Personalized, preventive medicine is on its way
We’ll use machine learning to predict illness before it starts, reducing pressure on the health-care system
(This illustration was created by Maclean’s art director Anna Minzhulina using the generative AI image program Imagine. Minzhulina spent weeks feeding prompts into the program, inspired by the essay.)
Share
Roxana Sultan is the chief data officer at the Vector Institute.
For some people, the words AI in health care might evoke the image of a robot conducting their physical exam. In reality, it is less sci-fi and far more hopeful: artificial intelligence is leading us toward an era of preventive and personalized health care that could change our medical system for the better.
The pandemic revealed to the Canadian public what health-care workers already knew: our system is stretched thin, and our hospitals are at capacity. People wait months for surgeries and MRIs, emergency rooms are clogged, and doctors and nurses are overworked. A big reason for that is that health care in Canada is often reactive: we rush to treat people only once they are already sick. If we instead identified specific risk factors in patients and treated them preventively, it would reduce their hospital time and improve their outcomes down the road. AI-powered algorithms can get us there.
A few hospitals are already leveraging AI and seeing great results. At the Vector Institute, where I work, we collaborated with Unity Health Toronto’s St. Michael’s Hospital to implement an AI model trained on historical data that determines which inpatients are most at risk of escalating to the ICU or dying, based on metrics like age, biological sex and vital-sign measurements. The hospital implemented this algorithm in 2020, and even in the context of the pandemic, it reduced ICU escalation and death by more than 20 per cent. We estimate that this translated to about 100 deaths avoided annually, and staff report that it has relieved stress and workload, allowing them to focus their attention on patients who need it most.
We also supported the development of an algorithm at Toronto’s University Health Network, or UHN, that manages patients with congestive heart failure. Traditionally, those patients would be seen regularly for check-ups and would need to visit the hospital whenever they felt ill. Now, AI-powered software collects data from wearable devices and sends an alert if the patient’s vitals go out of range. The alert first goes to the patient. If they cannot stabilize themselves, the software then sends an alert to the team at UHN. A nurse coordinator sets up a virtual consultation with the patient, and if they cannot resolve the issue, only then do they need to go to the hospital. This process cut heart-failure patient hospitalizations in half and allowed nurse coordinators to support six times the number of patients compared to before.
AI will revolutionize medical research. Historically, a lack of time and money prevented us from conducting randomized, controlled trials of drugs and procedures on highly diverse populations, as well as on patients with uncommon health conditions. In consequence, many of the therapies approved for care in the Western world—from pain relievers to chemotherapies—were not trialled on groups that reflected the diversity of the people using them. The beauty of AI is that we will be able to work with massive amounts of historical data from all types of people, which could open the door to more precise treatment of patients based on things like age, genetics and even socioeconomic status.
There is similar potential for diagnosing and treating rare diseases. Right now, the data we have on rare conditions at any one hospital is limited, which makes it difficult to gain insight into how to treat patients in a personalized way. By collecting data from across the country or beyond and training AI models on that data, we could identify patterns with more ease and potentially treat conditions even before they cause symptoms.
It’s challenging to predict when these technologies will become widespread, because much still has to happen in the way of data governance. The more people allow the use of their de-identified medical data for science, the more we will learn about diseases, and the better we will be able to deliver more personalized care. However, it is critical that data is shared in a way that does not compromise people’s safety or privacy. Work is being done to continue to improve privacy preservation in AI, and Vector is working with its health partners to test innovative, privacy-enabled ways of training models on data across multiple hospital sites. We are also having productive conversations with federal and provincial government branches to see how we can successfully implement more trustworthy and safe AI solutions in health.
To ensure the sustainability of our public health system, we cannot just keep building hospital wings and adding beds and hoping that our problems go away. We need to innovate, and AI technology could have a greater positive impact than anything I have seen in my two decades of working in health care. This is not about replacing human resources with robots. It’s about enhancing personalized care delivered by people who are armed with the tools to ensure that we can achieve optimal health outcomes for all.
We reached out to Canada’s top AI thinkers in fields like ethics, health and computer science and asked them to predict where AI will take us in the coming years, for better or worse. The results may sound like science fiction—but they’re coming at you sooner than you think. To stay ahead of it all, read the other essays that make up our AI cover story, which was published in the November 2023 issue of Maclean’s. Subscribe now.