Using Machine Learning Models to Study Medication Adherence in Hypertensive Patients Based on National Stroke Screening Data

Abstract

Stroke, with high incidence, prevalence and mortality, has brought a heavy burden to families as well as society in China nowadays. In 2009, the China national stroke screening and intervention program was launched by the Ministry of Health of China. In the national program, risk factors of stroke are screened and people aged over 40 with high-risk of stroke will be followed-up. From the experience, it is found that hypertension is an important risk factor of stroke. Improving the adherence of hypertension medication can effectively control blood pressure and further decrease stroke incidence. In this study, firstly, we employ oversampling and undersampling method to process the imbalanced dataset. Then, we build four machine learning models, namely logistic regression model, decision tree model, neural network model and random forest model, to classify medication adherence in hypertensive patients. We use the recall and precision to evaluate these models, and considering these two criteria, the model based on decision tree achieves best performance. The models constructed in this paper can be used to identify people with low adherence of antihypertensive drugs in the stroke screening program and improve the efficiency of the follow-up interventions, which can effectively control blood pressure and reduce the possibility of stroke.