What restaurant menus reveal about obesity and inequality in cities

A man stands in front of a counter at a cafe. (Pexels)

A first-of-its-kind global study from the Massachusetts Institute of Technology (MIT) has found that certain types of food available on restaurant menus may be tied to local obesity rates – raising questions about how the same patterns might apply to urban centres across Canada.

The study, published in Scientific Reports, used artificial intelligence to analyze nearly 30,000 restaurants and millions of food items across three cities: Boston, Dubai and London.

Researchers then linked those nutritional profiles to the socio-economic and health data of each area.

The results? Neighbourhoods with higher obesity rates tended to be “nutrient deserts,” saturated with calorie-dense and ultra-processed menu offerings. In contrast, lower-obesity neighbourhoods had greater access to nutrient-rich foods.

Sadaf Mollaei, assistant professor at the University of Guelph and the Arrell Chair in the Business of Food, told CTVNews.ca Thursday the findings are likely reflective of trends in Canada too.

“There have been previous studies using other methodology that have links of the food environment with health outcomes or the socio-economic characteristics of the people,” Mollaei said.

“If this study is replicated in the Canadian context, it will probably have the same outcome or show the same results to some extent,” she added.

What the study found

The study’s authors were careful to suggest that their findings were “observational in nature.”

In a email to CTVNews.ca, study co-author Guido Camps said, “The presence of more energy-dense, lower-cost menu items in higher-obesity areas could potentially reinforce or exacerbate existing dietary patterns.”

Using machine learning and natural language processing, researchers extracted nutritional data from publicly available menus. In 2023, they assessed each menu for the relative presence of health and unhealthy ingredients through popular food-delivery platforms.

Researchers evaluated the food items as rated by the U.S. Department of Agriculture’s (USDA) FoodData Central database, an information bank with 375,000 kinds of food products.

Boston saw the strongest results from the study’s AI-based method. With a high match rate of 71 per cent between restaurant menu items and the USDA food database. They found that areas with more dietary fibre on menus tended to have lower obesity rates.

Wealthier neighbourhoods also had better access to high-fibre, healthier foods.

In London, the study revealed similar patterns. Low fibre availability in restaurant food correlated with higher obesity rates. Researchers also found a strong link between housing prices – a stand-in for income – and healthier food environments. London had a slightly lower match rate of 56 per cent with the U.S. food database. The lower match was partly due to differences in food terms like “chips” versus “French fries.”

The analysis in Dubai was less conclusive, with only 42 per cent of the menu items successfully matched to U.S. nutrition data. Researchers couldn’t draw strong links between restaurant food and rental prices or nutrition levels.

The lack of culturally specific data – like Arabic food items not found in the U.S. database – was a major limitation.

However, one district, Al Ttay, stood out for having the lowest fibre content in the analysis, which the researchers flagged as a potential red flag for future obesity risk.

How it works

Pinpointing where nutrient deserts are could help drive policies to close the health equity gap, Mollaei suggested.

To better understand the quality of food being offered in restaurants, researchers used two tools called the Meal Balance Index (MBI) and the Nutrient-Rich Foods Index (NFR).

The MBI looks at how much of each key nutrient – like protein, fibre or sodium – is in a meal compared to how many calories it has. It then compares that to daily recommended amounts and assigns a score based on how well the meal meets those targets.

A higher meal MBI means the meal is more balanced and nutritious, while a lower score suggests it may be lacking important nutrients or has too much of others.

The NFR score looks at the good stuff – like fibre, protein and calcium, as well as the less healthy ones like saturated fat, added sugars and sodium. The higher the NFR score, the healthier the food is considered.

By measuring nutrients per 100 calories, it allowed for comparisons across different foods no matter the portion size.

Camps said it was surprising to see how difficult and easy it was to assess health from a menu.

“Cheesecake is probably always too high in saturated fats/sugar and an unhealthy option, but caesar salad can be relatively healthy and low calorie but also lathered in dressing with bread and bacon and therefore not especially healthy,” Camps shared in his email.

“If it is difficult for us to judge how healthy your dish is based on the available public data, it’s also difficult for the average consumer to assess who may want to make a healthy choice. Calorie and nutritional info may help in this regard, but that is not standard to give in menus throughout all cities,” he added.

Affordability a major factor

Camps said while they cannot conclude causation between lower menu prices and higher calorie density in cities with greater obesity prevalence, affordability may influence customers food choices and what restaurants choose to offer.

“Lower prices may make high-calorie options more accessible and appealing, especially in areas where budget constraints are more common,” Camps said, adding that further research is needed to explore this economic factor.

Mollaei said when affordability and accessibility are at stake, food environments play a crucial role in shaping the choices people make.

While menu labelling regulations exist in provinces like Ontario – where restaurants must show calorie counts and flag items high in sugar, sodium or fat – she says these measures alone may not shift consumer habits. “If it’s not affordable or accessible, then even changing the menu items is not going to have a big impact or change consumption patterns.”

Mollaei believes studies like the one from MIT, which use AI to analyze restaurant menus in real time, could offer a “more dynamic” way to assess Canada’s food landscape. “It will give us insight into what’s actually available and accessible to people at a very granular level,” she said.

“You don’t want to fall into that loop – low-income, low-quality food, no other options – because people will keep going back to what’s there,” she added.