Sir Mortimier B. Davis Jewish General Hospital

Published purpose

The urgency of the COVID-19 pandemic has triggered a surge in the development of algorithms that address forecasting, contact tracing, screening, and treatment of COVID-19 positive patients to aid clinicians in their decision-making. However, a common feature of this work is a concern that bias in the models may mean the results do not generalize well. Moreover, such models can present challenges to identifying the underlying functional mechanism by which predictions are made and, ultimately, when they do not perform to expectation. Taken together, these issues can present significant challenges to moving such methods into clinical practice. This Project will seek to demonstrate that user-centric development, when combined with explainable approaches to machine learning can result in greater uptake of new data-driven aids to care by clinicians. This will be done through development of best practices for introducing explainable machine learning (ML) into clinical informatics applications and a validation of these approaches in a prototype presentation layer using an explainable ML analysis that is relevant to clinical needs during pandemics as a test case.