Ten percent of admissions to Australian acute-care hospitals are associated with harm to patients or adverse events. The reporting of critical incidents by health professionals is now well established and the rate of reporting continues to increase worldwide. Current methods, which rely on retrospective manual review of incident reports, do not permit timely detection of safety problems and can no longer keep up with this growing volume of data. In the New South Wales public hospital system alone, more than 200,000 patient-safety incidents were reported in 2021.

Dr Wang will present her work on developing AI-based tools to identify incident reports using text classification. Her studies have identified the type and risk of incident reports and generalised to 10 common incident classes and 4 severity levels. She will also share the recent experience to improve incident classification performance on unbalanced data sets including rare event types.

As a parallel line of her research, she aims to integrate AI-based tools in clinical workflows and investigate their impact on improving incident management processes. Due to the black-box nature of AI models developed using deep learning methods, their internal functions are not directly understandable by humans. The lack of explainability continues to be an obstacle in real-world deployment of AI tools. Dr wang will share their recent insights on XAI technique assessment to overcome the last mile hurdles of AI delivery in incident management.

Speaker bio

Dr Ying Wang is a research fellow at Centre for Health Informatics, Australia Institute of Health Innovation, Macquarie University.  She obtained her PhD in pattern recognition and artificial intelligence.

Ying is passionate about developing and explaining advanced machine learning algorithms to support decision making at clinical and diagnostic levels in health. She currently leads a project to examine clinical safety events and identify risk levels, and hence to provide interpretations of AI decisions to patient safety experts.  She is responsible for the development of a platform to monitor AI associated safety risks in medical devices. Her interests also involve the development of safe usage of AI in clinical practice, with a particular emphasis on evaluating model transportability in real world and explainable AI in health.