Back pain is the most frequent cause of disability and inability to work, and it can result in dramatic social/economic/family costs, with an estimated U.S. prevalence of 27%. Prescribed medications are considered a first-line treatment to relieve back pain. Any drug may cause and adverse outcomes of adverse events (i.e., overdose, dependence, addiction) and side effects (e.g., sedation and respiratory depression). Multiple uses of drugs will lead to multiple adverse outcomes due to the interactions among them. Here, we propose to predict multiple adverse outcomes using the information on the first day from encounter admission time for each back pain transition. Our empirical results showed promising predictive power for each combination with the need to follow more advanced analytics approaches.
Learning Objective: Learn the challenges to model early prediction for multiple adverse outcomes in acute and chronic back pain transitions with different feature settings
Samir Abdelrahman (Presenter)
University of Utah
Michael Newman, University of Utah
Christina Porucznik, University of Utah