Food advertising is an important determinant of unhealthy eating. However, analysing a large number of advertisements (ads) to distinguish between food and non-food content is a challenging task. This study aims to develop a machine learning-based method to automatically identify and classify food and non-food ad videos.

Methodological study to develop an algorithm model that prioritises both accuracy and efficiency in monitoring and classifying advertising videos.

From a collection of Brazilian television (TV) ads data, we created a database and split it into three sub-databases (i.e. training, validation and test) by extracting frames from ads. Subsequently, the training database was classified using the EfficientNet neural network. The best models and data-balancing strategies were investigated using the validation database. Finally, the test database was used to apply the best model and strategy, and results were verified with field experts.

The study used 2124 recorded Brazilian TV programming hours from 2018 to 2020. It included 703 food ads and over 20 000 non-food ads, following the protocol developed by the INFORMAS network for monitoring food marketing on TV.

The results showed that the EfficientNet neural network associated with the balanced batches strategy achieved an overall accuracy of 90·5 % on the test database, which represents a reduction of 99·9 % of the time spent on identifying and classifying ads.

The method studied represents a promising approach for differentiating food and non-food-related video within monitoring food marketing, which has significant practical implications for researchers, public health policymakers, and regulatory bodies.

To map the scientific research on food environments in Brazil, based on the following questions: How many studies have addressed food environments?; What study designs and methodological approaches were applied?; What is the geographic scope of the studies?; What scenarios and dimensions of food environments were studied?; Which population groups were studied?; How were food environments conceptualised?; What are the main limitations of the studies?

Scoping review conducted in four databases, from January 2005 to December 2022, using different food environment-related terms to cover the main types and dimensions proposed in the literature. The studies were independently selected by two authors. A narrative synthesis was used to summarise the findings.

Brazil.

130 articles.

Scientific research on Brazilian food environments has been increasing. The analytical quantitative approach and the cross-sectional design were the most frequently used. Most articles were published in English. The majority of studies evaluated the community food environment, addressed aspects of the physical dimension, sampled the adult population, had food consumption as an outcome, used primary data, and were carried out in capital cities in the Southeast region. Furthermore, in most articles, no conceptual model was explicitly adopted.

Gaps in literature are related to the need for conducting studies in the Brazilian countryside, the support for the formulation of research questions based on conceptual models, the use of valid and reliable instruments to collect primary data, in addition to the need for a greater number of longitudinal, intervention and qualitative studies.

To describe the promotion of food and beverage and marketing strategies used by online food delivery services (OFDS) in a social media platform before and during the pandemic in Brazil.

Publicly available data were extracted from OFDS Instagram accounts. Posts published 6 months immediately before and after the first case of COVID-19 in Brazil were randomly sampled. Two independent authors coded the posts’ content. Food and beverage items featured in posts were classified according to the NOVA food system classification. Marketing strategies were coded according to protocols from previous studies.

Top three OFDS Instagram accounts in Brazil.

Posts published in the period studied (n 304).

During the pandemic, the proportion of posts featuring at least one food item decreased from 71·6 % to 40·2 %, and the proportion of ultra-processed foods decreased from 57·6 % to 27·9 %. Before the pandemic, the most widely used marketing strategies were branding elements (80·7 %), product imagery (unbranded) (48·9 %) and partnerships/sponsorship (35·2 %). While during the pandemic, branding elements (62·2 %) continued to be the most applied, but were followed by the use of videos/graphics interchange format/boomerangs (34·1 %) and corporate social responsibility (31·7 %). The most frequent COVID-19 marketing strategies were ‘social responsibility in the pandemic’ (30·5 %), ‘combatting the pandemic’ (28·0 %) and ‘accelerating digitalisation’ (20·7 %).

OFDS advertisements on a social media platform placed less emphasis on food items, but improved the nutritional quality of foods and beverages featured in posts. A COVID-washing approach was highlighted, especially through the use of social responsibility marketing during the pandemic.

To analyse food advertised on an online food delivery (OFD) platform during 16 weeks of the COVID-19 pandemic in Brazil.

Longitudinal study. We randomly selected foods advertised on the OFD app’s home page, classifying the food items into: water; natural juices and smoothies; vegetables; fruits; traditional meals and pasta; ultra-processed beverages; ice cream, candies, and salty packaged snacks; sandwiches; savoury snacks; and pizza. We also registered the marketing strategies used to promote the food items, such as photos, discounts, ‘combo deals’, and messages on healthiness, value for the money, tastiness, and pleasure.

Belo Horizonte, Minas Gerais, Brazil.

1593 food items.

In general, the OFD platform most commonly promoted traditional meals and pasta, ultra-processed beverages, and sandwiches – these food groups were offered 20–25 % of the time during the 16 weeks. There were no promotions for water during the whole period, and the least common promotions were those for natural juices and smoothies, vegetables, and fruits (< 5 %). The most common food promotion strategies were photos (> 80 %) and discounts (> 95 %), while approximately 30 % of the offers featured combo deals. Messages on tastiness, pleasure and value for the money varied from 15 % to 40 %.

Although the OFD platform offered traditional meals and pasta, most of the foods and beverages advertised were unhealthy and promoted with persuasive strategies. This finding highlights a growing public health concern: an increase in unhealthy eating patterns during the pandemic.

To investigate the advertising patterns on the posts of a fast-food chain in Brazil on three social media platforms in 2019.

An exploratory cross-sectional study.

Advertising strategies on the posts of a major fast-food chain on their official Facebook, Instagram and YouTube accounts. The strategies were investigated according to the INFORMAS protocol for food promotion monitoring. Principal component analysis (PCA) was employed to identify advertising patterns in each platform.

305 advertisements

Four advertising patterns were identified in the PCA of Facebook and Instagram. In both platforms, the components for kids and product exaltation were similar. On Facebook, a pattern corresponding to economic appeal was identified as price and discount, while on Instagram, this pattern also included a practical approach. On Facebook, the fourth component was named celebrity, while on Instagram it was celebrity/innovation since on this second social media the component also included the ‘new brand development’ variable. On YouTube, three advertising patterns were identified in PCA. Similar to the other platforms, the first and the second patterns were called for kids and price and discount, and the third component referred to both celebrity and commemorative dates.

The advertising patterns of the fast-food chain on three social media platforms were commonly directed to children and addressed price, discounts and the celebrities’ universe. The findings of this study corroborate other data in the literature regarding unhealthy food advertising on social media. This study discusses the urgency of regulating food advertising content on this medium.

Food delivery apps represent an important and emerging dimension of the digital food environment. This study aimed to examine food availability and the use of marketing strategies by two food delivery apps in a Brazilian metropolis.

An exploratory study was conducted in the city of Belo Horizonte, Minas Gerais. Food groups were identified and the use of price discounts and photos by the apps was observed.

Eighteen neighbourhoods and the ten best rated restaurants in each app.

Three hundred sixty-two commercial food establishments.

The proportion of ultra-processed beverages on offer in the apps (78·45 %) was much higher in comparison with water (48·89 %), natural juices or smoothies (27·07 %). Ultra-processed ready-to-eat meals represented almost 70 % of the food offered in the establishments’ menus, while traditional meals and vegetables represented just over 30 % of the offering. Ice cream, candies and salty packaged snacks were nine times more frequently presented than fruits. The use of photos and price discounts prevailed predominantly among ultra-processed beverages, sandwiches and ice cream, candies and salty packaged snacks. These marketing strategies were least used for promoting fruits and vegetables.

Restaurants registered on food delivery apps offered a significant amount of ultra-processed foods with price discounts and photos in comparison with unprocessed and minimally processed foods.

To analyse the extent and nature of food and beverage advertising on the three major Brazilian free-to-air television (TV) channels.

Cross-sectional study. A protocol developed for the International Network for Food and Obesity/Non-Communicable Diseases Research, Monitoring and Action Support was applied for data collection. A total of 432 h of TV programming was recorded from 06.00 to 24.00 hours, for eight non-consecutive and randomly selected days, in April 2018. All TV advertisements (ads) were analysed, and food-related ads were classified according to the NOVA classification system. Descriptive analyses were used to describe the number and type of ads, food categories and the distribution of ads throughout the day and time of the day.

The three most popular free-to-air channels on Brazilian TV.

The study did not involve human subjects.

In total, 14·2 % (n 1156 out of 7991) of ads were food related (858 were specific food items). Approximately 91 % of food items ads included ultra-processed food (UPF) products. The top three most promoted products were soft drinks, alcoholic beverages and fast-food meals. Alcoholic beverage ads were more frequently broadcast in the evening.

The high risk of exposure of the Brazilian population to UPF ads should be considered a public health concern given the impact of unhealthy food advertising on people’s food choices and health.

To estimate usual diets among Brazilian children regarding the consumption of school meals and social vulnerability risks.

A cross-sectional study. School meal consumers were considered those children who reported consuming school meals ≥3 times/week. Social vulnerability risk was classified by an index. Dietary intake was evaluated by one 24 h dietary recall for the whole sample; a second 24 h dietary recall was administered in a sub-sample (38·6 %). The National Cancer Institute’s method was used to estimate children’s usual intake of nutrients and food groups.

Municipal public schools from Belo Horizonte, Brazil.

Children (n 1357) aged 8–12 years.

Half of the sample lived in low/medium social vulnerability risk areas and 27·9 % were school meal non-consumers. School meal consumers more frequently lived in high/very high social vulnerability risk areas (76·2 v. 68·7 %). Children with low/medium social vulnerability risk had a higher mean intake of thiamin (1·13 v. 1·04 mg) and a lower mean intake of candy (1·35 v. 1·42 g). Consumption of school meals among children under high/very high social vulnerability risk was associated with higher mean consumption of vitamin C (31·9 v. 24·1 mg), unprocessed/minimally processed foods (956·3 v. 851·9 g), fruits (128·5 v. 90·9 g) and vegetables (58·2 v. 47·1 g). Ultra-processed food product consumption was lower among school meal consumers (136·2 v. 187·7 g), especially ultra-processed beverages (252·5 v. 305·7 g).

Consuming school meals was associated with a better usual diet quality, particularly among those with higher social vulnerability risk.