Introduction:

Monitoring the nutritional quality of national food supplies is key to nutrition research, policy development, and monitoring the food supply to curb diet-related non-communicable diseases.

Objective:

The Food Label Information and Price (FLIP) Program is a big data approach to the collection and evaluation of brand-name foods in Canada and Latin American and Caribbean Countries (LAC).

Methods:

FLIP is a longitudinal cloud-based database of packaged and chain restaurant foods and beverages collected since 2010 in Canada and since 2015 in LAC. The most recent iteration in Canada, FLIP 2020, used website “scraping”, artificial intelligence-enhanced optical character recognition (AI-OCR), natural language processing (NLP) and machine learning (ML) algorithms to collect, process and manage food label information on all foods and beverages available on seven major Canadian e-grocery retailer websites and 201 Canadian chain restaurants. Additionally, a pilot study was conducted to examine the feasibility of extracting Nutrition Facts tables (NFt) and ingredient list information from the food label photos in Spanish using AI-OCR.

Results:

FLIP-Canada is comprised of 119,541 packaged foods and 21,225 menu items. Automating categorizations and analyses of data allows for timely observations to support evidence-based public health policies including menu labelling, front-of-pack nutrition labeling, regulations restricting marketing of unhealthy foods to children, and sodium reduction guidance targets. FLIPLAC is comprised of 42,629 packaged foods and data were used to set the Updated PAHO Regional Sodium Reduction Targets, and to monitor progress against national, regional, and global sodium reduction targets as of 2022.

Conclusions:

FLIP, with its comprehensive sampling and granularity and recent use of AI-OCR, NLP and ML algorithms, is a powerful tool for greatly enhancing data collection and enabling the timely evaluation and monitoring of both the Canadian and LAC food supply.