Validity of self-reported cardiovascular disease risk from survey questions

Introduction: Simple questionnaires with dichotomous responses evaluating atherosclerotic risk factors are frequently used to assess cardiovascular disease (CVD) risk in research and clinical settings. Such methods are easier and less costly than direct measurements, but the validity of selfreported risk compared to calculated CVD risk is not well studied. Hypothesis: Self-reported history of hypertension, hyperlipidemia, diabetes and smoking will predict measured CVD risk. Methods: Using the ACC/AHA Pooled Cohort Risk Equations (PCE), we calculated 10 year CVD risk from a randomly selected population sample of 9,856 individuals aged 40-79 without a history of CVD in the Minnesota Heart Survey (MHS). Using log-linear regression models, we estimated sexspecific PCE risk from the individual's self-reported history of 4 dichotomous risk factors; hypertension, hypercholesterolemia, diabetes, and smoking. Age was included in all models. Model performance was assessed internally using leave-one-out cross-validation. Results: The median PCE 10 year CVD risk in women was 2.1% (IQR: 0.8%-5.6%), and in men was 6.3% (3.1%-13.0%). Using the newly developed equations, which included only the 4 selfreported risk factors and age, the estimated median risk was 2.2% (0.9%-5.8%) in women, and 6.9% (3.2%-13.1%) in men. The measured and estimated PCE risk using a threshold of 7.5% to categorize low and high risk gave an accuracy of 95% for women and 87% for men. The multivariable model-estimated PCE corresponded closely with the actual PCE (Figure). Conclusions: Self-reported history of atherosclerotic risk factors may be useful to estimate CVD risk in individuals, avoiding actual measurements.