Two of the four injury databases are based on trauma center patient records; one is based on statewide hospital discharge and death certificate data; and one is based on emergency department visits, hospitalization, and death certificate data. Although the four databases differ substantially, each of them showed a large disparity between the case-fatality rates for inflicted and unintentional injuries. The extent of the disparity apparently is inversely related to the incidence (or the recognition rate) of inflicted injury in each database. The lowest incidence rate is associated with the largest case-fatality rate (CFR) disparity ratio. This is probably due to selection for severity in defined cases of inflicted injury. Younger populations had a higher incidence of inflicted injuries during the period covered in each of the four databases. There is no reason to believe that inflicted injuries are intrinsically more severe than unintentional ones, however; and there is no practical method available to determine whether they are or are not. The disparity between inflicted and unintentional injuries is most likely due to artificial diagnostic or definitional factors. The authors conclude that current diagnostic criteria for physically abusive (inflicted) injuries are forensically driven (diagnosis must be beyond a reasonable doubt); however, such diagnostic criteria are too conservative for public health purposes. New public-health-oriented case definitions for "inflicted injury" are needed. Programs for reducing injury recidivism in young children should be a part of overall injury prevention. 7 tables and 41 references