Machine Learning Predictions of Alternate Level of Care (ALC) in Canada: From Emergency Department to the in-Hospital Stage

Abstract

In Canada, patients who occupy hospital beds but do not require that intensity of care are called Alternate Level of Care (ALC) patients. ALC has numerous negative implications on patient health and the health care system. Early identification of patients who are at risk of becoming ALC could help decision-makers better manage the situation and alleviate this problem. This thesis evaluates the use of various ML algorithms in predicting ALC at two different time points in the patient’s trajectory. Moreover, it identifies the most important predictors of ALC in each time point and provides insights on how adding more information, at the expense of time for decision-making, would improve the predictive accuracy.