Over a 2-year period 1,116 children admitted to an urban teaching hospital were studied prospectively. The overall group was analysed as to the nature of the injury and a subgroup of seriously injured children was identified and further analysed. All deaths were examined as to their cause and possible preventable as well as salvageable factors. The predictive value of the Trauma Score (T.S.) and Method of Injury (M.O.I.) were evaluated for their prospective prediction of serious injury as determined by the Injury Severity Score and outcome. Most of the children were not seriously injured, with the most common injury being due to a fall (57%) and involving a single injury to the upper limb. With the subgroup of 143 children (13% of the total) who suffered serious injuries, the cranial cavity (90%) was the most common site of injury, occurring most often in pedestrians (31% of the total injured). There were 16 deaths in the series, representing 1.4% of all paediatric trauma admissions and 11% of the admissions who were seriously injured. All deaths were related to motor vehicle accidents and associated with serious head injury. A Trauma Score less than or equal to 12 accurately included all deaths but when correlated with the I.S.S., the Trauma Score had a specificity of 99% and a positive predictive value of 86%; its sensitivity was only 27%. The Method of Injury was associated with an overtriage rate of 300% in relation to the I.S.S.. Of children admitted following pedal cycle accidents only 9% were wearing helmets. Of car occupants injured, 39% were unrestrained.(ABSTRACT TRUNCATED AT 250 WORDS)