Planning of accident prevention requires detailed information on accidents, injuries, costs for medical care, and other consequences for individual and society.
In Göteborg a system has been established for continuousr egistrationo fdata on traffrca ccidents(c fBunketorp and Romanus 1978) .This system is now expanded to include collection of data on injuries and medical costs.
The aim of this report is to describe the methodology, diffrculties and costs for data collection and also to discuss improvements of the system and exemplify the results obtained and their implications.
250 patients at "Ostra sjukhuset" in Göteborg (traffrc injuries) and "Huddinge sjukhus" (abdominal and thoracic injuries).
Data from the medical records have been collected by personnel with various qualifications; doctors, medical students and nurses. Types of information: individual, pretraumatic medical status, accident, injuries (type and severity according to the Abbreviated Injury Scale Classification), length and level of medical care, quantity of specifrc medical procedures (consultations, x-ray, lab tests, operations). The information has been structured into a form suitable for computer processing and entered via a terminal. Search for possible relations between trauma, costs, outcome, etc. has been made utilizing a specially developed program package.
The time required for data collection and entering into computer was successively reduced for 20-30min/patient, still the errors encountered during these procedures seem to be less than 4%, varying with the type of data and the qualification and experience of the personnel. From the present experience we are able to optimize the process of data acquisition concerning the amount of information, degree of uncertainty and costs. In this retrospective study, based on hospital records, we found a large percentage of uncertain (0-20%) and missing data (0-50%) varying with the type of information.
In order to reduce these figures a prosp€ctive study must be carried out. It is concluded that the present method would be well suited for analysis of accident-injury relationships and cost-efficiency of medical care in a large scale multicenter study.