Case definition in childhood accident studies: a vital factor in determining results.

Research into possible aetiological factors associated with childhood accidents has failed to produce a consistent picture. In this paper we investigate the extent to which these discrepancies are attributable to different methods of case ascertainment. The approach was to use three different criteria for identifying accidents and to apply a number of commonly used statistical techniques to eight social and environmental factors. The data base consists of a nationally representative sample of 13 135 children (the Child Health and Education Study). In this way, broadly similar profiles were obtained for children reported to have had at least one accident in the first five years and for those who were said to be accident repeaters--the major risk factors in common for these two outcomes were young maternal age and residence in 'average' or 'well-to-do' urban areas. On the other hand, there were major differences in the results when admission to hospital for an accident was taken as the outcome--although young maternal age was still strongly associated, large family size and the loss or replacement of a natural parent were now also dominant risk factors. There was no relationship with area of residence. The conclusion from these example analyses is therefore that variation in case selection can lead to different conclusions about the risk factors associated with childhood accidents.