Authors (including presenting author) :
Ng HL, Fujikawa T, Kwok WT, Ho YK, Chow CY, Chan WY, Lim K, Chang TC, Siu CH, Wong HL
Affiliation :
Division of Cardiothoracic Surgery, Department of Surgery, Prince of Wales Hospital
Introduction :
EuroSCORE II, widely used and validated in European cardiac surgery cohort, lacks specific validation for Asian populations. Limited research on its effectiveness in Asia has led to uncertainty about its applicability in Hong Kong. Local validation is essential before clinical use.
Objectives :
The study's primary aim is to evaluate EuroSCORE II's predictive accuracy for cardiac surgery outcomes at Prince of Wales Hospital (PWH) in Hong Kong.
Methodology :
A 10-year retrospective cohort study was conducted on 4,180 adult patients aged 18 and above in Hong Kong who underwent cardiac surgery at PWH from January 1, 2013, to December 31, 2023. Pre-operative and mortality data were sourced from the Hospital Authority Electronic Patient Record system and entered into the Dendrite Clinical System. Data variables included pre-operative patient-related, cardiac-related, and operation-related risk factors for EuroSCORE II calculation in cardiac surgery. The study's aim was to validate EuroSCORE II for predicting 30-day mortality in post-cardiac surgery patients. EuroSCORE II's discriminatory and calibration performance were evaluated based on statistical techniques such as test accuracy, area under the receiver operating characteristic curve (AUROC), and Hosmer-Lemeshow tests.
Result & Outcome :
The overall cohort exhibited an AUROC of 0.829, indicating strong discrimination by the EuroSCORE II system in cardiac surgery. Test accuracy metrics (sensitivity, specificity, negative predictive value, accuracy) were 0.776, 0.751, 0.987, and 0.752 respectively, using a maximum Youden’s Index cutoff point of 5.015. Stratified analysis revealed AUROCs of 0.729 and 0.846 in aortic and non-aortic groups, with a significant AUROC difference (p = 0.003) via the DeLong test. Notably, EuroSCORE II exhibited superior predictive ability for postoperative mortality in non-aortic surgeries. In CABG-related and valve-related cohorts, AUROCs were 0.874 and 0.830 respectively, with a non-significant difference (p = 0.329). Besides, EuroSCORE II categorizes patients into low risk (≤4%), intermediate risk (>4% but ≤8%), and high risk (>8%). For these groups, AUROCs were 0.698, 0.621, and 0.724 respectively. EuroSCORE II demonstrates strongest discriminatory power in high-risk patients. Significant difference on AUROC was found between intermediate-risk and high-risk cohorts (p = 0.025). However, there were no significant differences between low-risk and intermediate-risk cohorts (p = 0.146) and between low-risk and high-risk cohorts (p = 0.588) respectively. Furthermore, the Hosmer–Lemeshow test indicated good logistic regression fit for aortic (p = 0.667) and CABG-related (p = 0.119) cohorts. To conclude, EuroSCORE II effectively predicts postoperative mortality, especially in non-aortic cardiac surgeries at our center. This could enhance high-risk patient identification for adverse outcomes, prompting exploration of alternative treatment therapies.
