Cybersecurity, privacy and artificial intelligence in health

Canada’s adoption of the latest information technology in the healthcare system is a hot button issue, and the stakes couldn’t be higher.  

That was the consensus of experts at a recent Waterloo conference held in Ottawa called “Cybersecurity, Privacy, and Artificial Intelligence in Health Data: Advancements and Challenges Conference.” 

The good news is, with the right information technology, the Canadian health-care system has the potential to optimize most aspects of medical care to ensure patients get the right care at the right time.  

One of the main challenges is that advances in technologies like AI are moving so fast that administrators and front-line workers may not fully trust in them such that they can enable their full use across the healthcare sector.  

“Our health data is something we rightfully want to keep safe and secure. It’s deeply personal information,” says Dr. Vivek Goel, President and Vice-Chancellor at the University of Waterloo, and an expert in governance, health data and policy. “The challenge is ensuring that we appropriately steward data to improve outcomes for patients and communities, health system management, population and public health, and research and innovation while ensuring that we keep this information safe and secure.” 

Goel, who chaired The Expert Advisory Group for the Pan-Canadian Health Data Strategy, played a leading role in setting the national agenda for the healthcare sector’s use of data.  

The problem is, Goel continued, “our systems, processes and policies are geared towards an analog world, while we live in a digital age. In Canada, the collection, organization and management of health data is not consistent. The primary obstacle we encounter is not technological but cultural in nature. Our fragmented approach and deep-seated distrust are what’s slowing down what we already know we can achieve.” 

This incredible potential for the development of technologies in the healthcare sector was the key topic of discussion in Ottawa, where Waterloo’s Cybersecurity and Privacy Institute (CPI) and Waterloo Artificial Intelligence Institute hosted the special session. The day long-session, organized by Dr. Anindya Sen, professor of economics and CPI associate director, with funding from Health Canada, Statistics Canada, and the Canadian Institute for Health Information (CIHI), featured diverse experts from Waterloo, each exploring different facets of our healthcare data puzzle.  

Sen also taught a data and coding workshop for federal government employees, which was held the day before the conference. The workshop was funded by the university’s Master of Public Service (MPS) program and based on courses developed by Sen for the Data Analytics and Behavioural Insights (DABI) certificate offered by WatSPEED. 

Predictably, much of the focus was on the rapid emergence of AI-powered large language models (LLMs) such as ChatGPT and what role they should play in modernizing how health data is used in Canada.  

“There’s quite a few reasons, ranging from technical, to legal, to moral, that we may want to pump the brakes on adopting LLMs,” says Dr. Gautam Kamath, an assistant professor in the Cheriton School of Computer Science at Waterloo, recently named as a Canadian Institute for Advanced Research (CIFAR) AI Chair and a Vector Institute Faculty Member in recognition of his contributions to differential privacy, machine learning and statistics.

“Is the algorithm less effective at designing treatments for individuals from certain demographic groups? Has the algorithm picked up on spurious correlations in its training data, resulting in worse utility after deployment? Is it okay to train a model on sensitive medical data, and how do we ensure that it is used in a privacy-respecting manner? How do we handle the cases when the algorithm makes mistakes, and who should be held liable when it does?” he asks.  

Dr. Samantha Meyer, an associate professor at Waterloo’s School of Public Health Sciences, also presented at the conference and noted that marginalized communities could be rightfully cautious in the face of this new technology given their experience with systemic oppression across Canadian social institutions.  

“We are seeing diminished trust in social institutions across all demographic groups. This has implications for public acceptance of government policy, and the acceptance of recommended health behaviours – including data sharing,” Meyer says. “The COVID-19 pandemic brought the notion of trust, as it relates to health broadly, to centre stage. The amount of disinformation about the pandemic and vaccines was staggering. We need to better understand why individuals look to, and trust, alternative forms of information that might be harmful. It is also critical we consider how to make our institutions trustworthy, and particularly for marginalized populations, if we are asking people to share personal and confidential information about their health.” 

Meyer’s research explores how we can measure trust in an ever-changing political climate in hopes we can develop strategies, tailored for specific populations, to (re)build trust in our social institutions. This starts with transparency and accountability, she says.

Dr. Sirisha Rambhatla, assistant professor in Management Sciences at UWaterloo also presented at the conference. She works closely with healthcare providers such as Grand River Hospital on building AI models to address inequities in healthcare systems, and with clinician researchers at University Health Network, specifically those performing liver transplants. She is well aware of the ethical issues raised by Kamath and Meyers, while also being acutely aware of the practical aspects of incorporating AI into our healthcare sector.   

"Making Canadian healthcare systems, what I call "AI Ready", is an urgent priority." she says. “AI/ML offers tremendous potential for our resource constrained healthcare systems, however patient information in our current healthcare systems is extremely fragmented and siloed-off, and we really don't understand the populations we serve. There are ways to make AI models fair. But for any kind of AI/ML modeling, we need to ensure that patient healthcare records are consolidated and made accessible to patients, clinicians, and researchers via a central system to give us a full picture."  

“Updating these systems is costly. Our cash-strapped health systems need funding to support these efforts, and guidance to build systems which are interoperable. Otherwise, we will remain in this deadlock for decades, and we can't afford that with our growing and aging population.” Rambhatla continues. 

The evening prior to the main conference, WatSPEED, Waterloo’s professional education program, hosted a dinner and discussion with health leaders from industry, government and academia.  

While the event was organized by Waterloo Associate Vice-President, Innovation Sanjeev Gill, Sen, Goel and Deputy Minister of Health Canada Dr. Stephen Lucas were all also critical in its conception.  

An informal conversation following the dinner, and moderated by President Goel, was an opportunity to share ideas and perspectives on how data can be used securely to improve patient outcomes. The mood in the room was hopeful, as nearly everyone agreed that the promise held by innovative technologies can only go as far as the willingness among stakeholders to collaborate towards a common goal.  

Anil Arora, Chief Statistician of Canada, who was also instrumental in making the gathering happen, described the dinner as “generating positive momentum in the responsible use of health data for the benefits of all Canadians.” 

He added, “I believe we can and need to play a leadership role in setting data standards, in data linkage as well as in ensuring that world class frameworks and statistical infrastructure remain strong foundational elements for progress. I look forward to concrete next steps and to deepening our collaboration.”