Interactive Dimensionality Reduction for Improving Patient Adherence in Remote Health Monitoring

Remote Health Monitoring Systems (RHM) and Body Sensor Networks (BSN) are effective tools to monitor and improve health conditions by collecting continuous data from patients. Although Remote Health Monitoring systems have shown potential for improving the quality of care and reducing healthcare costs, research studies have shown that the low adherence of patients can significantly degrade the system efficacy. In RHMS, often times patients are required to answer numerous questions, which are necessary to identify health conditions associated with medical diagnoses or adverse events. In this paper, we propose a new framework for question selection in order to reduce patients' burden, and consequently increase patients' engagement and adherence. We envision a unit named Interactive Learning for Data Acquisition (ILDA) for on demand data acquisition in order to improve the quality of our prediction. The ILDA attempts to identify the best non-redundant question subset in order to improve the response rate. It is also responsible to identify what additional information should be requested from which subjects to improve the accuracy and certainty of prediction. The proposed framework has been tested and validated on a large dataset collected from 600 AIDS patients.