Delirium is an acute, fluctuating neurocognitive disorder characterized by severe confusion, disorientation and agitation that affects up to half of older hospitalized medical patients. It causes significant distress for patients and their caregivers, and can lead to dementia, longer hospital stays, increased health costs (an additional $10,899 per admission), and death.
The Challenge of Detection and Prevention
Despite its clinical significance, delirium remains difficult for doctors to predict and track. Current methods require either ongoing clinical assessments or the use of complicated screening tools that are often impractical and too costly for most hospitals to implement. As a result, administrative data significantly underestimates the true incidence of delirium, which makes it difficult to compare how different hospitals are performing, or to know if new interventions are truly working.
An AI Solution
GEMINI’s delirium project uses artificial intelligence (AI) to solve this challenge and has developed two automated tools: one to identify existing delirium cases, and another to predict a patient’s risk of developing it. This work is built on research by Lu Wang, Amol Verma, and colleagues, who have demonstrated the use of routine data extracted from electronic medical records to detect delirium and the application of machine learning to predict its onset.
One important concern in clinical AI is the risk of increasing health inequities by failing to account for social and economic variables. To address this, a multidisciplinary team including experts in clinical medicine, AI, and health equity, is rigorously testing these tools for bias to ensure that the detection and prediction of delirium are accurate and fair across diverse patient populations.
This project is a collaboration with Professor Mark Chignell and the University of Toronto (U of T), with support from the Vector Institute. Funding for this research came from the 2019 NSERC/CIHR Collaborative Health Research Projects (CHRP) grant.
From Research to Clinical Trial
Last year, this project entered into a new phase — and the most critical — moving from research to real-time clinical application. This spring, the AIM to Prevent Delirium Trial will launch across 13 Ontario hospitals. This trial is a partnership with Professor Eldan Cohen’s engineering lab at U of T, the Regional Geriatric Program (RGP) of Toronto and the Centre for Quality Improvement and Patient Safety (CQuIPS).
Led by an interdisciplinary steering committee of scientists, hospital leaders, and patient partners, the trial will evaluate the AI predictor tool in live clinical settings. By identifying high-risk patients at the point of admission, hospital teams can implement targeted, personalized interventions to prevent delirium before it starts.
This initiative represents a novel approach in delirium care and could change how quality indicators are measured. By shifting from reactive care to proactive, AI-enabled prevention, it represents an opportunity to modernize delirium care, aiming for a healthcare model that is both safer and more inclusive for all Canadian patients.