Predicting Heart Failure Recovery by Wearables and Machine Learning

What this Trial Is About

Researchers are studying how data collected from the Apple Watch can help predict the health status of patients with decompensated heart failure with reduced ejection fraction (HFrEF). This observational study focuses on using machine learning methods to analyze physiological data, such as single-lead electrocardiograms, oxygen levels, respiratory rate, step count, and nighttime temperature. The goal is to forecast risks like heart failure worsening and rehospitalization while also correlating wearable data with clinical parameters collected during guideline-compliant treatment. Participants will receive an Apple Watch to wear during their hospital stay, with data collected continuously. The watch will be given to patients upon admission and collected at discharge for offline analysis. Clinical data including blood tests, echocardiography, medication changes, and physical assessments will also be gathered at admission, discharge, and a follow-up visit 90 days after discharge. The study will use this combined data to develop and evaluate predictive models of heart failure outcomes. During the study, participants will be monitored through physiological data from the watch and clinical assessments including laboratory tests and echocardiography. The Kansas City Cardiomyopathy Questionnaire, NT-proBNP levels, and other blood markers will be measured at admission, discharge, and the 3-month follow-up. Researchers will assess rehospitalization rates and treatment adherence, using statistical methods to evaluate the accuracy of predictive models. The entire observation period spans from hospital admission through a 90-day post-discharge follow-up.

Prediction of Heart-Failure with Machine Learning