Does the 2 year Framingham coronary heart disease risk score apply to a nationally representative sample? Evidence using the Health and Retirement Study

Background: It is not well known how well Framingham Heart Study (FHS) 2-year coronary heart disease (CHD) risk score predicts events in a nationally representative sample. We calculate the conventional FHS 2-year CHD score and also a modified score using only self report items in the U.S.-based Health and Retirement Study (HRS). Methods: CHD and stroke-free participants in HRS aged 50-74 and part of the 2006 biomarker collection (n=2,619), were followed until 2008 for proxy or self-report of diagnosis or fatal heart attack, coronary heart disease, angina, or other heart problems (90 events). The self-report score excluded cholesterol and included self-report of hypertension prevalence and treatment. Logistic regression was used to predict events and Loess smoothing to produce predictions for specific scores. Results: Risk scores calculated with either biomarker (OR=1.09, 95% CI: 1.05-1.13) or self-report (OR=1.07, 95% CI: 1.03-1.11) data were significant predictors of events. The HRS sample had higher probabilities of events at lower scores (<18) and lower probabilities at higher scores than the FHS sample. For example, among women, an FHS score of 10 had a 1% probability of event in FHS sample and 3% in HRS sample; a score of 22 had a 16% probability of event in FHS sample and 6% in HRS sample. Conclusions: FHS scores strongly predict new onset of CHD in a national representative sample. Probabilities of events associated with FHS risk scores diverge from those in the original FHS sample. Self-report and biomarker based scores had similar predictive performance.