Determining the ‘healthiness’ of foods marketed to children on television using the Food Standards Australia New Zealand nutrient profiling criteria

Perception and Impact of Food and Beverage Marketing on Children's Eating Behaviors and Associated Health Issues

BACKGROUND AND OBJECTIVE

Children are more susceptible to food and beverage marketing than adults, but little is known about the specific effects of marketing through the media most used by children. This study aims to discover variables that can help inform childhood obesity prevention strategies. Our findings indicate an association between food advertisements and children's consumption, evidencing a need for the concerned authorities to create strict guidelines that consider the nutritional value of advertised foods. This study aims to study the attitudes and practices of children related to their preference for unhealthy meals due to food marketing and their association with childhood obesity.

METHODOLOGY

A cross-sectional study of randomly selected guardians of children who were screened for obesity. A structured questionnaire was given to the children's parents.

RESULTS

The study found that most of the participants' children prefer fast food (291, 78.0%), eat healthy meals (287, 76.9%), and eat fruits and vegetables every day (198, 53.1%). Furthermore, most participants (340, 91.2%) indicated that they were aware of unhealthy diets, and 105 (28.2%) said their children were overweight. Most participants (326, 87.4%) also indicated that watching television (TV) was associated with eating high-calorie foods.

CONCLUSIONS

There is strong evidence that children exposed to food marketing develop attitudes about and choose unlimited healthy food and unhealthy foods, which negatively impacts their health. It is recommended that future research employs a wide range of methodologies to study contemporaneous marketing strategies.

Comparison of nutrition profiling models for food marketing regulation

AIM

Regulation on food marketing to children is a recommended response to childhood obesity rates. Policy requires country-relevant criteria to determine which foods are eligible to be advertised. This study aims to compare six nutrition profiling models for use in food marketing regulation in Australia.

METHODS

Advertisements on the outside of buses in five suburban Sydney transport hubs were photographed. Food and beverages advertised were analysed using the Health Star Rating; three models developed for food marketing regulation: the Australian Health Council guide and two World Health Organization models; the NOVA system; and the Nutrient Profiling Scoring Criterion used in Australian advertising industry codes. The proportion and types of products advertised on the buses that would be permitted by each of the six models were then analysed.

RESULTS

A total of 603 advertisements were identified. Of those, over a quarter of the advertisements were for foods and beverages (n = 157, 26%) and 2.3% (n = 14) for alcohol. Among the food and non-alcoholic beverage advertisements, 84% were for unhealthy foods according to the Health Council guide. The Health Council guide would permit 31% unique foods to be advertised. The NOVA system would permit the least proportion of foods to be advertised (16%), while the Health Star Rating (40%), and Nutrient Profiling Scoring Criterion (38%) would permit the most.

CONCLUSION

The Australian Health Council guide is the recommended model for food marketing regulation because it aligns with dietary guidelines by excluding discretionary foods from advertising. Australian governments can use the Health Council guide to develop policy in the National Obesity Strategy to protect children from marketing of unhealthy food.

Natural language processing and machine learning approaches for food categorization and nutrition quality prediction compared with traditional methods

BACKGROUND

Food categorization and nutrient profiling are labor intensive, time consuming, and costly tasks, given the number of products and labels in large food composition databases and the dynamic food supply.

OBJECTIVES

This study used a pretrained language model and supervised machine learning to automate food category classification and nutrition quality score prediction based on manually coded and validated data, and compared prediction results with models using bag-of-words and structured nutrition facts as inputs for predictions.

METHODS

Food product information from University of Toronto Food Label Information and Price Database 2017 (n = 17,448) and University of Toronto Food Label Information and Price Database 2020 (n = 74,445) databases were used. Health Canada's Table of Reference Amounts (TRA) (24 categories and 172 subcategories) was used for food categorization and the Food Standards of Australia and New Zealand (FSANZ) nutrient profiling system was used for nutrition quality score evaluation. TRA categories and FSANZ scores were manually coded and validated by trained nutrition researchers. A modified pretrained sentence-Bidirectional Encoder Representations from Transformers model was used to encode unstructured text from food labels into lower-dimensional vector representations, followed by supervised machine learning algorithms (i.e., elastic net, k-Nearest Neighbors, and XGBoost) for multiclass classification and regression tasks.

RESULTS

Pretrained language model representations utilized by the XGBoost multiclass classification algorithm reached overall accuracy scores of 0.98 and 0.96 in predicting food TRA major and subcategories, outperforming bag-of-words methods. For FSANZ score prediction, our proposed method reached a similar prediction accuracy (R²: 0.87 and MSE: 14.4) compared with bag-of-words methods (R²: 0.72-0.84; MSE: 30.3-17.6), whereas structured nutrition facts machine learning model performed the best (R²: 0.98; MSE: 2.5). The pretrained language model had a higher generalizable ability on the external test datasets than bag-of-words methods.

CONCLUSIONS

Our automation achieved high accuracy in classifying food categories and predicting nutrition quality scores using text information found on food labels. This approach is effective and generalizable in a dynamic food environment, where large amounts of food label data can be obtained from websites.

Monitoring complaints about food marketing to children under the Australian industry Codes 2015-20: a qualitative analysis

OBJECTIVE

To analyse the case reports of complaints regarding food marketing to children over six years and compare Australian marketing Codes to best practice recommendations.

METHODS

Case reports on complaints about food marketing to children under the five industry Codes - the Responsible Children's Marketing Initiative, the Quick Service Restaurant Initiative and the Australian Association of National Advertisers Code of Ethics, Code of Advertising and Marketing to Children and Food and Beverages Code - were qualitatively analysed. Reports on the Ad Standards website in the food/beverage groceries and food/beverage venues categories from 2015-2020 were investigated. The most common clauses from the Codes were identified and quotes from reports used to illustrate the determinations. Codes were compared with World Cancer Research Fund recommendations on policy to protect children.

RESULTS

Only 14 of 119 complaints resulted in a reported breach of industry Codes. The most common reason for dismissing complaints involved clauses requiring advertisements to be 'primarily' directed to children. The Codes did not align with best practice recommendations.

CONCLUSIONS

Complaints by the public show concern for food advertising to children but the Australian industry Codes fall short of addressing those concerns. Implications for public health: Government regulation is required to protect children from unhealthy food marketing.

Defining unhealthy food for regulating marketing to children-What are Australia's options?

AIM

To compare six nutrient profiling models for suitability in food marketing to children regulation.

METHODS

Products (n = 220) advertised at transport hubs were classified as eligible/ineligible to be advertised to children using an Australian government developed guide (Council of Australian Governments), the Health Star Rating system before and after the modifications made in 2020, World Health Organization Western Pacific Region and Europe nutrient profile criterion and the NOVA food classification system. Agreement between models was determined using Cohen's Kappa.

RESULTS

The Council of Australian Governments' guide was able to classify more products than the other models (n = 210) and was easy to use as it did not require nutrition information. It agreed most closely with NOVA (moderate agreement). The proportion of foods classified as eligible to be marketed (most strict) was lowest for NOVA (10%), similar for Council of Australian Governments and the World Health Organisation models (16%-17%) and highest for Health Star Rating models (26%-28%).

CONCLUSIONS

The Council of Australian Governments' guide provides simple, easy to use profiling criteria aligned with Australian dietary advice. Political will is now required to incorporate a profiling model within government-endorsed food marketing regulation.

Bus Stops Near Schools Advertising Junk Food and Sugary Drinks

Children rarely understand the full extent of the persuasive purpose of advertising on their eating behaviours. Addressing the obesogenic environments in which children live, through a quantification of outdoor advertising, is essential in informing policy changes and enforcing stricter regulations. This research explores the proportion of bus stop advertisements promoting non-core food and beverages within walking distance (500 m) from schools in Auckland, New Zealand while using Google Street View. Information was collected on: school type, decile, address, Walk Score^®, and Transit Score for all 573 schools in the Auckland region. Ground-truthing was conducted on 10% of schools and showed an alignment of 87.8%. The majority of advertisements on bus shelters were for non-food items or services (n = 541, 64.3%). Of the advertisements that were for food and/or beverages, the majority were for non-core foods (n = 108, 50.2%). There was no statistically significant difference between the variables core and non-core food and beverages and School decile (tertiles), Walk Score (quintiles), and Transit Score (quintiles). 12.8% of all bus stop advertisements in this study promoted non-core dietary options; highlighting an opportunity for implementing stricter regulations and policies preventing advertising unhealthy food and drink to children in New Zealand.

Comparison of global nutrient profiling systems for restricting the commercial marketing of foods and beverages of low nutritional quality to children in Canada

Background: The Canadian government recently committed to introduce legislation to restrict the commercial marketing of unhealthy foods and beverages to children.Objective: We compared the degree of strictness and agreement between nutrient profile (NP) models relevant to marketing restrictions by applying them in the Canadian context.Design: With the use of data from the University of Toronto 2013 Food Label Information Program (n = 15,342 prepackaged foods), 4 NP models were evaluated: the Food Standards Australia New Zealand-Nutrient Profiling Scoring Criterion (FSANZ-NPSC), the WHO Regional Office for Europe (EURO) model, the Pan American Health Organization (PAHO) model, and a modified version of the PAHO model (Modified-PAHO), which did not consider the extent of food processing because the application of this characteristic was prone to ambiguity. The number and proportion of foods that would be eligible for marketing to children was calculated with the use of each model, overall and by food category.Results: The Modified-PAHO and PAHO models would permit only 9.8% (95% CI: 9.4%, 10.3%) and 15.8% (95% CI: 15.3%, 16.4%) of foods, respectively, followed by the EURO model [29.8% (95% CI: 29.0%, 30.5%)]. In contrast, the FSANZ-NPSC would consider almost half of prepackaged foods as eligible for marketing to children [49.0% (95% CI: 48.2%, 49.8%)]. Cross-classification analyses showed that only 8.1% of foods would be eligible based on all models (e.g., most pastas without sauce). Subanalyses showed that each model would be more stringent when evaluating food items that specifically target children on their package (n = 747; from 1.9% of foods eligible under Modified-PAHO to 24.2% under FSANZ-NPSC).Conclusions: The degree of strictness and agreement vary greatly between NP models applicable to marketing restrictions. The discrepancies between models highlight the importance for policy makers to carefully evaluate the characteristics underlying such models when trying to identify a suitable model to underpin regulations restricting the marketing of unhealthy foods to children.

An audit of food and beverage advertising on the Sydney metropolitan train network: regulation and policy implications

BACKGROUND

Increased marketing of energy-dense, nutrient-poor foods has been identified as a driver of the global obesity epidemic and a priority area for preventative efforts. Local and international research has focused on the unhealthiness of television advertising, with limited research into the growing outdoor advertising industry. This study aimed to examine the extent of food and beverage advertising on the Sydney metropolitan train network, and to assess the nutritional quality of advertised products against the Australian Guide to Healthy Eating.

METHODS

All 178 train stations on the Sydney metropolitan train network were surveyed in summer and winter. A survey tool was developed to collect information for all advertisements on and immediately surrounding the train station. Information included product, brand, location and advertisement format. Advertisements were coded by nutrition category, product subcategory and size. Chi-square, ANOVA and ANCOVA tests were conducted to test for differences in the amount of food and beverage advertising by season and area socioeconomic status (SES).

RESULTS

Of 6931 advertisements identified, 1915 (27.6%) were promoting a food or beverage. The majority of food and beverage advertisements were for unhealthy products; 84.3% were classified as discretionary, 8.0% core and 7.6% miscellaneous. Snack foods and sugar-sweetened beverages were the most frequently advertised products, regardless of season. Coca-Cola and PepsiCo were the largest advertisers on the network, contributing 10.9% and 6.5% of total advertisements respectively. There was no difference in the mean number of food and beverage advertisements by area SES, but the proportion of advertising that was for discretionary foods was highest in low SES areas (41.9%, p < 0.001).

CONCLUSIONS

The results indicate that, irrespective of season, food and beverage advertisements across the Sydney metropolitan train network are overwhelmingly for unhealthy (discretionary) products. The results of this study highlight the inadequacy of Australia's voluntary self-regulatory system in protecting members of the public from exposure to unhealthy food advertising. Regulatory action by government, such as placing a cap on the amount of unhealthy food advertisements, or requiring a proportion of all advertising to be for the promotion of healthy foods, is required to address this issue.

Community perspectives on the use of regulation and law for obesity prevention in children: A citizens' jury

INTRODUCTION

Childhood obesity is a significant challenge for public health internationally. Regulatory and fiscal measures propagated by governments offer a potentially effective response to this issue. Fearing public criticism, governments are often reluctant to use such measures. In this study we asked a descriptively representative and informed group of Australians their views on the use of legislation and fiscal measures by governments to address childhood obesity.

METHODS

A citizens' jury, held in South Australia in April 2015, was asked to consider the question: What laws, if any, should we have in Australia to address childhood obesity?

RESULTS

The jury agreed that prevention of obesity was complex requiring multifaceted government intervention. Recommendations fell into the areas of health promotion and education (n=4), regulation of food marketing (n=3), taxation/subsidies (n=2) and a parliamentary enquiry. School-based nutrition education and health promotion and mandatory front-of-pack interpretive labelling of food and drink were ranked 1 and 2 with taxation of high fat, high sugar food and drink third.

CONCLUSION

The recommendations were similar to findings from other citizens' juries held in Australia suggesting that the reticence of decision makers in Australia, and potentially elsewhere, to use legislative and fiscal measures to address childhood obesity is misguided. Supporting relevant informed public discussion could facilitate a politically acceptable legislative approach.

Television food advertising to children in Slovenia: analyses using a large 12-month advertising dataset

OBJECTIVES

The marketing of energy-dense foods is recognised as a probable causal factor in children's overweight and obesity. To stimulate policymakers to start using nutrient profiling to restrict food marketing, a harmonised model was recently proposed by the WHO. Our objective is to evaluate the television advertising of foods in Slovenia using the above-mentioned model.

METHODS

An analysis is performed using a representative dataset of 93,902 food-related advertisements broadcast in Slovenia in year 2013. The advertisements are linked to specific foods, which are then subject to categorisation according to the WHO and UK nutrient profile model.

RESULTS

Advertising of chocolate and confectionery represented 37 % of food-related advertising in all viewing times, and 77 % in children's (4-9 years) viewing hours. During these hours, 96 % of the food advertisements did not pass the criteria for permitted advertising according to the WHO profile model.

CONCLUSIONS

Evidence from Slovenia shows that, in the absence of efficient regulatory marketing restrictions, television advertising of food to children is almost exclusively linked to energy-dense foods. Minor modifications of the proposed WHO nutrient profile model are suggested.

Sponsorship of junior sport development programs in Australia

OBJECTIVE

This study investigated the nature and extent of unhealthy food, beverage, alcohol and gambling sponsors of children's sport development programs.

METHODS

Websites of junior development sport programs (n=56) associated with sporting organisations that received funding from the Australian Sporting Commission were analysed. Sponsors were considered unhealthy if they were alcohol or gambling companies or sold food and/or beverages that failed independent nutrition criteria. The websites of the sport development programs were also analysed for types of promotion.

RESULTS

There were 246 sponsors identified. Eleven (4.5%) sponsors were food, beverage, alcohol or gambling companies of which 10 (91%) were unhealthy. Surf Lifesaving (n=4) and athletics (n=3) websites had the highest number of unhealthy sponsors. Promotions associated with unhealthy sponsorship included logo placement on homepages (100%), naming rights (31%), logo on sport uniforms (27%) and branded participant packs (31%).

CONCLUSIONS

The majority of food and beverage company sponsors in sport development programs are companies associated with unhealthy products. Two websites hosting junior development program information included an alcohol company sponsor and a gambling company sponsor.

IMPLICATIONS

Unhealthy product sponsorship of children's sport should be addressed as part of a comprehensive regulation designed to reduce exposure to marketing of unhealthy foods.