Author(s)

Yue Yu

Corresponding Author

Yue Yu

ABSTRACT

Background: As a chronic disease, cardiovascular disease (CVD) has greatly affected people's quality of life. Limited studies have investigated the relationship between many risk factors and infecting CVD, and few of them use a large sample size to conduct the study. We investigated the relationship between several different risk factors, such as blood pressure and cholesterol, and CVD. Methods: A dataset which contains 70000 samples is obtained from Kaggle, and binary logistic regression analysis was used to investigate the impact of clinical and lifestyle factors on the prevalence of CVD. Variable selection, such as forward selection and backward selection, was performed to build the final model. Results: The result of forward-backward selection and backward selection show that all features, except gender, are significant under 0.1% level, and the AIC of them are both 77216. The percentage of accuracy of the model is 0.7267. The proportion of actual positive samples identified correctly (recall/sensitivity) is 0.663, the proportion of predict positive samples identified correctly (precision) is 0.755, and the proportion of actual negative samples identified correctly (specificity) is 0.789. Conclusion: Our study suggested that age, height, weight, systolic blood pressure, diastolic blood pressure, cholesterol, glucose, smoke, alcohol intake, and physical activity are the most important factor in determining CVD. The findings from our study have important public health implications and call for future studies to explore the potential mechanism of these findings. CCS Concept•Mathematics of computing  Probability and statistics  Statistical paradigms  Regression analysis

KEYWORDS

non-communicable diseases, cardiovascular disease, logistic regression, binary logistic regression