The Risk Factors of Hypertension and Their Predictive Power in Identifying Patients Using a Decision Tree

Abstract

Hypertension (HTN) is the most important controllable risk factor for non-communicable diseases that can have various causes, which vary in different subgroups. This secondary analysis was conducted using the data obtained through the recruitment phase of Ravansar non-communicable cohort study (RaNCD). The multivariable logistic regression was used to determine the risk factors of HTN, and a decision tree with the CART algorithm was used to determine the predictive power of these variables. Of the 10,046 individuals aged 35 to 65 participating in RaNCD, 1579 (15.72%) of the participants had HTN. Aging and diabetes were the most important risk factors of HTN. The sensitivity and specificity of the decision tree for the training and testing models were very similar, such that the sensitivity of training was 69.0% and testing 68.0%, and their specificity was 73.0% and 71.0%, respectively. Overall, the accuracy rate of the training and testing models was 70% and 68%, respectively. The variable that best discriminated people with HTN from non-HTN was diabetes. In people with diabetes, the incidence of HTN was 5 years higher than those without diabetes. Since the predictive power and effect of the risk factors of HTN vary from one group to another, the decision tree can be of great help in identifying people with HTN due to the latent nature of the disease.