Background: Heart failure (HF) costs the US an estimated $30.7 billion annually, of which approximately 80% is associated with hospitalization. Reducing this cost by lowering hospitalization rates through in-home patient monitoring is appealing. Adherence is a major challenge to in-home heart monitoring systems, especially when most high-risk patients are elderly and sick.
Objective: We use a novel toilet seat that is an inconspicuous monitoring system with potential for daily use. The seat has integrated sensors for multiple signals including ECG, PPG, BCG, body weight. We examine the initial deployment success and adherence rate to this new technology.
Method: Heart failure patients discharged from hospitalization are enrolled for a 90-day monitoring study using the seat. For deployment success rate, we examine successful enrollment rate and causes for failed enrollment. For adherence, we examine data missingness. For predictive analysis, we train a random forest classifier in a five-fold cross validation for a monitoring window of 7 days to predict adverse events within 21 days.
Results: 139 HF patients consented to participate over a period of a year, of which 49 subjects (35.5%) were enrolled. The main causes for failure to enroll were: hardware compatibility (43%), further illness (20%), loss of follow-up (17%), lack of interest (10%). Out of 45 subjects that were analyzed, we observed the following averaged number of days of monitoring data: 71±36 days of heart rate (hr), heart rate variability (hrv) and weight(sw); 64±31 days of QRS duration and QT corrected interval; 47±40 days of SPO2; 37±41 days of blood pressure. Initial predictive analysis using hr, hrv, and sw, resulted in an AUC of 0.74 (sensitivity 57, specificity 75) in a five-fold cross validated dataset.