POSSUM versus ACS NSQIP for Morbidity and Mortality Prediction in Indian Surgical Patients

Abstract

Non-cardiac surgery carries substantial and variable perioperative risk, and accurate preoperative risk stratification remains central to informed consent, resource planning, and quality benchmarking; however, the performance of widely used risk models such as POSSUM and the ACS NSQIP Surgical Risk Calculator in Indian surgical populations remains insufficiently characterized, given differences in comorbidity burden, nutritional status, and disease presentation compared with the Western cohorts in which these tools were developed. This prospective observational study included 163 adults undergoing elective or emergency non-cardiac surgery at a South Indian tertiary care centre; physiological, operative, and preoperative variables were recorded, and predicted morbidity and mortality risks from POSSUM and ACS NSQIP were compared against observed 30-day outcomes using AUROC with DeLong testing, Hosmer–Lemeshow calibration, and Youden-index-derived sensitivity, specificity, and predictive values. Morbidity occurred in 47 patients (28.8%) and mortality in 9 (5.5%). POSSUM showed numerically higher discrimination for morbidity (AUROC 0.708 vs. 0.638), whereas ACS NSQIP discriminated mortality better (AUROC 0.882 vs. 0.839), though neither difference reached statistical significance (p=0.279 and p=0.365, respectively); both models showed acceptable calibration on Hosmer–Lemeshow testing, with under-prediction of risk in higher-risk strata for morbidity. These findings indicate that POSSUM and ACS NSQIP perform comparably overall, without demonstrable superiority of either model, but may serve complementary clinical roles, with ACS NSQIP being useful for preoperative risk communication and POSSUM for postoperative audit.