OBJECTIVES: The authors determine whether assessments of effects of rural emergency medical service (EMS) system characteristics on trauma outcomes (using patient-level data) are significantly biased if the Injury Severity Score (ISS) is not available. METHODS: The data are from ambulance trip reports merged with the trauma registry data for the Georgia EMS region VI trauma center hospital, located in Augusta. All 294 trauma patients for the rural counties surrounding Richmond County for the calendar year 1991 who were not dead at the scene and were treated at the trauma center are included. A 20% random sample of trauma patients from Richmond county from May to September 1991 not dead at the scene and treated at the trauma center yielded an additional 96 cases. Excluding 43 patients with missing data yields 347 trauma cases with 18 trauma deaths. A logistic regression model for trauma mortality is estimated using the Revised Trauma Score (RTS), ISS, type of trauma, and patient age (analogous to the standard Trauma Related Injury Severity Score model). The predicted probability of patient mortality from this model is compared with the predicted probability of mortality when the logistic regression model omits ISS. Correlations between the difference in predicted probability (i.e., the error in predicted probability associated with the omitted ISS variable) and EMS system characteristics are determined. RESULTS: Although ISS adds to the predictive power of the trauma outcome model, the errors in predicted probabilities associated with the omission of ISS generally are small and uncorrelated with patient or EMS system characteristics, with the exception of patient gender. CONCLUSIONS: In rural settings, where a patient's ISS generally is not available, studies of rural EMS system characteristics and trauma outcomes may use RTS, patient age, and type of trauma to control for expected survival. The patient's ISS does not appear to be essential, at least for the rural area analyzed in this study.
OBJECTIVES: The authors determine whether assessments of effects of rural emergency medical service (EMS) system characteristics on trauma outcomes (using patient-level data) are significantly biased if the Injury Severity Score (ISS) is not available. METHODS: The data are from ambulance trip reports merged with the trauma registry data for the Georgia EMS region VI trauma center hospital, located in Augusta. All 294 traumapatients for the rural counties surrounding Richmond County for the calendar year 1991 who were not dead at the scene and were treated at the trauma center are included. A 20% random sample of traumapatients from Richmond county from May to September 1991 not dead at the scene and treated at the trauma center yielded an additional 96 cases. Excluding 43 patients with missing data yields 347 trauma cases with 18 trauma deaths. A logistic regression model for trauma mortality is estimated using the Revised Trauma Score (RTS), ISS, type of trauma, and patient age (analogous to the standard Trauma Related Injury Severity Score model). The predicted probability of patient mortality from this model is compared with the predicted probability of mortality when the logistic regression model omits ISS. Correlations between the difference in predicted probability (i.e., the error in predicted probability associated with the omitted ISS variable) and EMS system characteristics are determined. RESULTS: Although ISS adds to the predictive power of the trauma outcome model, the errors in predicted probabilities associated with the omission of ISS generally are small and uncorrelated with patient or EMS system characteristics, with the exception of patient gender. CONCLUSIONS: In rural settings, where a patient's ISS generally is not available, studies of rural EMS system characteristics and trauma outcomes may use RTS, patient age, and type of trauma to control for expected survival. The patient's ISS does not appear to be essential, at least for the rural area analyzed in this study.