(joint work with Randall P. Ellis, Karen Lasser, Heather Hsu, Corinne Andriola, Tzu-Chun Kuo, Jeffrey Siracuse, Allan Walkey, & Arlene Ash). In this working paper, we build on the DXI framework discussed in our other working paper by exploring the possibility of nonlinear interactions in risk adjustment models which leverage within-disease variation in aspects such as disease severity, location, and timing. We propose parsimonious and easily estimable approaches to integrate these details even with large data sets, and show that doing so improves model performance significantly. These improvements have important implications for the equity and fairness of risk adjustment models as well as for provider payments.
The full abstract/text of this paper is not available as it is currently prepared for submission at clinical journals that do not allow online circulation of working papers. Look for more detail about this paper coming soon.