Subjects and methods

From 1 January to 30 November 1996, 19 190 children less than 15 years old with unintentional injuries were registered in the EDISS database from “A Kyriakou” Children’s Teaching Hospital in Athens. Of those, 129 children were car passengers less than 12 years old who sought hospital medical attention for an injury of variable severity after a motor vehicle accident. The injury severity score22 was up to 4 for 109 children, 5 to 9 for 16 children, and 10 or more for the remaining four children. Eleven of these children had fracture(s) and 12 had a head injury with at least some central nervous system involvement. Four health visitors associated with CERERPI interview every child and his or her guardian who seek medical attention for an injury on the basis of a precoded questionnaire. In addition to sociodemographic variables and injury characteristics, information on whether the child was seated in the front or rear seat and whether a child restraint system was used is also recorded for children injured in a road traffic accident as rear passengers.

In 1996 CEREPRI also undertook, in collaboration with the road traffic police department, a random inspection survey on the use of seat belts or child car restraints among occupants of passenger cars in the Athens area. During the survey, which lasted 40 days, teams of one CEREPRI interviewer and a road traffic policeman randomly stopped 1400 passenger cars, excluding taxis, in 10 sites on secondary roads, five sites on main road arteries in Athens, and five sites on highways linking the city with the rest of the country. The policeman stopped the car and explained to the driver that a study was being carried out in collaboration with the University of Athens and that no action would be taken in this instance against non-users of seat belts, even though their use is mandatory in Greece. Subsequently, the policeman withdrew and the interviewer inspected the availability and use of seat belts by the car’s occupants aged 12 years or older and the use of child restraint systems by younger children. In addition, information was recorded about basic demographic variables, the date, and the time of the day or night. There was no refusal to cooperate and the short inspection and interview ended with advice about the substantial health benefits imparted by the regular use of car restraints. A total of 191 children less than 12 years old were identified and for each of them gender, age, front or rear seating, and the use of a car restraint system at the time of the inspection was recorded (for children less than 5 years a restraint cot or safety seat; for children 5–11 years a seat belt in combination with booster).

The data were analysed as a case-control investigation considering the unselected sample of injured children as cases and children in the inspection survey of the underlying population as controls. Three variables were evaluated as relevant exposures: the non-use of a child restraint system (v use), front seating (vrear seating), and being less than 5 years old (v being at least 5 years old). The Mantel-Haenszel procedure allows for the control of the variables that are not focused upon during alternative comparisons. The method evaluates the statistical significance of each association after adjustment for confounding influences on the part of the other variables and generates the odds ratio (OR), which is an estimate of how more frequent the injury under study is when a certain categorical exposure is present rather than absent. Finally, the method provides the 95% confidence interval (CI), which is a measure of the plausible chance variation of the respective OR.

On the basis of the proportion of children exposed to each of the two main factors (front seating, non-use of a child restraint system) and the corresponding Mantel-Haenszel derived ORs, it is possible to calculate the population attributable fraction.23 This is the proportion of childhood injuries among car passengers of similar age in the study population of Greater Athens which would have been avoided if all the children were seated in the rear of the car or were using an appropriate restraint system (or both).

Results

Table 2 gives the frequency distribution of 129 injured children from the EDISS database and the 191 non-injured children from the inspection survey by age, seating position, and use of a child restraint system. The data from the inspection survey indicate that in the population of Athens most children are travelling without proper restraint systems. This is particularly true with respect to older children, for whom a booster seat must be used. Even among children less than 5 years old, however, a restraint system is used by less than one third (17/61). Of note, however, is that most children (96%) are travelling in the rear seat.

Unrestrained children are at significantly higher risk of sustaining a motor vehicle injury than restrained children, the OR being 3.3 (lower panel in table 2). This estimate is based on data for rear seated younger children because none of the older children used an appropriate restraint system that would allow proper comparison. Front seating increases the risk for injury fivefold. This OR mainly reflects the occurrence of injury among unrestrained children because the restrained children were all less than 5 years old and, among them, all but one were seated in the rear. As expected, the OR is very high when children are both unrestrained and seated in the front (OR 6.4), although this particular contrast is based on very few children in the younger age group. Finally, after adjustment for seating position and restraint use, younger children appear to be at 60% higher risk of injury than older children (p = 0.11).

Miettinen’s formula24 allows the calculation of the proportion of road traffic injuries among children that could have been avoided if they were all properly restrained or, alternatively, all of them seated in the rear, under the assumption that the relative effectiveness is uniform across all strata. To that effect, the proportion of exposed (unrestrained or, alternatively, front seated) children among the cases is multiplied by (OR−1) and divided by the OR. It appears that the proper restraint of all children could reduce the number of those injured by 67% and that the rear seating of all children could reduce this number by 16%.

Discussion

This study was designed and implemented as an analytical population based and population controlled epidemiological investigation. This characteristic distinguishes it from valuable previous investigations that were either ecological in nature (before/after group comparisons or group randomisation) or relied on the matched pairs design that does not involve a population based group of non-injured children.5 6 9-12 14 25 Other strengths of this investigation were the medical rather than police ascertainment of injuries, which assures better data quality,26 and the use of a control group that captures the exposure pattern of the actual population at risk. The inspection survey used in this study is likely to provide more valid data than those derived from observation surveys based on random intersection or shopping mall observations.27 Finally, the situation in Greece allows an efficient evaluation of the protection afforded by rear seating alone, because in this population unrestrained rear seating is common.

Among the weaknesses of this investigation are power limitations and the inability to address the relative safety of seating position (frontv rear) among restrained children because most of the restrained children were seated in the rear. Furthermore, the protection imparted by child restraint systems was evaluated only among children younger than 5 years because none of the older children was restrained.

Comparability of cases with controls is always an issue in case-control studies, but, in this investigation, the cases were a random sample of all cases in the study base and the controls were, by design, a random sample of that study base. Accordingly, there should be no selection bias in the study23 unless unobserved changes in traffic patterns have compromised the representative nature of the random inspection survey, which appears unlikely. An argument could also be made for using as control subjects children who were involved in car crashes but were not themselves injured. This type of control series would have some advantages, but it is difficult to assemble because uninjured children who were involved in a car crash are not easily accessible. The almost unavoidable losses would have created a strong potential for selection bias. Lastly, the exclusion from the case series of children who died as car passengers may have led to an underestimation of the protective potential of car restraints, but historical data suggest that their number is likely to be minimal (less than five).

The results of the present study provide statistically significant evidence that a lack of restraint system among infants and toddlers increases the risk of injury more than threefold. This translates into a protective effect of child restraints for younger children of almost 70%. This figure is slightly lower than, but not incompatible with, the 93% protective effect reported from a UK study.28Overall, front seating increases the risk of injury fivefold (protective effect 80%), but the relative protection appears to be higher among older (5–11 years) than among younger (0–4 years) children. There is some evidence that back seating and restraint use may have an additive protective effect, but the data do not allow firm conclusions. These findings are generally compatible with those reported by other investigators who have used different study designs and the convergence of results adds to confidence in their validity.2 3 6 14 17 18

As indicated, rear seating conveys substantial protection compared with front seating, at least among unrestrained children. The proportion of road traffic injuries that could have been avoided in the underlying population by the universal rear seating of children is deceptively small (16%) because most children in Greece are already transported in the rear seats and the potential benefit has already been harvested. Indeed, road traffic injury mortality among child car passengers is as low in Greece as in Nordic countries, and substantially lower than in the USA (table 1). It seems that when the universal use of child car restraint systems is not achievable, a policy that stresses the transportation of children in the rear can provide a substantial protective effect and outstanding cost efficiency for injury prevention while avoiding the risks associated with airbags, however minimal.

On the negative side, this study indicates that most children in Greece (91%) are transported in passenger cars without the proper use of car restraints and that a regular use of such restraints could reduce the number of childhood injuries after car crashes by two thirds. Even if we restricted consideration to children less than 5 years old, among whom 72% did not use a car restraint, and the realistic objective was to reduce this proportion to the 34% observed in the USA population,27 26% of all injuries in this age group could be prevented.

The non-use of child car restraints is more extensive in Greece than in most developed countries, but the problem is universal. Several important studies have evaluated the effectiveness of legislative, regulatory, and other public health measures in increasing compliance with passive safety measures for children in passenger cars. These studies and the effectiveness of alternative approaches have also been reviewed by Towner et al.19 29-31 Our findings confirm that the universal and proper use of child restraint systems is the ultimate objective, but they also suggest that until this goal is accomplished, children should only be transported in the rear seats. If our findings are supported by the results of additional studies, population policies and individual advice for the car transportation of children should emphasise rear seating as much as the use of restraints.