Development of a claims-based EuroSCORE II in patients with aortic stenosis needing surgical or transcatheter aortic valve replacement using electronic hospital records: a nationwide study

Background Prediction of operative risk in patients with aortic stenosis (AS) undergoing surgical aortic valve replacement (SAVR) or transcatheter aortic valve implantation (TAVI) remains a challenge, particularly in high-risk patients. The EuroSCORE II is now commonly used to improve risk prediction. Large analyses from administrative database have provided opportunities for conducting health research in the field of structural heart disease interventions but may have a lack of granularity and do not routinely include EuroSCORE II, which may result in a risk of uncontrolled biases. We sought to approximate the EuroSCORE II using only administrative claims data to enable the operative risk to be assessed without clinical or paraclinical performance measures. Methods Based on the administrative hospital-discharge database, we collected information for all patients with AS treated with SAVR or TAVI between 2010 and 2019 in France. A total of 78,085 SAVR and 60,821 patients with AS treated with transcatheter aortic valve replacement (TAVR) were found in the database. For each patient, the EuroSCORE II was estimated using the formulas available at the EuroSCORE website. Age, gender, extracardiac arteriopathy, poor mobility, previous cardiac surgery, chronic lung disease, active endocarditis, diabetes on insulin, recent MI, dialysis are items available in the PMSI database using the ICD-10 or CCAM codes. For renal impairment, NYHA class, LVEF, pulmonary hypertension, “critical preoperative state” and urgent intervention, different proxies were built based on ICD-10 codes likely to represent increasing severity of these items. Results In the cohort of patients with SAVR, mean estimated EuroSCORE II was 3.3±1.1 while all-cause death at day 30 after SAVR was 3.8%. In the cohort of patients with TAVI, mean estimated EuroSCORE II was 3.8±1.0 while all-cause death at day 30 after TAVI was 5.5%. In the whole cohort, the area under the curve (AUC) of the estimated EuroSCORE II for predicting the risk of all-cause death at day 30 was 0.72 (95% CI 0.71–0.73) and was higher in patients treated with SAVR (AUC 0.76, 95% CI 0.75–0.77) than in those treated with TAVI (AUC 0.67, 95% CI 0.65–0.68, p<0.00001 for DeLong test). The observed versus predicted risks of all-cause death at day 30 post-TAVI OR SAVR within risk deciles are shown in Figure 1. Calibration of the prediction score was satisfying across the 10 deciles and a predicted 30-day mortality rate of approximately 15%. Conclusions Claims data alone can be used to identify individuals with AS at operative risk when they are considered for SAVR or TAVI. The Claims-based EuroSCORE II might be used in research with large datasets for confounding adjustment or risk prediction. It provides hospitals and health systems with a low-cost, systematic way to identify a group of patients who are at greater risk of adverse outcomes with these interventions and for whom a more specific approach might be useful.