An mHealth Lifestyle Intervention Service for Improving Blood Pressure using Machine Learning and IoMTs

Abstract

In this paper, we present an AI-driven lifestyle intervention service for patients with hypertension. The automated intervention platform consists of a remote monitoring system that ingests lifestyle and blood pressure (BP) data and builds a personalized machine learning (ML) model to generate tailored lifestyle recommendations most relevant to each patient's BP. Lifestyle data is collected from a wearable device and questionnaire mobile app which includes activity, sleep, stress and diet information. BP data is remotely collected using at-home BP monitors. With this data, the system trains random forest models that predict BP from lifestyle features and uses Shapley Value analysis to estimate the impact of features on BP. Precise lifestyle recommendations are generated based on the top lifestyle factors for each patient. To test the system's ability to improve BP, we enrolled hypertensive patients into a three-armed clinical trial. During the 6-month trial period, our system provided weekly recommendations to patients in the experimental group. We evaluate the system's effectiveness based on multiple BP improvement metrics and comparison with a control group. Patients in the experimental group experienced an average BP change of −4.0 and −4.7 mmHg for systolic and diastolic BP, respectively, compared to −0.3 and −0.9 mmHg for the control group. Our results demonstrate that the platform can effectively help patients improve their BP through precise lifestyle recommendations.