681

Introduction: We compared the prognostic value for mortality of a recently derived pre-hospital critical illness score to established intensive care unit (ICU) and emergency department (ED) scoring systems. Hypothesis: We hypothesized that the ICU and ED scores would outperform a pre-hospital emergency care scoring system. Methods: 227 subjects were enrolled in a prospective observational cohort of critically ill patients admitted to the ICU via the ED. We compared the pre-hospital critical illness prediction score developed by Seymour et al (JAMA 2010) to the ICU scores Acute Physiology and Chronic Health Evaluation (APACHE) II, APACHE III, Simplified Acute Physiology Score (SAPS) II, and the ED scores Modified Early Warning Score (MEWS), Rapid Emergency Medicine Score (REMS), and Prince of Wales Emergency Department Score (PEDS). The primary endpoint was 60-day mortality. We compared the receiver operating characteristic curves of the different scores and assessed calibration using the Hosmer-Lemeshow goodness-of-fit test. Results: The areas under the curve (AUC) for the ICU scores were higher (AUC 0.779-0.799) than the ED scores (AUC 0.698-0.709). Performance of the Seymour score was intermediate between the ED scores and ICU scores, with an AUC of 0.743; this AUC was not significantly different than that of the best performing ICU score, APACHE III (AUC 0.799; p=0.09). In contrast, the ED scores did not discriminate as well as APACHE III (p=0.005 to 0.01 for comparison of ED scores to APACHE III). The differences in discrimination among the Seymour score and the ED scores were not significantly different (AUC 0.698-0.743; p=0.45). There were no significant differences among the ICU-based scoring systems (AUC 0.779-0.799; p=0.60). All scores had acceptable calibration. Conclusions: Although primarily developed to predict need for critical illness prior to arrival to the hospital, the Seymour score shows promise in predicting mortality for critically-ill patients admitted from the ED to the ICU and should be further evaluated in larger cohorts. Support: HL090833, Flight Attendant Medical Research Institute, KL2RR024130 from NCRR (component of NIH)