With Montreal’s Scale AI, hospitals across Canada are testing new AI applications

McGill University Health Centre (MUHC) in Montreal identified opportunities to perform hundreds of additional surgeries each year. Toronto’s Princess Margaret Cancer Centre decreased the mortality rate from delayed radiology treatment by up to 13 percent. And the Centre Hospitalier de l’Université de Montréal (CHUM) achieved a 5 percent increase in efficiency in its infusion clinic – amounting to 11 hours of extra treatment capacity per day – by accurately predicting patient treatment times.

How did they do it? By applying artificial intelligence (AI) to find efficiencies and optimize limited resources, the proverbial ‘do more with less’ strategy that Montreal-based innovation hub Scale AI is working to accelerate across Canada’s healthcare sector.

Funded by the federal and Quebec governments, Scale AI is on a mission to build a national AI ecosystem by helping organizations to implement real-world applications.

In October 2023, the hub announced investments of $21 million for nine AI healthcare projects as part of the Pan-Canadian Artificial Intelligence Strategy. The announcement marks the second time Scale AI decided to open a funding call geared specifically to hospitals.

The first took place during the early days of the pandemic and resulted in projects that optimized surgery schedules at MUHC and radiation patient scheduling at Princess Margaret Hospital, and led to the development of an AI-guided tool to maintain safety and quality of cancer treatment at CHUM.

“We hear a lot about how AI can be used for therapeutic applications such as pre-analysis of medical imagery … or genAI apps to facilitate report writing and patient visit summaries,” said Scale AI senior director, Investments, Marc Vaucher. “But there is a third use of AI that we promote a lot at Scale AI, which is to improve productivity, optimizing the operations around the healthcare system and allowing the healthcare system to treat more patients with the same resources.”

The nine projects included in the recent round of healthcare-focused funding range from virtual triaging and queue management in busy emergency rooms, to optimizing workflow operations and patient care management, to improving demand forecasting.

Each received a Scale AI investment of $1.5 million and though led by and deployed into specific hospitals, all projects involve a solution provider or system integrator with the goal of making the final deliverables available to any Canadian hospital.

Part of Scale AI’s role is to ensure every funded project sets out a clear change management plan that engages end-users from the beginning. The hub also structures projects in levels so that impacts can be measured at each stage of deployment and challenges can be addressed, with the intent of ensuring implementations are seen through to completion.

“We know we have limited resources (in healthcare),” said Vaucher. “If you allocate staff better, then you avoid overtime hours and reduce turnover, improve quality of life for your staff, and at the end of the day, it’s also improving the quality of care you provide to your patients.”

Global professional services firm Deloitte has partnered on several Scale AI-funded projects in the Canadian healthcare sector, including one that is applying AI to build a physician scheduling model at a large western Canadian health authority and another that is harnessing AI to improve medication adherence at a home and community healthcare provider by segmenting patients into personalized care pathways.

An earlier project, representing a total investment of $6 million with $1.8 million coming from Scale AI, was aimed at helping SE Health, a homecare and community care provider, to better manage its Ontario home care workforce – responsible for roughly 120,000 client visits per week – at a macro level.

The goal was to create an AI-enabled decision support tool capable of better matching staff resources to patient demand while supporting the social enterprise’s over-arching goal of providing continuity of care.

“Everything we do for home care is geography based,” explained SE Health vice-president of Business Transformation, Jennifer Hayward, noting that teams operate from 13 regional service delivery centres. “Our staff are going into people’s homes across the province, so the challenge is trying to figure out what team sizes and regional boundaries are needed to optimize capacity,” she said.

The AI model built by Deloitte’s data sciences team uses postal code data from Canada Post and demographic data aggregated by Manifold Data Mining Inc. to forecast patient demand and the type of caregiver skill required at a regional level.

The decision support tool – referred to as the GO Tool for Geographic Optimization, and used by clinical and administrative managers in each region – uses the model to generate recommended team geographical boundaries on a quarterly basis, suggesting where to draw borders in order to minimize the distance travelled by caregivers so they spend more time caring for patients.

From the start, the GO Tool was designed with end-users in mind, said Deloitte Digital Studio partner Arslan Idrees, who focuses on leveraging technology to improve the citizen experience across Canada as the lead for Deloitte’s national life sciences, healthcare and aging well market cluster.

“We wanted non-technical clinical staff to be able to go in and play around, to move the line and see how it changes the whole dynamic,” said Idrees, explaining that the AI model does the heavy lifting in the background to calculate the impact on business metrics.

“They can see the business impact of moving the geographic boundary – including how many patients they can serve – in real-time and then make a decision to go forward with the change or not,” he added.

Before using the GO Tool, workforce planning was a tedious manual ask that wasn’t performed very often and didn’t always provide reliable information. Now teams are better aligned, and managers have a deeper understanding of capacity and the impact that small changes can make.

“In some cases, we were pleasantly surprised that the tool validated that what we were doing was good, now we can just do it quicker,” Hayward said. “In other cases, it was ‘Oh my goodness! I can’t believe I can generate that many extra visits or better utilization just by changing this boundary.’”

Since being deployed, the AI solution has led to an increase in staff utilization, a decrease in travel time, higher patient volumes and increased employee satisfaction, she added.

Idrees said the idea behind the Scale AI innovation model is simple: we need to solve challenges in the Canadian healthcare sector and doing that requires marrying a human-centred design approach with technology to ensure the right problem is being solved in the right way.

An excellent example of that is the Scale AI-funded virtual triaging and queue management project currently under way at Humber River Health in Toronto, supported by Deloitte and the MEDITECH collaborative.

When Humber River Health first launched the project as a way to reduce patient wait times in a busy emergency department, they were focused on using AI to help screen potential patients from home and make a judgment as to whether they should or should not visit emergency.

When Scale AI and partners came on board last fall, the focus shifted. “We asked, ‘What’s the novel problem here?’ And it’s actually not the screening function but the queuing function,” said Humber River Health CIO Peter Bak. “And that has proven to be really interesting.”

The main challenges in the emergency department are overcrowding and extended wait times due to scarce workforce and limited capacity. The idea is to solve the problem using AI, without adding resources.

The new ED Queue app is designed to provide emergency department patients with a personalized time slot that is updated according to real-time conditions in the ER, historical analysis and other factors, so that they can wait at home as long as possible before arriving at the hospital.

Currently in the pilot stage, ED Queue will be rolled out over the next three months and the project is expected to be completed by March 2025.

As Idrees explained, there are three possible scenarios: a patient is advised to call 911 immediately, get themselves to the emergency department right away, or wait at home until their time slot is available.

“We ask simple questions about how you are feeling and the magic happens at the backend,” said Idrees, stressing the time slot is not an appointment. Rather it’s a moving target that will change as circumstances change, but patients will always be in the know rather than sitting helplessly in a busy waiting room.

The shift in focus has opened up new possibilities for the app, such as partnering with a digital twin of the emergency department to better drive performance. Bak envisions a scenario in which patients are categorized into care pathways. If it’s likely they’re going to require bloodwork or an X-ray, the algorithm can be massaged to take into consideration radiology and lab capacity.

“We can actually start to coordinate the time slot not just based on activity in the emergency department, but on activity in the emergency department and the supporting services that are required,” he said.

Deloitte’s and HRH’s initial analysis of ED queue demonstrates that patient wait times in the ED could decrease by as much as 78 percent in the Fast-track zone. Possible future enhancements to the algorithm include adding real-time traffic data so that a patient’s estimated drive time can be included, or partnering with ride hailing apps so patients without a vehicle can book a ride from within ED Queue.

To support Scale AI’s goal of developing a Canadian AI ecosystem, the queuing app is designed to be replicable in other hospitals. The front-end user interface is decoupled from the backend so that hospitals can use their existing portals, digital platform or other form of patient engagement.

Right now, the backend data stream is integrated with MEDITECH, but the integration layer is “light touch”, meaning support for other electronic medical record systems can easily be added, said Idrees.

When it comes to implementing cutting-edge AI solutions in healthcare, Idrees believes the lack of a robust AI governance framework remains a critical challenge. Whereas other sectors focus on outcome when embarking on an innovation project, solving security, privacy, ethical or regulatory challenges as they come along, healthcare entities tend to have those discussions upfront causing long delays.

“In healthcare, the paradigm of innovation is flipped upside down and what happens is you’re asking questions that should be handled downstream,” said Idrees. “An AI governance framework would allow champions of the idea to make sure concerns are handled at a certain stage and the rigour depends on the stage you’re at.”

A key benefit of the Scale AI approach, said Bak, is that it allows hospitals to come with up ideas, validate them and then rely on system integrators and partners to help commercialize them.

“We’re very grateful for Scale AI to have funded us. We’ve worked well with Deloitte and we’re very interested in commercialization,” he said. “What we’re hoping is that this is something that does indeed get scaled-out for the benefit of the Canadian health system.”