Biomedical & Clinical Informatics Lab

Department of Computational Medicine & Bioinformatics

System for Continuous Monitoring and Prediction of Hospital Readmission

Hospital readmission prediction could improve the delivery of often resource-intensive interventions to patients. In addition, due to the significant costs associated with hospital visits, excessive hospital readmission rates have been a major burden on the healthcare economy. This financial burden can be alleviated using an accurate method for assessment / prediction of readmission. The main shortcomings and poor performance associated with the current readmission prediction systems could be attributed to the following factors: 1) assessments are made based only on data collected at hospital 2) monitoring / data collection does not include informative high-density waveform physiological data. The proposed research hypothesizes that a home-based smart monitoring of patients’ health, based on the analysis of all patient information including their physiological waveform may accurately predict readmission risk. The objective of this research is to design and test the computational methods for a smart monitoring system that monitors the person’s high-density physiological signals using a portable monitor to predict readmission. The study will be done retrospectively using a dataset of ICU patients monitored at home for 30 days after discharge. Although the system will be specialized towards patients discharged from ICU, the methods can be easily extended to monitoring of other readmission applications. The proposed system is to: I) create computational methods that form predictions based on all specific patient details, producing the readmission predictions by integrating the information in all patient data (e.g. continuously-monitored physiological signals, demographics, medical history, etc) collected at hospital and at home, II) explore the earliest time, within a window of 2 weeks to 3 days before readmission, in which the resulting model can accurately and reliably predict readmission, III) test the accuracy and reliability of this system retrospectively using the database discussed above and an extensive cross validation process.