Validation of an injury surveillance epidemiological data system used within emergency departments

Objectives To establish ascertainment rates, validity, and the presence of bias within the Victorian Emergency Minimum Data set (VEMD). Methods Experienced researchers conducted interviews using a paper-based instrument to collect injury details. These data were obtained in parallel with the routine electronic collection of injury data within four of the 25 VEMD hospitals. Data were matched using the unique date and time of presentation, as well as birth date. Interviews were conducted with 481 injured persons. Results Electronic and paper records were successfully matched in 382 cases. A high mean capture rate (82.5%) across hospitals was found when the interview data and VEMD data were compared. Data were mostly coded with some inaccuracy, 87% of cases had at least one error, yet when compared with interview descriptions, the coded injury and incident data were usually valid (83.9%). However, narrative data provided information beyond coded data for only 14.1% of cases. Conclusions These results suggest that the VEMD is a reliable and valid computerised data set, but that the case narratives require attention.