Development and validation of a postoperative risk calculator (POP-score) for patients undergoing cardiac surgery: a retrospective cohort study.

Abstract

This study aimed to identify intraoperative and perioperative factors influencing 30-day mortality after cardiac surgery and to develop a risk score (POP-score) for its prediction. Retrospective cohort study with multivariable regression analysis. A tertiary care cardiac surgery centre in Austria; data from consecutive patients undergoing cardiac surgery between 2010 and 2020 were analysed. A total of 8072 patients were included. The cohort was randomly divided into a derivation cohort (75%) and a validation cohort (25%). The primary outcome measure was 30-day mortality. We analysed associations between intraoperative and perioperative variables and 30-day mortality, assessed via multivariable regression analysis. Several factors were significantly associated with 30-day mortality, including intraoperative RBC transfusion (OR 3.407 (95% CI 2.124-5.464)), postoperative high-sensitive cardiac troponin T cut-off levels (OR 2.856 (95% CI 1.958 to 4.165)), need for dialysis/haemofiltration (OR 2.958 (95% CI 2.013 to 4.348)) and temporary extracorporeal membrane oxygenation support (OR 5.218 (95% CI 3.329 to 8.179)) (p<0.001 for all). The newly developed POP-score demonstrated superior predictive performance for 30-day mortality compared with the EuroSCORE II alone (area under the ROC curve (AUC) 0.884 vs 0.800, p=0.013), based on peak troponin values assessed within the first 7 postoperative days. As 98% of peak troponin elevations occurred within 72 hours, the POP-score can be calculated at this earlier time point for clinical implementation. The validated POP-score provides an improved tool for predicting 30-day mortality after cardiac surgery by incorporating intraoperative and perioperative factors alongside the EuroSCORE II. Although model performance was evaluated using 7-day peak troponin data, the score can be calculated within the first 72 hours postoperatively in most patients, supporting its clinical applicability for early decision-making, resource allocation and patient counselling. Further research is warranted to assess its clinical utility in diverse populations.