Assessing Severity of Illness in Patients Transported to Hospital by Paramedics: External Validation of 3 Prognostic Scores

Introduction: Emergency Medical Services (EMS) are the first healthcare contact for the majority of severely ill patients. Physiologic measures collected by EMS, when incorporated into a prognostic score, may provide important information on patient illness severity. This study compares the predictive ability of 3 common prognostic scores for predicting clinical outcomes in EMS patients. Methods: Discrimination and calibration for predicting the primary outcome of hospital mortality, and secondary outcomes of 2-day mortality and ED disposition, were assessed for each of the scores using a one-year cohort of patients transported to hospital by EMS in Alberta, Canada. For each score, binary logistic regression was used to predict hospital mortality and 2-day mortality and ordinal logistic regression was used to predict ED disposition. Discrimination for each outcome was assessed using C-statistics, and calibration was assessed using calibration curves comparing predicted versus observed outcomes. Results: The Critical Illness Prediction CIP, Modified Early Warning Score MEWS, and National Early Warning Score NEWS were compared using 121,837 adult patients who were transported by paramedics. All scores had good discrimination for hospital mortality (C-statistic CIP: 0.79, MEWS: 0.71, NEWS: 0.78) and 2-day mortality (CIP:0.85, MEWS: 0.80, NEWS:0.85) but only moderate discrimination for ED disposition (CIP: 0.68, MEWS: 0.61, NEWS: 0.66). Calibration was reliable for hospital mortality in all scores but over-predicted risk for 2-day mortality at higher scores. Overall, the CIP score had the best discrimination, good calibration, and the greatest range of predicted probabilities (0.01 at a CIP score of 0 to 0.92 at a CIP score of 8) for hospital mortality. Conclusions: Prognostic scores using physiologic measures assessed by paramedics have good predictive ability for hospital mortality. These scores, particularly the CIP score, may be considered as a tool for mortality risk stratification or as a general measure of illness severity for patients included in EMS studies.