discussion

Controlled for potential confounding from factors like length of journey, time of day, and accident proneness, reported seat restraint was associated with a 93% lower risk of child accident injury (95% confidence interval 79% to 98%). This case-control study thus supports the findings of our previous report from an observational study which did not account for several potential confounding variables.9 However, some methodological issues in the present study merit further consideration.

In order to obtain a valid estimate of the odds ratio in a case-control study, the cases must be incident and representative of the population of those with the disease, and the controls must be representative of the population of those without the disease. The cases were incident events, but exposure data were collected on only 40 of the 78 cases. Controls were identified through A&E outpatient clinics which took place between 9 am and 5 pm, and this may wholly or partly explain the strong association observed between injury and journey time. The association between injury and journey length may also be spurious because few children attending outpatient clinics in Grampian area will travel more than 10 miles. Because it was not feasible to obtain community based controls in this study, the controls were hospital based, although not selected from attenders to A&E involved in a road traffic accident, as in previous studies.8 9

While obtaining data on exposures like seat restraint use, there is always a fear that respondents may provide a socially acceptable rather than a truthful response. To try to overcome this problem, we attempted the random response technique. The method proved to be of little help because it has a large random error associated with its estimate of the proportion of parents reporting that their child was restrained. The sample size available to the study was therefore not sufficient to make meaningful interpretations. Furthermore, the estimate of prevalence of seat restraint use obtained by this technique was almost identical to that obtained by the direct method. This suggests that parents may report the correct answer on direct questioning and that information bias may not have been a problem, or that parents give socially acceptable answers even with the random response technique.

The association between the number of accidents incurred by the driver during the past five years, the number of points on the licence, and injury is an important new finding. In this study, a case was defined as a child sustaining an injury, with an MAIS score of > 0, while travelling in a motor vehicle. All cases will therefore by definition also have been exposed to a motor vehicle accident of some kind. Controls, on the other hand, were children travelling in a motor vehicle and not sustaining an injury. Although it is conceivable that a control child might have been involved in a motor vehicle accident and not have sustained an injury, in fact none of the control children was involved in a car accident. Thus the observed association indicates a relation between previous accident history and the risk of having an accident, and not the risk of sustaining an injury per se. However, this finding of a link between previous accidents and risk of childhood RTAs has important implications for the prevention of accident injury. It suggests that effective strategies aimed at reducing the risk of a further accident among drivers with a history of an accident in the last five years, or with accumulated licence endorsements, might have an additional impact on reducing childhood accident injuries.

The data from this study can be used to estimate the likely minimum and maximum effects (based on the confidence intervals of the odds ratios), in terms of percentage reduction in the incidence of non-fatal injury, of achieving new target levels of seat restraint use.9This kind of information on the marginal benefits of increased seat restraint usage—in terms of expected reductions in death and injury—combined with data on the estimated marginal costs of meeting seat restraint target levels can help policy makers and health service managers to decide on what extra resources should be allocated to this particular area of health promotion.