Integration of a learning prostate cancer risk prediction tool into the electronic medical record

Project ID: 930

Cube proposed by: Gregory Auffenberg

Unit: Medicine

Cycle: MCubed 2.0

Cube Accomplishments Click here for more details
3 accomplishments!
About this project: 
Prostate cancer will account for 26% of new cancer diagnoses in American men in 2015. Most diagnoses are made with a biopsy performed after consideration of a patient’s perceived risk based on demographics, family history, physical exam, and lab testing. Existing risk prediction models are inadequate because they are fixed, poorly calibrated, and do not account for differences in practice patterns. Additionally, modern electronic health records require statistical models to be hard-coded in ways that make model updates extremely cumbersome. We aim to address the grand challenge of improving risk stratification through the following goals: 1) develop a new statistical learning model that continuously calibrates to achieve better risk prediction; 2) build a learning health platform to contain and continuously update the model; and 3) integrate the learning health platform with the electronic health record to provide this information seamlessly to clinicians at the point of care.