An adjusted logistic EuroSCORE using variable correction factors improved calibration precision (Hosmer-Lemeshow p=0.891) while maintaining the moderate accuracy (AUC 0.695) of the crude score.
Observational (n=5,738)
Yes
Does an adjusted logistic EuroSCORE model improve the precision of mortality risk prediction in high-risk cardiac surgery patients compared to the crude logistic EuroSCORE?
An adjusted logistic EuroSCORE model improves the precision of mortality risk prediction in high-risk cardiac surgery patients by correcting the crude model's tendency to overestimate risk.
Effect estimate: AUC 0.695
p-value: p=0.891
OBJECTIVE: To verify the accuracy and precision of the logistic European system for cardiac operative risk evaluation (EuroSCORE) in high-risk cardiac surgery patients and to develop and externally validate a new system of recalibration. METHODS: The development series included 4279 high-risk patients who had undergone cardiac operations at the IRCCS Policlinico S. Donato. Performance, accuracy, and precision of the logistic EuroSCORE were assessed in this series, using a deciles-based comparison between expected and observed mortality rates, a receiver operating characteristic analysis, and a Hosmer-Lemeshow test for calibration. Differences between predicted and observed mortality rates were mathematically evaluated to develop an adjusted logistic EuroSCORE. This adjusted risk score was subsequently validated with the same approach on an external series of 1459 high-risk patients who had undergone cardiac operations at the Siena hospital. RESULTS: The adjusted logistic EuroSCORE was based on five different correction factors applied to the crude logistic EuroSCORE depending on its value. At the external validation, this model provided a good performance, with observed mortality rates not significantly different from the expected in 8 out of 10 deciles of risk distribution. The adjusted EuroSCORE had the same moderate balanced accuracy of the crude logistic EuroSCORE (area under the curve: 0.695), with a better precision (Hosmer-Lemeshow calibration test: chi(2): 3.6, p=0.891). CONCLUSIONS: Recalibration of the logistic EuroSCORE in high-risk patients is needed due to its tendency to overestimate the mortality risk. The application of a variable correction factor results in a better performance, increased precision, with unaltered balanced accuracy.
Ranucci et al. (Thu,) conducted a observational in High-risk cardiac surgery (n=5,738). Adjusted logistic EuroSCORE vs. Crude logistic EuroSCORE was evaluated on Performance, accuracy, and precision of mortality prediction (AUC 0.695, p=0.891). An adjusted logistic EuroSCORE using variable correction factors improved calibration precision (Hosmer-Lemeshow p=0.891) while maintaining the moderate accuracy (AUC 0.695) of the crude score.