MONTREAL – Artificial intelligence software deployed by CIUSSS de l’Ouest-de-l’Île-de-Montréal will be able to customize users’ menus based on their preferences, aversions and specific dietary needs.
We stress that technology also reduces the risk of error regarding allergies, cross-contamination and other food accidents.
The programme has been gradually rolled out, since June 2023 and after nearly ten years of work, to Lakeshore, Saint Mary’s and Lasalle hospitals, as well as to Nazaire-Piché and Denis-Benjamin Viger CHSLDs. It can now be rolled out to other institutions.
“There will always be human intervention,” said Debbie Berto, a nutrition research officer. But the goal is to improve the presentation while reducing interactions.
If the user has the right to have a cake for dessert, she cites as an example, the software will take care of calculating the carbohydrate content, “allowing us to offer more variety, (…) we no longer see only foods that are allowed or prohibited.” This also has the advantage of freeing technicians or nutritionists from these tasks.
It would be difficult to do the same thing manually, because “it would take a lot of resources,” said Marie-Helene Cyr, a nutritionist in a consulting role.
“It would take a nutritionist doing the calculations manually to know, ‘OK, this user needs 60 grams of carbohydrates at each meal,'” she explained. She would have to do the math, and then if there was an aversion or a preference, that would change her calculations. Depending on the offer on the list, you will have to recalculate every day. We will never have enough human power to do what a computer does for us now.
The project “offers a new approach to nutritional prescription,” we explained via email, one that “relies on describing needs rather than pre-specified diagnosis.”
Managers explain that the software is able to automatically adjust the foods served, which will therefore increase users’ variety and satisfaction while respecting their dietary restrictions.
They point out, for example, that AI makes it possible to provide 61 additional foods for a potassium-restricted diet. A list containing potassium was identified, providing an average of 1,800 calories and 75 grams of protein, compared to 1,750 calories and 70 grams of protein in traditional diets.
The program will also increase the variety of foods offered, as the standardized menus will include 16% of the available foods. For users who request multiple restrictions, the menu has been added, with the menu now offering 1,775 calories instead of 1,400 calories, an increase of 25%.
Management
Nutritionists first gather users’ food preferences and aversions, as well as their specific nutritional needs. The program is then configured with standardized nutritional values, recipes, and serving sizes.
The AI then automatically adjusts the menus based on the data collected, without the need for manual intervention. Any changes in the nutritional values of foods or recipes are automatically taken into account by the system.
Lakeshore General Hospital already had, several years ago, a computer program to oversee and manage certain aspects of users’ nutrition. On this basis, work began in 2014, then accelerated in 2019 and led to the emergence of the current tool.
“It takes a lot of data in the background that has to be accurate in the computer system,” Ms. Sir explained. “And now we’re going to have to keep them updated.”
Personalization
Menu customization is now faster and more accurate, reducing the need for manual corrections. Users will therefore receive foods that adapt to their restrictions, with less risk to their health.
When faced with a diabetic user, Ms. Cyr cited as an example, “We had to make sure the total carbohydrates in each meal weren’t too high.”
“We had to remove the cake for everyone,” she said. A person with diabetes cannot have cake at all times. This (almost) never happened. But now, if my goal for dinner is 75 grams (of carbs) and the total tray allows for that, and I have plenty of room left over, that evening the person will be able to eat cake for dinner.
As the clients they serve get older, nutritional deficiencies often become a factor to consider, Ms. Sear added. Then a contradiction emerged: As much as we wanted to provide the user with protein- and energy-rich foods, we often had to remove these foods because of the diabetic menu they had to follow.
“Now we can meet both of those needs,” she said. We can make sure all calorie values are met, but also respect the total carbohydrates in each meal. So we increase the variety at every meal and people want to eat more.
Ms. Berto recalled that the project was first deployed on one floor of Lakeshore General Hospital. She explained that users who were transferred to the hospital on that floor were quick to express their displeasure when they were transferred elsewhere and lost access to this dedicated menu.
“Patients did not want to go back to the exceptions that had been advocated for,” Ms. Berto said. They wanted to maintain the new diet that was more diverse. “They no longer wanted the old system, which they found too restrictive.”
Ms. Sir said it was still necessary to educate and reassure patients who suddenly saw foods appear in front of them that they were not used to and that they thought were forbidden.
We also had to take the time to explain to the employees involved, who were wondering what impact the new system would have on their work, the benefits they would gain from it, but also the benefits to the users,” added Ms. Berto.
“There was some hesitation at first because it was a change in practice, but we learned from our previous experiences and worked (with staff) to find solutions,” she said. “The staff contributed to the solutions, which made the rollout easier.”
Ms. Ser added that employees ultimately found that the new system did not take away work, quite the opposite. Instead, it allowed them to focus on users with more complex needs, and in the end, “we had more users who were better fed.”