A0559
Title: A personalized remote patient monitoring system using daily measurements of bodyweight, heart rate, and blood pressure
Authors: Mehran Moazeni - Utrecht Unveristy (Netherlands) [presenting]
Abstract: Mortality rates and readmissions are prohibitively high for heart failure patients. These events are preceded by a period in which either one or a combination of bodyweight, heart rate, and blood pressure shift from a healthy baseline. This preceding shift offers an opportunity to early detect heart failure. Facilitating early detection of changes in biometric values, remote patient monitoring systems have been developed to record biometric values. Previously, simple algorithms were introduced to distinguish normal biometric observations from observations signalling heart failure by using absolute thresholds for all patients, rule-of-thumb and a moving average convergence-divergence algorithm. However, these algorithms have a poor performance in detecting heart failure as they display a high rate of false alarms. To alleviate this, we propose a novel personalized algorithm for two settings: single and combined biometric measurement monitoring. The algorithm is informed by cross-sectional and longitudinal data and uses a linear mixed-effect model to predict a personalised expected biometric value. Then, differences between the expected and observed value are used for a statistical process control chart, providing patient-specific thresholds for determining alarms. Comparing area-under-the-curve values showed that our personalised algorithm outperforms the above-mentioned algorithms for both settings. We discuss further improvements that could be incorporated to the algorithm
