Background: The American Heart Association's Life's Essential 8 (LE8) is an updated construct of cardiovascular health (CVH), including blood pressure, lipids, glucose, body mass index, nicotine exposure, diet, physical activity, and sleep health. It is challenging to simultaneously measure all eight metrics at multiple time points in most research and clinical settings, hindering the use of LE8 to assess individuals' overall CVH trajectories over time.
Materials and methods: We obtained data from 5,588 participants in the Nurses' Health Studies (NHS, NHSII) and Health Professionaĺs Follow-up Study (HPFS), and 27,194 participants in the 2005-2016 National Health and Nutrition Examination Survey (NHANES) with all eight metrics available. Individuals' overall cardiovascular health (CVH) was determined by LE8 score (0-100). CVH-related factors that are routinely collected in many settings (i.e., demographics, BMI, smoking, hypertension, hypercholesterolemia, and diabetes) were included as predictors in the base models of LE8 score, and subsequent models further included less frequently measured factors (i.e., physical activity, diet, blood pressure, and sleep health). Gradient boosting decision trees were trained with hyper-parameters tuned by cross-validations.
Results: The base models trained using NHS, NHSII, and HPFS had validated root mean squared errors (RMSEs) of 8.06 (internal) and 16.72 (external). Models with additional predictors further improved performance. Consistent results were observed in models trained using NHANES. The predicted CVH scores can generate consistent effect estimates in associational studies as the observed CVH scores.
Conclusions: CVH-related factors routinely measured in many settings can be used to accurately estimate individuals' overall CVH when LE8 metrics are incomplete.
Keywords: cardiovascular health; life’s essential 8; life’s simple 7; machine learning; prediction model.
© 2023 Zheng, Huang, Guasch-Ferre, Hart, Laden, Chavarro, Rimm, Coull and Hu.