Navneet Singh1, Damian Gimpel2, Grant Parkinson3, Paul Conaglen3, Felicity Meikle3, Zaw Lin3, Nand Kejriwal2, Nicholas Odom3, David J McCormack2, Adam El-Gamel4. 1. School of Medicine, The University of Auckland, Auckland, New Zealand; Waikato Cardiothoracic Unit, Waikato Hospital, Hamilton, New Zealand. Electronic address: nsin825@aucklanduni.ac.nz. 2. Waikato Cardiothoracic Unit, Waikato Hospital, Hamilton, New Zealand; Faculty of Medical and Health Sciences, The University of Auckland, Auckland, New Zealand. 3. Waikato Cardiothoracic Unit, Waikato Hospital, Hamilton, New Zealand. 4. Waikato Cardiothoracic Unit, Waikato Hospital, Hamilton, New Zealand; The University of Waikato, Hamilton, New Zealand.
Abstract
BACKGROUND: The updated European System for Cardiac Operative Risk Evaluation (EuroSCORE II) is a well-established cardiac surgery risk scoring tool for estimating operative mortality. This risk stratification system was derived from a predominantly European patient cohort. No validation analysis of this risk model has been undertaken for the New Zealand population across all major cardiac surgery procedures. We aim to assess the efficacy (discrimination and calibration) of the EuroSCORE II for predicting mortality in cardiac surgical patients at a large New Zealand tertiary centre. METHODS: Data was prospectively collected on patients undergoing cardiac surgery from September 2014 to September 2017 at Waikato Hospital, New Zealand. Patient demographic information, preoperative clinical risk factors and outcome data were entered into a national database. Included patients received either isolated coronary artery bypass grafting (CABG), isolated valve surgery, isolated thoracic aortic surgery, or a combination of these procedures. The primary outcome was the discrimination and calibration of predicted EuroSCORE II risk scores compared with observed 30-day mortality events. RESULTS: 1,666 cardiac surgery patients were included during the study period, with an average EuroSCORE II of 2.97% (95% confidence interval (CI): 2.76-3.18). Nine hundred thirty-three (933) patients underwent isolated CABG, 384 underwent isolated valve surgery, 48 received isolated thoracic aortic surgery and 301 received combination procedures. Thirty-day (30-day) mortality events in each of these groups was 7, 4, 2 and 13 deaths respectively. There were 26 deaths across the total cohort at 30-days (observed mortality rate 1.56%). Discrimination analysis using receiver operating characteristic curves demonstrated the area under the curve (AUC) of the EuroSCORE II in each of these groups as 93.4% (95% CI: 91.6-94.9, p<0.0001), 66.3% (95% CI: 61.3-71.0, p=0.37), 37.0% (95% CI: 15.7-58.2, p=0.23) and 74.8% (95% CI: 69.5-79.6, p<0.0001) respectively. The total cohort AUC was 83.1% (95% CI: 81.2-84.9, p<0.0001). Calibration analysis using Hosmer-Lemeshow tests for the subgroups revealed p-values of 0.848, 0.114, 0.638 and 0.2 respectively. The total cohort Hosmer-Lemeshow p-value was 0.317. CONCLUSIONS: EuroSCORE II showed a strong discriminative ability for isolated CABG 30-day mortality in a New Zealand patient cohort. However, the scoring system discriminated poorly across valvular, thoracic aortic or complex combination cardiac surgical procedures. Good calibration of the EuroSCORE II was achieved across both the total cohort and subgroups. It is important to consider the performance of other cardiac surgery risk stratification models for the New Zealand population.
BACKGROUND: The updated European System for Cardiac Operative Risk Evaluation (EuroSCORE II) is a well-established cardiac surgery risk scoring tool for estimating operative mortality. This risk stratification system was derived from a predominantly European patient cohort. No validation analysis of this risk model has been undertaken for the New Zealand population across all major cardiac surgery procedures. We aim to assess the efficacy (discrimination and calibration) of the EuroSCORE II for predicting mortality in cardiac surgical patients at a large New Zealand tertiary centre. METHODS: Data was prospectively collected on patients undergoing cardiac surgery from September 2014 to September 2017 at Waikato Hospital, New Zealand. Patient demographic information, preoperative clinical risk factors and outcome data were entered into a national database. Included patients received either isolated coronary artery bypass grafting (CABG), isolated valve surgery, isolated thoracic aortic surgery, or a combination of these procedures. The primary outcome was the discrimination and calibration of predicted EuroSCORE II risk scores compared with observed 30-day mortality events. RESULTS: 1,666 cardiac surgery patients were included during the study period, with an average EuroSCORE II of 2.97% (95% confidence interval (CI): 2.76-3.18). Nine hundred thirty-three (933) patients underwent isolated CABG, 384 underwent isolated valve surgery, 48 received isolated thoracic aortic surgery and 301 received combination procedures. Thirty-day (30-day) mortality events in each of these groups was 7, 4, 2 and 13 deaths respectively. There were 26 deaths across the total cohort at 30-days (observed mortality rate 1.56%). Discrimination analysis using receiver operating characteristic curves demonstrated the area under the curve (AUC) of the EuroSCORE II in each of these groups as 93.4% (95% CI: 91.6-94.9, p<0.0001), 66.3% (95% CI: 61.3-71.0, p=0.37), 37.0% (95% CI: 15.7-58.2, p=0.23) and 74.8% (95% CI: 69.5-79.6, p<0.0001) respectively. The total cohort AUC was 83.1% (95% CI: 81.2-84.9, p<0.0001). Calibration analysis using Hosmer-Lemeshow tests for the subgroups revealed p-values of 0.848, 0.114, 0.638 and 0.2 respectively. The total cohort Hosmer-Lemeshow p-value was 0.317. CONCLUSIONS: EuroSCORE II showed a strong discriminative ability for isolated CABG 30-day mortality in a New Zealand patient cohort. However, the scoring system discriminated poorly across valvular, thoracic aortic or complex combination cardiac surgical procedures. Good calibration of the EuroSCORE II was achieved across both the total cohort and subgroups. It is important to consider the performance of other cardiac surgery risk stratification models for the New Zealand population.