External validation of EuroSCORE II, additive EuroSCORE, and the STS predicted risk of mortality at 2,600 meters above sea level: a retrospective cohort study from a Latin American cardiac surgery centre

Background EuroSCORE II, the Additive EuroSCORE, and the Society of Thoracic Surgeons Predicted Risk of Mortality (STS PROM) are widely used cardiac surgery risk models, but their performance has not been evaluated in a cardiac surgery centre located at high altitude (≥2,500 m asl). Methods We conducted a retrospective external validation study using prospectively collected registry data from consecutive adult cardiac surgery procedures performed at Hospital Universitario Mayor-Méderi (2,600 m asl), Bogotá, Colombia, between January 2019 and February 2026 ( N = 1,486). EuroSCORE II and the Additive EuroSCORE were calculated from the registry with the published models and evaluated in the full cohort; STS PROM was obtained from the official STS online calculator (2018 models) for the procedures within its scope ( n = 1,076). The outcome was in-hospital mortality. Discrimination was assessed with the area under the receiver operating characteristic curve (AUC), and calibration with the observed to expected (O/E) ratio, calibration in the large, calibration slope, and calibration plots. Results In-hospital mortality was 7.9% (117/1,486), and 82.3% of procedures were non-elective. The AUC was 0.736 (95% CI 0.687 to 0.785) for EuroSCORE II, 0.733 (0.684 to 0.783) for the Additive EuroSCORE, and 0.815 (0.757 to 0.873) for STS PROM. All three models underestimated mortality: the O/E ratio was 2.50 (2.09 to 3.00) for EuroSCORE II, 1.50 (1.25 to 1.79) for the Additive EuroSCORE, and 2.72 (2.14 to 3.47) for STS PROM. Calibration slopes were 1.04 (0.82 to 1.26), 1.94 (1.46 to 2.41), and 1.31 (1.01 to 1.60), respectively. Underestimation was present across procedure types, risk strata, and calendar years, and persisted in sensitivity analyses of variable mapping and missing data (EuroSCORE II O/E 2.37 to 3.05). Conclusions In this high-altitude Colombian centre, EuroSCORE II, the Additive EuroSCORE, and STS PROM ranked patients acceptably, STS PROM best, but all underestimated in-hospital mortality, EuroSCORE II and STS PROM by more than twofold. The study design cannot separate an effect of altitude from differences in case mix, urgency, and local care. Local recalibration is needed before these models are used for counselling or benchmarking in this setting.