Life’s Simple 7 (LS7) developed by the American Heart Association (AHA) is a new index of cardiovascular health (CVH). LS7 is comprised of 7 metrics, including 3 health factors (blood pressure [BP], total cholesterol, and glucose) and 4 health behaviors (body mass index [BMI], cigarette smoking, diet, and physical activity). To date, CVH estimates are mainly obtained from national surveys, clinical trials, and cohort studies. Real-world data (RWD) such as electronic health records (EHRs) and claims data and real-world evidence generated from these data are playing an increasing role. However, the lack of information on all 7 CVH metrics in RWD limits the use of the CVH concept in research and preventions based on RWD. Using data from the 1999-2016 National Health and Nutrition Examination Survey (NHANES), we developed predictive models for CVH among adults using 3 metrics (i.e. BP, BMI, and smoking) and sociodemographic factors (i.e. age, gender, race/ethnicity, education, and marital status) which are widely available in RWD. Each CVH metric was categorized into ideal (2 points), intermediate (1 point), or poor (0 point), and then weighted accordingly following LS7 to generate an overall CVH score (0-14 points) with a higher score indicating better CVH. Individuals with more than 4 ideal CVH metrics were determined as having ideal CVH. In addition, we also developed models using 4 CVH metrics (i.e. BP, BMI, and smoking + one of the other 4 metrics). The data were randomly divided into training (80%) and testing (20%) sets. Gradient boosting decision trees models were trained using the CatBoost library with hyper-parameters tuned by a grid search based on 5-fold cross validations. A total of 45,614 individuals aged 18 years and older in 1999-2016 NHANES were included. The models with 3 CVH metrics (i.e. BP, BMI, and smoking) as predictors achieved a test-AUC of 0.95 and a test-RMSE of 1.39. Including one of the other 4 CVH metrics (i.e. total cholesterol, glucose, diet, or physical activity) as a predictor in the models along with the previous 3 metrics (i.e. BP, BMI, and smoking) further improved the predictive performance (test-AUC>0.96 and test-RMSE<1.38). These findings suggested that the 3 CVH metrics (i.e. BP, BMI, and smoking) that are widely available in RWD can be used to accurately estimate CVH among adults in the United States.
Mechanisms for Online Organ Matching
