Children 4-14 are generally too large for child seats and too small for seat belts designed for adults. We wished to examine the effectiveness of seat belts for these children, and the factors associated with altered effectiveness (age of child, seat position). Records of all crashes in Ontario for 1988-92 where either a child occupant or the driver was killed or admitted to hospital were included. Using Evans's double pair comparison method, we estimated the odds ratio for death or hospitalization of the children in these crashes as an indicator of the effectiveness of the belts.

There were 2089 crashed vehicles where either the child or the driver was killed or hospitalized. Belt use was 73.3% for children and 80.0% for drivers. The odds ratio for death or serious injury among belted children, controlling for driver's belt status, was 0.58 (95% confidence limits:​ 0.46 to 0.75). After controlling for age, seat position and other factors the odds ratio is 0.60 (0.47 to 0.77). These results suggest that seat belts are at least as effective for children as for the adults for whom they were designed.