Hospitals, clinics, universities and other health-centric organizations regularly collect data on everything from spinal cord scans to sleep research results, but much of their valuable intelligence is covered in the company.
This is because researchers employing artificial intelligence and other data analytics tools missed the opportunity to improve patient health outcomes.
“Many organizations collect data,” says David Rotenberg, chief analytics officer at the Center for Addiction and Mental Health (CAMH). “But even if it's high quality, it often remains trapped and can be difficult to share. It limits what we can learn from it.”
Enter Health Data Nexus (HDN), the foundation stone provided by the University of Toronto Temerty for AI Research and Education in Medicine (T-cairem), which is part of the Temerty Dedicine of Medicine. The Health Database repository provides a safe and secure way to share data that has been stripped of individual patient information. It is also easy to access for those with academic or research qualifications, and is organized to read easily by AI algorithms.
In short, HDN is a silobust open source home for health data that helps solve the old “trash, trash output” problems of AI.
“Connecting data across an institution allows you to discover insights that a single team cannot find on its own,” says Rotenberg, who is also a co-leader at T-Cairem. “We are working on open science to advance medicine and advance how we apply AI to medicine.”
T-Cairem was launched as a research centre in December 2020 and proposed to meet the latter pillars, focusing on three pillars: research, education and data infrastructure. Six months later, HDN launched with three datasets.
“Half of the first year was laying the foundations through privacy impact assessments, threat risk assessments, initial governance and document resolution,” said Adams in January, who runs HDN as data governance and quality analyst for T-Cairem.
In fact, the repository has extensive data governance policies regarding information, ethics, consent and sharing.
According to Adams, HDN took its first major test in 2023 at a two-day datason, with around 40 researchers and students asking questions about the Nexus flagship dataset. The set includes 22,000 encounters of 14,000 unique patients over eight years, tracking forwarding, death, discharge and other outcomes.
HDN has since grown to 10 datasets – and Rotenberg says the team wants to add another five this year.
With the recent publication of journal articles and the growth of calendars of events, the team hopes to build HDN awareness while continuing to expand its scope.
“We are moving rapidly to grow our Nexus, but awareness is important. We want researchers to know. This is your go-to place to be your AI ready,” he says.
HDN is not the only health data repository available to researchers. Founded by the National Institutes of Health in 1999, Physionet is running out of Massachusetts Institute of Technology (MIT). (Adams says he regularly meets with the team behind Physion to share his infrastructure and regulatory ideas.) There is Nightingale Open Science, Medical Imaging, run by the University of Chicago business school.
However, Rotenberg says HDN is unique in its range. “Our dataset spans the entire range of medicine, including wearables, ultrasound, speech, text, imaging, and more. We bring together a wide range of health information in one place. That diversity leads to AI discovering patterns across disciplines, leading to breakthroughs that are impossible in a single discipline.”
Qualified researchers can sign up to access HDN's database after completing an online training course on research ethics. You can then minify your HDN information to use it yourself or enrich your own data. You can also work with remote partners. “It's easier to collaborate when you no longer have to cross-reference datasets, compare results, and navigate the infinite barriers for partners to access,” says Rotenberg.
The T-Cairem team plans to continue improving its repository and is working to support agencies adding their own datasets. It offers a $50,000 grant to help researchers prepare their data.
“This is the issue of making it a usable and valuable format. This means that the machine is easy to read, so these models interface well,” Adams says.
In addition to providing materials for health research, the repository demonstrates its potential as an educational tool. It is used in the U of T graduate data science course by Azadeh Kushki, a senior scientist at Holland Bloorview and an associate professor at the Institute of Biomedical Engineering.
While governments south of the border are restricting data collection and access, AI algorithms increasingly provide promises to better understand human health, Rotenberg says the need for better data solutions is bigger than ever, and HDN is useful. “It's a unique Canadian model built on safe, collaborative, and trust. It changes the way we interact with data and accelerates discoveries that benefit people everywhere.”
/Public release. This material of the Organization of Origin/Author is a point-in-time nature and may be edited for clarity, style and length. Mirage.news does not take any institutional position or aspect, and all views, positions and conclusions expressed here are the views of the authors alone.
