Quantifying Dynamic Risk with Predictive Analytics
Predictive analytics enables innovative approaches for leveraging real-time clinical data to assess risk. Predictive methodologies and applications that use dynamic patient information to update health risk assessments over time are especially valuable. For this purpose, two healthcare applications have been developed that assess either risk of hospital readmission or probability of elective, surgical procedures. Both techniques use electronic health record information as a source of features for predictive modeling. These applications are rooted in health system operations and may provide innovative value for monitoring and managing clinical operation risk.
This paper provides readers with a walkthrough for using predictive analytics to assess risk by describing the general workflow and providing two real-world examples. The predictive analytics methods discussed have broad applicability to numerous applications and industries. The authors intend to present a stepping off point for readers looking to use predictive analytics to assess and manage clinical operation risk in a forward-looking way. While these examples are not specific to clinical trial execution, they provide process templates for risk identification in a dynamic data environment. To describe the scope of the paper, the first portion of the paper called “Predictive Modeling Workflow” provides a primer from data selection to predictive modeling. The second portion of the paper, which is broken into two sections called “Healthcare Application 1” and “Healthcare Application 2,” provides two detailed, practical examples of predictive modeling related to risk management that the authors encountered in a real-world, healthcare setting.