Predictive ability of the SCORE Belgium risk chart for cardiovascular mortality

Background: The identification of persons at high cardiovascular risk is of primary importance in the context of cardiovascular prevention. Accuracy and precision of risk assessment are essential properties. We developed a calibrated SCORE risk chart (Systematic COronary Risk Evaluation) for calculating the absolute 10-year probability of developing a fatal cardiovascular event, adapted to national mortality statistics and risk factor distributions in Belgium and critically evaluated its predictive accuracy. Methods: Our SCORE Belgium risk chart was validated using data from 6212 non-diabetic men and women free of CHD participating in a prospective cohort study carried out in the eighties (Belgian Interuniversity Research on Nutrition and Health). Agreement between numbers of predicted and observed CVD deaths across the entire spread of risk was studied using chi-square and Hosmer-Lemeshow statistics. Discrimatory power of risk estimates was evaluated according to Harrell's c-statistic. Results: During the period of 10 years, 274 CVD deaths were observed while the recalibrated risk chart predicted 263 events. The SCORE Belgium risk chart showed very good accuracy over the complete range of predicted risk (Hosmer-Lemeshow: P=0.14). ROC analysis revealed excellent discriminatory power in labelling future cases of fatal cardiovascular disease with a c-statistic of 0.86. The 5% threshold for the probability of 10-year cardiovascular death yielded an optimal balance of sensitivity and specificity. Conclusions: The SCORE Belgium risk chart proves to be well suited as an accurate and precise estimation tool for the assessment of cardiovascular risk in Belgium. (C) 2009 Elsevier Ireland Ltd. All rights reserved.