An ordinal prediction model of the diagnosis of non-obstructive coronary artery and multi-vessel disease in the CARDIIGAN cohort
Background: The extent of coronary artery disease (CAD) is relevant for the evaluation and the choice of treatment of patients and consists of the severity of stenoses and their distribution within the coronary tree. Diagnosis is not easy and severe CAD should not be missed. For low-risk patients one wants to avoid the invasive angiography. We aim to propose a diagnostic prediction model of CAD respecting the degree of disease severity. Methods: We included 4888 patients from the Coronary Artery disease Risk Determination In Innsbruck by diaGnostic ANgiography (CARDIIGAN) cohort. An ordinal regression model was applied to estimate the probabilities of five incrementally disease categories: no CAD, non-obstructive stenosis, and one-, two- and three-vessel disease. We included 11 predictors in the model: age, sex, chest pain, diabetes, hypertension, dyslipidaemia, smoking, HDL and LDL cholesterol, fibrinogen, and C-reactive protein. Bootstrapping was used to validate model performance (discrimination and calibration). Results: Age, sex, and three laboratory measures had a large predictive effect. The model poorly separated most adjacent disease categories, but performed well for categories far apart, with little optimism. The overall discrimination added up to a c statistic of 0.71 (95% CI 0.69 to 0.73). The model enables the estimation of individual patient probabilities of disease severity categories. Conclusions: The proposed ordinal diagnostic risk model, employing routinely obtainable variables, allows distinguishing the extent of CAD and can especially discriminate between non-obstructive stenosis and multi-vessel disease in our CARDIIGAN patients. This can help to decide on treatment strategy and thereby reduce the number of unnecessary angiographies.