669 Mortality Risk Prediction for Emergency Laparotomy: Are We Utilising the Best Tool?
Abstract Aim Emergency laparotomy is one of the common operations performed in the UK. To aid in more objective decision making, several scoring systems have been formulated. In this project, we aimed to explore the predictive power of both NELA and P-Possum mortality scores against 30 days and 90 days observed mortality for emergency laparotomy patients. Method Patient details from two large district general hospitals were extracted from the NELA database over a period of three years. Pre-operative NELA, post-operative NELA, and P-POSSUM predicted mortality were calculated and compared with the observed 30 days and 90 days mortality for the entire cohort. Model discrimination (statistical accuracy) was tested by calculating the area under the receiver operating characteristic curve (AUC), which was used to assess how accurately the model could discriminate. Results There were 378 patients eligible for inclusion with a median age of 64. 39 patients (10.3%) died within 30 days and 52 patients (13.8%) died within 90 days. P-POSSUM score, pre-operative NELA, and post-operative NELA scores predicted the 30 days mortality as (2.7%, 3.7%, and 2.4%) and 90 days mortality as (2.9%, 4.8%, and 4%) respectively. The discriminative power for 30 days and 90 days mortality was highest for the pre-operative NELA score (AUC 0.870, CI: 0.824 – 0.916), (AUC 0.826, CI: 0.769 – 0.884) respectively. Conclusions Both NELA and P-Possum scores underpredicted actual 30 days and 90 days mortality. It was however noted that the pre-operative NELA mortality score showed more accurate mortality discriminative power than the other 2 tested tools.