In 1981-82, 5, 963 children's accidents were reported over a one-year period in 2 public and 11 private hospitals in a given public health district. The nature of treated accidents in the public hospitals differs from that in the private facilities. Reporting of cases reaches 75% in the public hospitals; only about 50% in the private hospitals. Characteristics of reported accidents differ from those unreported. While the quality of clinical data is satisfactory, the site of the accident is unknown in 14%, and the causal agent, in 35% of the cases. The estimation of incidence implies the extension of registrations to the private sector, source of 31% of reported cases, and where significant differences from the public sector were observed in the nature of the accidents. A permanent system requires use of a rolling sample and training of personnel. Systematic registration is a tool to measure incidence of accidents, but is not a suitable instrument for the surveillance of dangerous products and objects.