In the early, middle, and late samples over the 24-year period, there were 200, 240, and 232 fractures, respectively, in children under 3 years old. The proportion of cases rated as due to abuse decreased from 22.5 percent in the early period to 10 percent in the middle period, and leveled off at 10.8 percent in the late period. Over the period of the study, the odds of a given fracture in a child under 3 years old being rated as abuse decreased by just over 50 percent from the early period to the middle and late period. No statistically significant difference was found between the odds for an abusive fracture for the middle group and the late group. The significant decrease between the early and middle samples may reflect early recognition of less serious forms of maltreatment and the availability of services to high-risk families. Medical records were examined for all children under 3 years old who were seen at the hospital and determined to have a bone fracture. The time periods examined were 1979-1983, 1991-94, and 1999-2002. After reviewing the medical records, including radiographic information, two clinicians (one an expert on child abuse) and two pediatric radiologists each rated the likelihood of abuse being the cause of the fracture. All of the examiners used explicit criteria and a seven-point scale that ranged from "definite unintentional injury" to "definite abuse." Ratings were done independently. When disagreements in ratings occurred, the case at issue was discussed and a joint rating was reached if possible. The proportions of cases rated as abuse were compared over the three time periods, and logistic regression was used to calculate adjusted odds ratios. 7 tables and 16 references