School of Public Health and Health Systems, University of Waterloo
Personalized predictive analytics based on electronic medical data and patient similarity metrics
As hospitals and doctors’ offices in Canada rapidly adopt electronic medical records (EMRs), the enormous clinical value of ever-increasing EMR data is receiving the spotlight. In particular, massive EMR data can facilitate personalized medical treatment through identification and analysis of past patients who are similar to a current patient.
To seize this opportunity, the proposed research aims to quantitatively define patient similarity and apply the patient similarity definitions to predict the future state of a given patient in a personalized manner. Two public EMR datasets, containing a total of 160,000 admissions to intensive care units, will be analyzed in this research.
The ultimate objective is to show that the proposed personalized data-driven approach leads to better predictive performance than traditional methods that treat each unique patient like the average patient. The outcome of this research will provide important scientific evidence for the emerging field of medical data analytics.