THE ABILITY OF THE ‘SCORE’ RISK MODEL TO PREDICT 10-YEAR CARDIOVASCULAR DISEASE MORTALITY: CANADIAN HEART HEALTH SURVEYS (CHHS) FOLLOW-UP STUDY AS AN EXAMPLE

Abstract

The purpose of this study is to evaluate the predictive accuracy of the SCORE (Systematic COronary Risk Evaluation) risk prediction model to predict the risk of 10-year cardiovascular disease (CVD) mortality in Canadian Heart Health Surveys (CHHS) Follow-up Study population. The study used data from The Canadian Heart Health Surveys (CHHS) Follow-up Study which is a linked dataset of Canadian Heart Health Surveys (1986-1992), a baseline cross-sectional complex surveys and Canadian Mortality Database by Statistics Canada. SCORE CVD risk prediction models for both high- and low-risk regions were used to calculate the predicted deaths from cardiovascular disease over a period of 10-year for both men and women aged 40-69 years with no previous history of cardiovascular disease. The predicted number of deaths was compared with observed CVD deaths (stratified by age and sex) in the CHHS Follow-up Study population. . Over a period of 10-year, the observed absolute risk of cardiovascular death in CHHS population was found to be 1.9% (95% CI: 1.87, 1.92) in male and 0.9% (95% CI: 0.87, 0.90) in female. By using SCORE high-risk model, the observed vs. expected death (O/E) ratio was found to be 0.54, 0.34, 0.49 in men and 0.55, 0.52, 0.50 in women respectively for the age group 40-49, 50-59 and 60-69 years. The SCORE low- model predicted O/E ratio of 1.06, 0.63, 0.87 in men and 0.88, 0.80, 0.75 in women respectively for age group 40-49, 50-59, and 60-69 years. Twenty-eight percent of males and seven percent of females were identified to have a risk of 5% or more of having a fatal CVD event within next 10 years by the SCORE high-risk model and the same was found to be in 12% of males and 5% of females respectively by the SCORE low-risk model. Application of the original SCORE models for high- and low-risk regions in Canadian Heart Health Survey Follow-up Study data has resulted in over prediction of the number of CVD deaths. SCORE low-risk model performed better than the high-risk version. Recalibration of SCORE risk prediction model is suggested if it is considered to be used in Canada.