Evaluating and improving current risk prediction tools in emergency laparotomy.

Abstract

Emergency laparotomy (EL) encompasses a high-risk group of operations, which are increasingly performed on a heterogeneous population of patients, making preoperative risk assessment potentially difficult. The UK National Emergency Laparotomy Audit (NELA) recently produced a risk predictive tool for EL that has not yet been externally validated. We aimed to externally validate and potentially improve the NELA tool for mortality prediction after EL. We reviewed computer and paper records of EL patients from May 2012 to June 2017 at Middlemore Hospital (New Zealand). The inclusion criteria mirrored the UK NELA. We examined the NELA, Portsmouth Physiological and Operative Severity Score for the enUmeration of Mortality (P-POSSUM), Acute Physiology and Chronic Health Evaluation II (APACHE-II), and American College of Surgeons National Surgical Quality Improvement Programs risk predictive tools for 30-day mortality. The Hosmer-Lemeshow test was used to assess calibration, and the c statistic, to evaluate discrimination (accuracy) of the tools. We added the modified frailty index (mFI) and nutrition to improve the accuracy of risk predictive tools. A total of 758 patients met the inclusion criteria, with an observed 30-day mortality of 7.9%. The NELA was the only well calibrated tool, with predicted 30-day mortality of 7.4% (p = 0.22). When combined with mFI and nutritional status, the c statistic for NELA improved from 0.83 to 0.88. American College of Surgeons National Surgical Quality Improvement Programs, APACHE-II, and P-POSSUM had lower c statistics, albeit also showing an improvement (0.84, 0.81, and 0.74, respectively). We have demonstrated the NELA tool to be most predictive of mortality after EL. The NELA tool would therefore facilitate preoperative risk assessment and operative decision making most precisely in EL. Future research should consider adding mFI and nutritional status to the NELA tool. Level IV; Retrospective observational cohort study.