An artificial neural network as a model for prediction of survival in trauma patients: Validation for a regional trauma area

Background: To develop and validate an artificial neural network (ANN) for predicting survival of trauma patients based on standard prehospital variab les, emergency room admission variables, and Injury Severity Score (ISS) us ing data derived from a regional area trauma system, and to compare this mo del with known trauma scoring systems. Patient Population: The study was composed of 10,609 patients admitted to 2 4 hospitals comprising a seven-county suburban/rural trauma region adjacent to a major metropolitan area. The data was generated as part of the New Yo rk State trauma registry. Study period was from January 1993 through Decemb er 1996 (1993-1994: 5,168 patients; 1995: 2,768 patients; 1996: 2,673 patie nts). Methods: A standard feed-forward back-propagation neural network was develo ped using Glasgow Coma Scale, systolic blood pressure, heart rate, respirat ory rate, temperature, hematocrit, age, sex, intubation status, ICD-9-CM In jury E-code, and ISS as input variables. The network had a single layer of hidden nodes. Initial network development of the model was performed on the 1993-1994 data, Subsequent models were generated using the 1993, 1994, and 1995 data. The model was tested first on the 1995 and then on the 1996 dat a. The ANN model was tested against Trauma and Injury Severity Score (TRISS ) and ISS using the receiver operator characteristic (ROC) area under the c urve [ROC-A(z)], Lemeshow-Hosmer C-statistic, and calibration curves. Results: The ANN showed good clustering of the data, with good separation o f nonsurvivors and survivors. The ROC-A(z) was 0.912 for the ANN, 0.895 for TRISS, and 0.766 for ISS, The ANN exceeded TRISS with respect to calibrati on (Lemeshow-Hosmer C-statistic: 7.4 for ANN; 17.1 for TRISS), The predicti on of survivors was good for both models. The ANN exceeded TRISS in nonsurv ivor prediction, Conclusion: An ANN developed for trauma patients using prehospital, emergen cy room admission data, and ISS gave good prediction of survival. It was ac curate and had excellent calibration. This study expands our previous resul ts developed at a single Level I trauma center and shows that an ANN model for predicting trauma deaths can be applied across hospitals with good resu lts.