This study examines the validity of the Revised Estimated Survival Probability (RESP) index in a set of trauma patients admitted to three hospitals. For each patient four different severity indices were computed: 1) RESP derived from in-hospital assigned International Classification of Disease (ICD) codes; 2) RESP based on written face sheet discharge diagnoses; 3) RESP based on a full review of the medical record; and 4) Injury Severity Score (ISS) based on full review of the medical record. These four severity indices were then correlated with six measures of outcome or construct validity, including mortality, duration of hospitalization, intubation or tracheostomy performed, ambulance transport to hospital, admission to the intensive care unit, and ventilatory assistance received. The results indicate that for every validity measure examined, the ISS index was superior to the RESP index, regardless of the abstraction procedure. However, the RESP index was independently associated with mortality, length of hospitalization, and ventilatory assistance even after adjusting for the ISS. In addition, the performance of the RESP index improved dramatically as the quality of information improved. Last, strong evidence is presented which questions the utility of calculating any type of severity index using data from computerized discharge abstracts without careful quality control measures.