Development and Validation of the CR-DECIDE Models to Predict Major Adverse Cardiovascular Events and Health Status in Stable Coronary Artery Disease

Background: Guidelines emphasize individualized care in the management of stable coronary artery disease (CAD). We aimed to develop and validate clinical prediction models for major adverse cardiovascular events (MACEs) and health status among patients with stable CAD to support individualized, shared decision-making. Methods: For model development and internal validation, we used registries of outpatients with obstructive CAD on coronary angiography in British Columbia (2004-2015) and Alberta (2004-2020). Models were externally validated in ISCHEMIA trial participants with obstructive CAD on coronary computed tomography angiography. Outcomes included MACE (death, myocardial infarction, or stroke) within 3 years, angina-free status, and good-to-excellent physical functioning at 1 year, based on the Seattle Angina Questionnaire. Results: Median age was of study patients was 66-67 years, and 77% were male in both the MACE (n = 34,990) and health status (n = 13,312) model development cohorts. MACEs occurred in 9% (2026 patients) at 3 years. A 14-variable model had a C statistic of 0.68, calibration slope of 0.98, and positive net benefit in decision-curve analysis. At baseline, 41% were angina-free and 21% had good-to-excellent physical functioning, which increased to 64.5% and 72% at 1 year, respectively. C statistics for the angina-free and physical functioning models were 0.67 and 0.78, respectively, and calibration slopes were 0.98-0.99. In external validation, discrimination was modestly reduced and all models slightly underpredicted their respective outcomes, yet the MACE model retained positive net benefit. Conclusions: The CR-DECIDE models had moderate ability to predict MACEs and health status in patients with stable CAD and warrant further assessment of their impact at the point of care.