In Frontiers in cardiovascular medicine
To explore the impact and risk of short sleep duration (sleep duration < 6 h/night) on new-onset cardiovascular and cerebrovascular diseases (CVDs) in people with metabolic syndromes (Mets), this study used the 2011 baseline and 2015 follow-up data from the China Longitudinal Study of Health and Retirement (CHARLS) to conduct a prospective study of people aged ≥ 45 years in China. A total of 5,530 individuals without pre-existing CVDs in baseline were included. Mets were defined according to the harmonized criteria. We applied the Logistic Regression (LR), the Deep Neural Networks (DNN), and the Adaptive Boosting (AdaBoost), to evaluate the association between Mets components, short sleep, and the risk of new-onset CVDs, and the importance of multiple variates for new-onset CVDs. During the 4-year follow-up period, 512 individuals developed CVDs, and short sleep increased the risk of CVD in individuals with Mets. The odds ratio for prevalent CVD in Mets with short sleep group was 3.73 (95%CI 2.95-4.71; P < 0.001) compared to the normal group, and 1.99 (95% CI 1.58-2.51; P < 0.001) compared to the Mets without short sleep group. The DNN method reached the highest precision of 92.24% and f1-score of 95.86%, and the Adaboost method reached the highest recall of 99.92%. Both DNN and Adaboost have better predictive performance than LR and revealed short sleep duration and components of Mets are all the strongest predictors of CVD onset.
Sun Jiaxin, Chen Yizhou, Sun Yazhou, Yang Bo, Zhou Jining
2022
cardiovascular and cerebrovascular diseases, machine learning, metabolic syndromes, risk prediction, short sleep duration
