Transforming diabetes care: SRI researchers secure $900K grant for AI prediction and prevention network
A new AI-powered solutions network for predicting and preventing diabetes developed by SRI researchers Laura Rosella, Jennifer Gibson, and Shion Guha has received $900K in funding from CIFAR’s AI for Health Solution Networks grant program.
In a groundbreaking stride toward revolutionizing diabetes care, the AI for Diabetes Prediction & Prevention Solution Network, co-led by Schwartz Reisman Institute (SRI) Faculty Affiliate Laura Rosella, has secured over $900,000 in funding from CIFAR. Over the course of three years, this visionary project aims to forge a pioneering framework utilizing already-validated machine learning models to predict diabetes risk in Ontario’s Peel region.
Rosella, an associate professor of the University of Toronto’s Dalla Lana School of Public Health and Temerty School of Medicine and a Canada Research Chair in Population Health Analytics, will direct this innovative endeavour alongside Jay Shaw, Canada Research Chair in Responsible Health Innovation in U of T’s Department of Physical Therapy. The project’s team also includes SRI Faculty Fellow Shion Guha, an assistant professor in the Faculty of Information who specializes in computer science with research interests in algorithmic decision-making and the nascent field of human-centered data science, SRI Faculty Affiliate Jennifer Gibson, an associate professor and director of U of T’s Joint Centre for Bioethics at the Dalla Lana School of Public Health, as well as Ibukun Abejirinde and Dr. Lorraine Lipscombe, scientists at Women’s College Hospital.
From left to right: SRI Faculty Affiliate Laura Rosella and Jay Shaw will co-direct the AI for Diabetes Prediction & Prevention Solution Network alongside team members SRI Faculty Affiliate Jennifer Gibson and Faculty Fellow Shion Guha.
A revolutionary approach to diabetes prevention
The network has already achieved significant milestones. Operating on routinely collected health system data, the models have demonstrated the ability to forecast diabetes up to five years before formal diagnosis, presenting a revolutionary approach to diabetes prevention.
“Our team developed and validated models that can predict diabetes incidence and complications in advance,” Shaw explains in an interview with the Temerty School of Medicine. “These models have already been validated, meaning that their performance for accomplishing their goals of predicting diabetes onset and complications has already been established, allowing us to focus on how best to implement these models so that they are used effectively and responsibly.”
The team will collaborate closely with health system practitioners, decision-makers, and community members, and adopt a participatory approach to surmounting barriers hindering adoption and implementation. Their collective efforts aim to construct a new framework tailored for the responsible deployment of AI-based technologies within health systems. This innovative framework will subsequently be integrated with the machine learning models, with deployment, monitoring, and evaluation slated for the project's culminating year.
The next phase of this study will see these validated models used to create a dynamic dashboard tailored for health system decision-makers. This tool will enable precise planning of interventions to address diabetes prevention needs, strategically targeting high-risk populations.
“One of the most seemingly intractable problems in public health”
Diabetes continues to affect more Canadians than ever before. The most recent figures released by Diabetes Canada show a disconcerting upward trend in diabetes rates, with 11.7 million Canadians, or around 30 percent of the population, living with diabetes or pre-diabetes in 2022. The number is projected to climb to nearly 14 million by 2030, potentially incurring almost $5 billion in direct costs to the health system. The intricacies of disease progression and diagnosis, compounded by persistent challenges in ensuring quality, equity, and accessibility of care, particularly influenced by socioeconomic factors, have resulted in poorer rates and outcomes for certain demographic groups.
An important consideration in choosing Peel region as the prime location for deploying these models is the region’s substantial diabetes burden. Peel has one of the highest rates of diabetes in Ontario. The City of Mississauga reports that approximately one in ten Peel adults has diabetes, and that number is likely to escalate to one in six by 2025. Furthermore, Peel region houses a diverse demographic, with 51 percent of its population being immigrants and 62 percent identifying as visible minorities.
“Diabetes is one of the most seemingly intractable problems in public health today, with enormous costs to the health, quality of life, and survival of millions of people in Canada, and wide inequities driven by socio-economic factors,” says Rosella.
“Our hope is that by the end of the three-year project, we can share and scale the resulting AI solution and framework for better diabetes prediction and prevention as well as shape how AI applications are designed for ethical impact in health systems in Canada and internationally,” she added.
The AI for Diabetes Prediction & Prevention Solution Network is one of two Solution Networks selected from a competitive pool of proposals to receive three years of funding from CIFAR. The initiative aims to advance novel strategies for the responsible use of AI to help solve challenging problems in the Canadian health system. The other Solution Network, based in Quebec, focuses on the integrated use of AI for health imaging.
