Garth Gibson is President & CEO of the Vector Institute, an independent, not-for-profit research institute dedicated to advancing artificial intelligence, excelling in machine and deep learning. The Vector Institute is funded by the Province of Ontario, the Government of Canada through the Pan-Canadian AI Strategy administered by CIFAR, and industry sponsors from across the Canadian economy.
Improving patient care while simultaneously reducing costs is the Holy Grail for hospitals. St. Michael’s Hospital of Unity Health Toronto is taking a novel approach to achieving this goal. Earlier this year, the hospital used machine learning (ML), a field of artificial intelligence, to predict peaks in ER admissions with 95-per-cent accuracy, automate ER nurse assignments and optimize nurse staffing. The result: reduced wait times, a reduction in the time spent on assigning nurses’ tasks – down from hours to seconds – and optimized nurse staffing for an estimated savings of approximately $1-million annually.
None of this would have been possible without the hospital’s in-house data science and ML expertise and analysis of records they collect every time a patient is admitted.
ML technologies are powerful. The technologies that understand voice commands in smartphones can also detect signs of dementia in speech. Computer vision tools that suggest faces to tag on social media can also pinpoint cancer on radiology scans. And algorithms, such as the ones that predict movie and music preferences, can also predict the likelihood of a heart attack. Health-care providers and ML researchers are – unsurprisingly – keen to tap into this potential.
But in order to work, these technologies require vast amounts of data. The greater the volume and diversity (in terms of sociodemographic makeup) of that data, the more accurately algorithms can predict the health of individuals and populations.
The St. Michael’s project took place within a single hospital. Imagine the improvements to care and cost savings that could be realized if ML researchers and practitioners – such as those working at the Vector Institute – could apply these technologies across multiple organizations or even the entire health-care system.
The large amounts of data collected by Canada’s public-health systems are ideal for enabling ML insights and solutions, and the world-class ML researchers recruited under Canada’s $125-million AI strategy are keen to affect positive change. Major advancements are within reach, but ML experts lack sufficient access to the data underpinning such technological breakthroughs. Most are limited to studying data from a single hospital (often from a single department) or using population data sets from the United States and the United Kingdom.
Health insights gathered from smaller and more homogeneous data sets will not necessarily be suitable for Canadians, a fifth of whom are foreign born (51 per cent in the GTA). Data samples that are too small or are not reflective of our own population and health system are likely to produce results that are incorrect, ill-suited or – worst case scenario – cause harm for Canadians.
The opportunities and risks of inaction are too great. Defining a health-data governance framework that safeguards privacy without stymieing the development of life-saving technologies is a critical public policy challenge. Unaddressed, our public health system risks playing catch-up in frameworks created for the benefit of people and companies in other countries. Moreover, we risk losing our world-class ML experts to other jurisdictions.
There is no silver bullet. But in an effort to forge a pragmatic approach, last month the Vector Institute and the Canadian Institute for Health Information met with health-data stakeholders including ICES, HPC4Health (a joint initiative of the Hospital for Sick Children and University Health Network), Sunnybrook Research Institute, MaRS, Population Data BC, Compute Ontario and the Global Alliance for Genomics and Health. The goal was to pool local and international experiences and identify essential elements for a modern health-data governance framework.
Four key requirements emerged: First, the need for overarching ethical principles for data access, including accountability and transparency. Secondly, we must have data management and governance that lays out clear processes for the collection, storage and use of health data. Further, we must have requirements for researchers and practitioners who access health data, including training and accountability measures. Lastly, we require continuing stakeholder and public engagement about how health data are used, by whom and for what purposes.
This is not a case of privacy- or ethics-washing (terms coined to describe companies that don’t walk the talk). Vector and its partners want to help create a practical, enforceable health-data governance framework with teeth. Only then can we determine our own data destiny.
Keep your Opinions sharp and informed. Get the Opinion newsletter. Sign up today.