25 When Scoring Systems Collide: Predicting Laparotomy Outcomes with NSQiP & NELA - A Retrospective Analysis from Two Centres

Abstract Aim Accurate surgical risk prediction in emergency laparotomy is essential for informed, shared decision-making. While the NELA calculator is widely used in the UK for predicting 30-day mortality, this study investigates whether the ACS-NSQIP model, which incorporates additional patient factors (e.g. age, ASA class, functional status), provides superior predictive accuracy. Method A retrospective analysis of 138 emergency laparotomy patients was conducted, comparing pre- and post-operative NELA models with ACS-NSQIP. Predictive performance was assessed using area under the ROC curve (AUC) and Brier scores. Statistical significance of differences between models was evaluated using DeLong’s test. Results Post-operative NELA achieved the highest AUC (0.878) and the lowest Brier score (0.077), indicating slightly superior performance compared to both Pre-op NELA and ACS-NSQIP (AUC 0.863 and 0.826, Brier scores 0.078 and 0.094 respectively). However, DeLong’s test between Post-op NELA and NSQIP (p=0.616) indicated no statistically significant difference. Post-op NELA also showed higher sensitivity (0.94) but lower specificity (0.73), at 10% threshold, than ACS-NSQIP (0.81 and 0.76, respectively). Among ASA IV patients (n=25), pre-op NELA more closely approximated actual 30-day mortality (21.5% vs. 28%). Overall, 30-day mortality was 11.6% (16/138). Conclusions While Post-operative NELA demonstrated marginally better predictive metrics, there was no statistically significant difference between ACS-NSQIP and NELA in predicting 30-day mortality following emergency laparotomy. Due to the small sample size, these findings should be interpreted cautiously. Larger studies are warranted to validate and clarify the comparative utility of these tools.