AI scours 1.2 million appointments to cut doc office wait times
A Quebec tech company said it can cut wait times in doctors’ offices to less than 20 minutes and save patients 6.5 million hours a year in the Canadian province alone.
Bonjour Sante, which manages bookings for more than 300 clinics in Quebec, teamed up with Montreal’s flagship AI lab to create an algorithm predicting delays. The model looks at tens of thousands of past appointments through 1,900 parameters—from the weather to the type of medical-file software the doctor uses. Two hours before the scheduled time, patients receive a text estimating when the doctor will actually see them.
The service is currently offered at 10 clinics in the Bonjour Sante network at no additional charge. Founder Benoit Brunel said potential paying clients include hospitals and eventually, other industries struggling with needless waits.
“People are always looking for time; it’s the most valuable thing in life,” Brunel says. “We’re in the very early days. I have at least 20 projects I’d like to start.”
Businesses could create as much as $5.8 trillion in value per year globally adopting deep learning techniques—a subset of machine learning that’s inspired by the structure of the brain—according to a study by the McKinsey Global Institute. Yet the shift is only starting and complicated by one condition not all companies can meet: It takes a lot of data to "train" an algorithm and determine which model works the best.
Enter the medical field, where a treasure trove of information—including blood-pressure readings, lab results and doctors’ notes—are spurring companies to develop new strategies for improving efficiency and saving lives.
Microsoft, for example, is exploring how it can use AI to detect early warning signs of disease, while Google in May published a paper on an AI system that would provide a score on how likely a patient is to die based on almost 50 billion pieces of data.
For its part, Bonjour Sante had a data set of 1.2 million appointments to start working with MILA-Quebec Artificial Intelligence Institute, the center that has contributed to making Montreal one of the world’s AI hubs under the leadership of neural networks pioneer Yoshua Bengio. A group of experts there mentors selected companies under a technology-transfer program, with a focus on innovation, says Myriam Cote, who heads the government-subsidized initiative.
While Canadians enjoy universal health coverage, the difficulty of finding a doctor is pushing people into already overcrowded emergency rooms. Same or next-day appointments are harder to get in Canada than in France or New Zealand, a survey by the Canadian Institute for Health Information shows.
Bonjour Sante carved out a niche helping people score an appointment at a walk-in clinic in its home province. The platform, which some have criticized for introducing two tiers in the government -funded system, is free for doctors but charges patients about $13 for a search outside of their usual clinic. It’s part of closely held Tootelo Innovation, a company founded and headed by Brunel that offers customer-service and emergency response outsourcing.
Brunel, who owns 68 percent of Tootelo, says he’d consider going public if there was a sudden surge in demand, but has no immediate plan to do so.
The company previously tried to estimate delays using traditional statistical methods and a handful of criteria, but couldn’t lower the average wait below 45 minutes, according to Pierre Lafrance, who’s leading the project. In deep learning, data are repeatedly run by models, which can detect patterns—for instance, a doctor’s tendency to be late on a given day. Data keep being added, too.
At Polyclinique Levasseur, a facility on the island of Montreal that’s using the new program, director Julie Lessard says she’s been able to add doctors without facing a corresponding surge of people in the waiting room. “It’s very positive for us,” she said. “When people get into the consultation, they haven’t been here for two hours, chomping at the bit.”
Brunel says the algorithm could be exported to other provinces in Canada, where it has about 50 clients, or to countries with similar healthcare systems. Even in the U.S., a 2016 survey showed that 63 percent of patients find the wait in the lobby the most stressful part of going to the doctor.
Already, Lafrance is mulling the addition of new parameters to improve the algorithm, including a very local one: The game schedule for The Montreal Canadiens hockey team. “It’s less busy when the Canadiens plays,” Lafrance said. “Some clinics can literally be short on patients to fill the slots.”