Issue:
A major pharmaceutical client sought to better characterize the risk strata based on ejection fraction (EF) and associated risk drivers associated with U.S Medicare patients that had decompensated acute heart failure (AHF). Key outcomes of interest were re-admission rates and mortality risk.
Solution:
- We constructed a Classification and Regression Tree (CART) model which is a structured machine learning approach, as well as a logistic regression model to identify key variables predictive of either re-admission risk or mortality.
- The results of this analysis led to the creation of a predictive desktop tool that could be used by clinicians to prognostically predict at a patient-level risk adverse re-admission and mortality outcomes at various pre-identified risk strata. A decision was made by the client to not pursue deploying the tool in practice settings due to potential regulatory/legal issues.
- This research was published in CMRO. References available upon request.
