Predicting Hypertension Status Using Mental and Behavioral Health Indicators

About half of the adult population in the United States has hypertension. If hypertension is left uncontrolled or untreated, high blood pressure can escalate to more severe conditions that can be costly to healthcare organizations. It would be beneficial to know if mental and behavioral health factors could predict hypertension to enable timely diagnosis and treatment in these patient populations. Using the 2023 National Health Interview Survey, a dataset of around 9,200 adults with about 40 variables related to anxiety, depression, and demographics was used to predict hypertension status. After evaluating multiple classification model types, feature groups, and tuned hyperparameters, a logistic regression model achieved an accuracy metric of 82 percent. However, the area under the curve (AUC) metric was only 58 percent, which indicated insufficient model performance. Despite published research showing mental and behavioral indicators such as depression and anxiety increase hypertension risk, the logistic regression model could not be used to draw operational or clinical conclusions. Therefore, the next step is to gather detailed and discrete data on hypertension in a healthcare organization’s Electronic Health Record (EHR) system. Then, the recommendation is to pursue model development with the EHR data as the data source to predict uncontrolled hypertension and research other mental and behavioral health factors (ADHD, OCD, etc.) as potential predictors of hypertension in adults.