Machine Learning of Hospital Readmission after Heart Failure based on Wearable Devices

Hospitalization for symptom management among patients with heart failure is frequent, costly and negatively affects quality of life. Currently, heart failure related costs in the U.S. are estimated at nearly $31 billion annually and expected to rise to $70 billion by 2030. Interventions aimed at reducing readmission in the short term have been largely ineffective indicating the need to characterize readmission risk better. Further, there is a critical need to identify those at highest risk for poor outcomes. We will address this need by combining state-of-the-art wearable sensors and machine learning algorithms for physiological data analysis with the latest clinical understanding in the field. We will collect data on heart rate variability, physical activity and bio-behavioral factors in relation to clinical events among 10 adults with heart failure. Improved understanding of factors associated with heart failure admission risk will inform development of interventions to avert unplanned urgent and emergent care. Accurate risk stratification also will determine who needs more frequent clinical oversight versus targeting interventions to improve self-care skills supporting judicious use of healthcare resources.