To describe the incidence of severe traffic injuries before and after implementation of a comprehensive, hospital-initiated injury prevention program aimed at the prevention of traffic injuries to school-aged children in an urban community. Hospital discharge and death certificate data on severe pediatric injuries (ie, injuries resulting in hospital admission and/or death to persons age <17 years) in northern Manhattan over a 13-year period (1983-1995) were linked to census counts to compute incidence. Rate ratios with 95% CIs, both unadjusted and adjusted for annual trends, were calculated to test for a change in injury incidence after implementation of the Harlem Hospital Injury Prevention Program. This program was initiated in the fall of 1988 and continued throughout the study period. It included 1) school and community based traffic safety education implemented in classroom settings in a simulated traffic environment, Safety City, and via theatrical performances in community settings; 2) construction of new playgrounds as well as improvement of existing playgrounds and parks to provide expanded off-street play areas for children; 3) bicycle safety clinics and helmet distribution; and 4) a range of supervised recreational and artistic activities for children in the community. Traffic injuries were a leading cause of severe childhood injury in this population, accounting for nearly 16% of the injuries, second only to falls (24%). During the preintervention period (1983-1988), severe traffic injuries occurred at a rate of 147.2/100 000 children <17 years per year. Slightly <2% of these injuries were fatal. Pedestrian injuries accounted for two thirds of all severe traffic injuries in the population. Among school-aged children, average annual rates (per 100 000) of severe injuries before the intervention were 127.2 for pedestrian, 37.4 for bicyclist, and 25.5 for motor vehicle occupant injuries. Peak incidence of pedestrian and bicyclist injuries occurred during the summer months and afternoon hours, whereas motor vehicle occupant injuries showed little seasonal variation and were more common during evening and night-time hours. Age-specific rates showed peak incidence of pedestrian injuries among 6- to 10-year-old children, of bicyclist injuries among 9- to 15-year-old children, and of motor vehicle occupant injuries among adolescents between the ages of 12 and 16 years. The peak age for all traffic injuries combined was 15 years, an age at which nearly 3 of every 1000 children each year in this population sustained a severe traffic injury. Among children hospitalized for traffic injuries during the preintervention period, 6.3% sustained major head trauma (including concussion with loss of consciousness for >/=1 hour, cerebral laceration and/or cerebral hemorrhage), and 36.9% sustained minor head trauma (skull fracture and/or concussion with no loss of consciousness >/=1 hour and no major head injury). The percentage of injured children with major and minor head trauma was higher among those injured in traffic than among those injured by all other means (43.2% vs 14.2%, respectively; chi2 = 336; degrees of freedom = 1). The percentages of children sustaining head trauma were 45.4% of those who were injured as pedestrians, 40.2% of those who were injured as bicyclists, and 38.9% of those who were injured as motor vehicle occupants. During the intervention period, the average incidence of traffic injuries among school aged children declined by 36% relative to the preintervention period (rate ratio:.64; 95% CI:.58,.72). After adjusting for annual trends in incidence, pedestrian injuries declined during the intervention period among school aged children by 45% (adjusted rate ratio:.55; 95% CI:.38,.79). No comparable reduction occurred in nontargeted injuries among school-aged children (adjusted rate ratio:.89; 95% CI:.72, 1.09) or in traffic injuries among younger children who