Description
Identifying clinically stable patients who are likely to be discharged within a few days allows more time for planning. Machine learning algorithms can use existing EHR data to predict whether a patient will be sufficiently stable for discharge within 24 to 36 hours. This pilot study was designed evaluate four classification algorithms for best performance in predicting patient stability within 36 hours. A random forest model showed better performance than the competitors with positive predictive value of 48%
Learning Objective: Selection of the best-performing algorithm for automated prediction of patient discharge within 36 hours.
Authors:
Kevin Coppa (Presenter)
Northwell Health
Marsha Myetlis, Northwell Health
Jamie Hirsch, Northwell Health
Jan Horsky, Northwell Health
Presentation Materials: