(VP035) DEVELOPMENT AND INTERNAL VALIDATION OF THE CORONARY REVASCULARIZATION-TOOL FOR EVIDENCE-BASED INDIVIDUALIZED SHARED DECISION-MAKING (CR-DECIDE) QUALITY OF LIFE PREDICTION MODEL

Background: Coronary revascularization guidelines emphasize shared decision-making (SDM) for chronic coronary artery disease (CAD) management. Decision aids that provide individualized estimates of potential benefits and harms could improve the relevance and evidence base of SDM. This study aimed to develop and internally validate a model to predict quality of life (QoL) 1 year after invasive coronary angiography in patients with chronic CAD for integration into a point-of-care decision aid.

METHODS AND RESULTS: We used the prospective Alberta Provincial Project for Outcome Assessment in Coronary Heart Disease registry from February 2004-2020 to assemble a cohort of adults with obstructive CAD identified during invasive coronary angiography indicated for stable angina. The minimum required sample size was 2746. Participants were invited to complete the 19-item Seattle Angina Questionnaire (SAQ) at baseline (≤30 days post-angiography) and 1 year post-angiography.

Model 1 predicted angina-free (SAQ-Angina Frequency score 100/100) status at 1 year. Model 2 predicted good-to-excellent physical functioning (SAQ-Physical Limitations score ≥75/100) at 1 year. We selected and retained 18 candidate variables based on clinical expertise, prior literature, and availability in the development database. We also considered interactions between diabetes, number of diseased vessels, and treatment received; however, models with interaction terms performed worse than models without interaction terms (likelihood ratio test p< 0.01), so interactions were removed from final models. Missing data were handled by multiple imputation. Continuous variables were fit with restricted cubic splines with up to 3 knots. Models were developed using logistic regression including all candidate predictors without elimination. Internal validation was performed (1000 bootstraps) to generate optimism-corrected estimates of discrimination (c-statistic), calibration curves, and overall model performance (Cox-Snell R2 and Brier scores).

We included 13,091 participants (median age 66, 77% male, 22% underwent percutaneous coronary intervention, 19% underwent coronary artery bypass graft surgery). Responses to SAQ sub-scores were available for 69-74% at baseline and 44-46% at 1 year. Of these patients, 41% were angina-free and 55% had good-to-excellent physical functioning at baseline, which improved to 65% and 72% at 1 year. In both models, the single strongest predictor of 1-year QoL was baseline QoL. For model 1, c-statistic=0.67, calibration slope=0.983, Cox-Snell R2=0.14, and Brier score=0.211. For model 2, c-statistic=0.77, calibration slope=0.988, Cox-Snell R2=0.256, and Brier score=0.159.

Conclusion: The CR-DECIDE QoL models had moderate ability to predict QoL 1 year after angiography in patients with chronic CAD, and warrant further validation in an external dataset and impact analysis as part of a point-of-care decision aid.