Description
In the context of high-risk, nonspecific conditions like pediatric sepsis, using AI as a clinical decision-support tool may be challenging due to high variability and uncertainty in clinical sings and symptoms. Globally, healthcare systems lack consistent standardized rules and procedures for pediatric sepsis, hindering AI performance and introducing ethical issues if relied on uncritically. An ecological approach to system design that guides both clinicians and AI in decision-making is crucial. This approach would require an AI to learn the clinical boundaries of a child's health and predict deviations from a baseline state, and for clinicians to easily recognize the socio-technical system constraints within which AI predictions are being made, to understand how and when to appropriately rely on them. This doctoral thesis project focuses on the domain of pediatric sepsis because of its incomplete diagnostic definition which leads to inconsistency in clinical applications and treatment across institutions. These inconsistencies provide an ideal context to investigate the advantages of the proposed constraint-based approach to system design in support of human-AI teaming. The outcomes of this work will be (1) an overview of the current literature on pediatric sepsis prediction technologies in healthcare, (2) an improved understanding of how clinicians conceptualize uncertainty in pediatric sepsis decision-making and the use of AI predictions, (3) a comprehensive Cognitive Work Analysis model of the pediatric sepsis work domain and how AI uncertainty information may impact decision-making, (4) the development and evaluation of ecologically inspired human-AI interface graphics that visualize AI and work domain uncertainty information, and (5) understanding how these graphics may impact expert and non-expert healthcare professionals in a team huddle.
Research Partners
- The Hospital for Sick Children (SickKids)
Research Team
- Dr. Mark Ansermino, Department of Anesthesiology, The University of British Columbia, Vancouver, British Columbia, Canada
- Dr. Kate Mercer, Libraries, University of Waterloo, Ontario, Canada
- Jennifer Graham, Department of Psychology, University of Waterloo
- Juliet Kern, Systems Design Engineering, University of Waterloo
Related Publications
Tennant, R., Graham, J., Kern, J., Mercer, K., Ansermino, J. M., & Burns, C. M. (2024). A scoping review on pediatric sepsis prediction technologies in healthcare. Npj Digital Medicine, 7(1), 353. https://doi.org/10.1038/s41746-024-01361-9