Sir.—The excellent article by McClelland and Heiple (Journal 1982;136: 26-29) on fractures in infancy points to a way in which this challenging problem may be studied. However, one comment about its design is necessary.
The authors defined nonaccidental fractures as those accompanied by discrepant or variable histories, those accompanied by delay in treatment or other nonaccidental injuries, or those with multiple unexplained injuries. Accidental fractures seem to have been defined as all others observed.
Although this definition effectively prevented accidents from being diagnosed as nonaccidents, it did not necessarily do the reverse. If a plausible accident was described, the history was accepted as true and the fracture classified as accidental. Documentation of the accident was not required.
This method of definition should dictate a change in the way the data are summarized. Of the 34 patients with fractures observed, the injuries to 19 or more were defined as