Predicting Laparotomy Outcomes Using the American College of Surgeons National Surgical Quality Improvement Program...

Background Risk prediction in emergency laparotomy supports peri-operative decision-making, counselling and resource allocation in acutely unwell surgical patients. In the United Kingdom (UK) practice, the National Emergency Laparotomy Audit (NELA) score is widely used because it is specific to emergency laparotomy, whereas the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) calculator incorporates broader comorbidity and functional status variables. Direct comparisons between these tools in UK emergency laparotomy cohorts remain limited. The primary objective of this study was to compare the discriminatory performance of pre-operative NELA, post-operative NELA and ACS-NSQIP for predicting 30-day mortality after emergency laparotomy. Secondary objectives were to compare overall prediction accuracy using the Brier score and to describe threshold-based test characteristics. Methods We performed a retrospective observational two-centre cohort study at two district general hospitals. Adult patients undergoing emergency laparotomy between November 2023 and October 2024 were identified from electronic records. Mortality risk was calculated using ACS-NSQIP, pre-operative NELA and post-operative NELA. Pre-operative NELA was derived from variables available at the time of operative decision-making, whereas post-operative NELA incorporated intra-operative variables. The primary outcome was 30-day post-operative mortality. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and Brier score. DeLong's test was used to compare AUCs. Results A total of 138 patients were included; 30-day mortality was 11.6% (16/138). Post-operative NELA had the highest AUC at 0.878 (95% CI: 0.762-0.943), followed by pre-operative NELA at 0.863 (95% CI: 0.762-0.943) and ACS-NSQIP at 0.826 (95% CI: 0.765-0.942). Brier scores were 0.077, 0.078 and 0.094, respectively. Although post-operative NELA showed numerically higher discrimination and lower overall prediction error, differences between models were not statistically significant, including comparison with ACS-NSQIP (DeLong's test; p = 0.616). At a descriptive 10% mortality threshold, post-operative NELA had higher sensitivity than ACS-NSQIP but slightly lower specificity. Conclusion In this two-centre retrospective cohort, all three models demonstrated good performance for predicting 30-day mortality after emergency laparotomy. Post-operative NELA performed numerically best, but differences between models were statistically non-significant and should be interpreted cautiously given the small sample size and limited number of mortality events. These findings are exploratory and support further evaluation in larger multicentre cohorts.