Predicting adults likely to develop heart failure using readily available clinical information An analysis of heart failure incidence using the NHEFS

Background: Heart failure is a heavy burden on the health care system in the United States. Once heart failure develops, the quality of life and longevity are dramatically affected. As such, its prevention is critical for the well-being of at risk patients. We evaluated the predictive ability of readily available clinical information to identify those likely to develop heart failure. Methods: We used a classification and regression tree (CART) model to determine the top predictors for heart failure incidence using the NHANES Epidemiologic Follow-up Study (NHEFS). The identified predictors were hypertension, diabetes, obesity, and myocardial infarction (MI). We evaluated the relationship between these variables and incident heart failure by the product-limit method and Cox models. All analyses incorporated the complex sample design to provide population estimates. Results: We analyzed data from 14,407 adults in the NHEFS. Participants with diabetes, MI, hypertension, or obesity had a higher incidence of heart failure than those without risk factors, with diabetes and MI being the most potent predictors. Individuals with multiple risk factors had a higher incidence of heart failure as well as a higher hazard ratio than those with just one risk factor. Combinations that included diabetes and MI had the highest incidence rates of heart failure per 1000 person years and the highest hazard ratios for incident heart failure. Conclusions: Having diabetes, MI, hypertension or obesity significantly increased the risk for incident heart failure, especially combinations including diabetes and MI. This suggests that individuals with these conditions, singly or in combination, should be prioritized in efforts to predict and prevent heart failure incidence.