Using Natural Language Processing to Explore Social Media Opinions on Food Security: Sentiment Analysis and Topic Modeling Study

BACKGROUND

Social media has the potential to be of great value in understanding patterns in public health using large-scale analysis approaches (eg, data science and natural language processing [NLP]), 2 of which have been used in public health: sentiment analysis and topic modeling; however, their use in the area of food security and public health nutrition is limited.

OBJECTIVE

This study aims to explore the potential use of NLP tools to gather insights from real-world social media data on the public health issue of food security.

METHODS

A search strategy for obtaining tweets was developed using food security terms. Tweets were collected using the Twitter application programming interface from January 1, 2019, to December 31, 2021, filtered for Australia-based users only. Sentiment analysis of the tweets was performed using the Valence Aware Dictionary and Sentiment Reasoner. Topic modeling exploring the content of tweets was conducted using latent Dirichlet allocation with BigML (BigML, Inc). Sentiment, topic, and engagement (the sum of likes, retweets, quotations, and replies) were compared across years.

RESULTS

In total, 38,070 tweets were collected from 14,880 Twitter users. Overall, the sentiment when discussing food security was positive, although this varied across the 3 years. Positive sentiment remained higher during the COVID-19 lockdown periods in Australia. The topic model contained 10 topics (in order from highest to lowest probability in the data set): "Global production," "Food insecurity and health," "Use of food banks," "Giving to food banks," "Family poverty," "Food relief provision," "Global food insecurity," "Climate change," "Australian food insecurity," and "Human rights." The topic "Giving to food banks," which focused on support and donation, had the highest proportion of positive sentiment, and "Global food insecurity," which covered food insecurity prevalence worldwide, had the highest proportion of negative sentiment. When compared with news, there were some events, such as COVID-19 support payment introduction and bushfires across Australia, that were associated with high periods of positive or negative sentiment. Topics related to food insecurity prevalence, poverty, and food relief in Australia were not consistently more prominent during the COVID-19 pandemic than before the pandemic. Negative tweets received substantially higher engagement across 2019 and 2020. There was no clear relationship between topics that were more likely to be positive or negative and have higher or lower engagement, indicating that the identified topics are discrete issues.

CONCLUSIONS

In this study, we demonstrated the potential use of sentiment analysis and topic modeling to explore evolution in conversations on food security using social media data. Future use of NLP in food security requires the context of and interpretation by public health experts and the use of broader data sets, with the potential to track dimensions or events related to food security to inform evidence-based decision-making in this area.

Re-Licious: Co-Design with Adolescents to Turn Leftovers into Delicious and Healthy Meals-A School-Based Pilot Intervention

One-third of the food produced globally is lost or wasted, and one cause is consumer leftovers. Re-licious was an eight-week pilot intervention aiming to increase awareness of food waste and healthy eating by building adolescents' ability to prepare and cook leftovers. Re-licious used a co-design approach and was piloted in a secondary school, half of which was during a COVID-19 lockdown period. Students watched videos on food waste and healthy eating during class. They identified leftover ingredients at home and repurposed ingredients to create recipes. Students co-created recipe criteria to ensure the personal relevance of the recipes. They completed pre- and post-intervention questionnaires (n = 40) about food waste and motivation and interest in healthy eating. Four group interviews were conducted. The factors identified as important in the co-creation sessions were preparation time, cost, healthiness, and sustainability. Participants with low motivation and interest in healthy eating decreased, and participants with high interest increased (p < 0.001). The intention to reduce food waste increased (p = 0.007), as did resourcefulness (p < 0.001) and personal norms (p = 0.048). Interviews highlighted the students' increased awareness of food waste and enjoyment of the intervention. With improvements based on this pilot, Re-licious could be adapted and re-trialled in a face-to-face format to educate young people about food waste.

The use of sentiment and emotion analysis and data science to assess the language of nutrition, food and cooking related content on social media: A systematic scoping review

Social media data is rapidly evolving and accessible which presents opportunities for research. Data science techniques, such as sentiment or emotion analysis which analyse textual emotion, provide an opportunity to gather insight from social media. This paper describes a systematic scoping review of interdisciplinary evidence to explore how sentiment or emotion analysis methods alongside other data science methods have been used to examine nutrition, food and cooking social media content. A PRISMA search strategy was used to search nine electronic databases in November 2020 and January 2022. Of 7325 studies identified, 36 studies were selected from 17 countries, and content was analysed thematically and summarised in an evidence table. Studies were published between 2014-2022 and used data from seven different social media platforms (Twitter, YouTube, Instagram, Reddit, Pinterest, Sina Weibo and mixed platforms). Five themes of research were identified including Dietary patterns, Cooking and recipes, Diet and Health, Public Health and Nutrition and Food in general. Papers developed a sentiment or emotion analysis tool or used available open-source tools. Accuracy to predict sentiment ranged from 33.33% (open-source engine) to 98.53% (engine developed for the study). The average proportion of sentiment was 38.8% positive, 46.6% neutral and 28.0% negative. Additional data science techniques used included topic modelling and network analysis. Future research requires optimising data extraction processes from social media platforms, the use of interdisciplinary teams to develop suitable and accurate methods for the subject and the use of complementary methods to gather deeper insights into these complex data.

Psycho-Behavioural Segmentation in Food and Nutrition: A Systematic Scoping Review of the Literature

Inadequate dietary intakes are a key modifiable risk factor to reduce the risk of developing non-communicable diseases. To encourage healthy eating and behaviour change, innovative public health interventions are required. Social marketing, in particular segmentation, can be used to understand and target specific population groups. However, segmentation often uses demographic factors, ignoring the reasons behind why people behave the way they do. This review aims to explore the food and nutrition related research that has utilised psycho-behavioural segmentation. Six databases from were searched in June 2020. Inclusion criteria were: published 2010 onwards, segmentation by psycho-behavioural variables, outcome related to food or nutrition, and healthy adult population over 18 years. 30 studies were included; most were quantitative (n = 28) and all studies used post-hoc segmentation methods, with the tools used to segment the population varying. None of the segments generated were targeted in future research. Psycho-behavioural factors are key in understanding people's behaviour. However, when used in post-hoc segmentation, do not allow for effective targeting as there is no prior understanding of behaviours that need to change within each segment. In future, we should move towards hybrid segmentation to assist with the design of interventions that target behaviours such as healthy eating.

Beyond Body Weight: Design and Validation of Psycho-Behavioural Living and Eating for Health Segments (LEHS) Profiles for Social Marketing

Obesity, sedentary behaviour, and poor dietary habits amongst young adults are growing concerns, with this age group being in a worse state of health and nutrition than adolescents and adults. This paper presents the procedures for establishing a new instrument for defining behaviours in relation to healthy lifestyle and food choices amongst young adults (Living and Eating for Health Segments: LEHS). The aim of this paper is to outline the instrument design protocol for external validation and to permit replication in other studies. The instrument design process used a multi-step social marketing instrument design method. This approach has previously been used in designing valid and reliable measures in marketing and consumer research, including social marketing. The protocol established six psycho-behavioural LEHS profiles for young adults. These profiles are: Lifestyle Mavens (15.4%), Aspirational Healthy Eaters (27.5%), Balanced-all Rounders (21.4%), Health Conscious (21.1%), Contemplating Another Day (11.2%), and Blissfully Unconcerned (3.4%). Each of these profiles provided insights into psycho-behavioural characteristics that can be used in designing apposite social media social marketing campaigns.

Assessing the Credibility and Authenticity of Social Media Content for Applications in Health Communication: Scoping Review

Background

Nutrition science is currently facing issues regarding the public’s perception of its credibility, with social media (SM) influencers increasingly becoming a key source for nutrition-related information with high engagement rates. Source credibility and, to an extent, authenticity have been widely studied in marketing and communications but have not yet been considered in the context of nutrition or health communication. Thus, an investigation into the factors that impact perceived source and message credibility and authenticity is of interest to inform health communication on SM.

Objective

This study aims to explore the factors that impact message and source credibility (which includes trustworthiness and expertise) or authenticity judgments on SM platforms to better inform nutrition science SM communication best practices.

Methods

A total of 6 databases across a variety of disciplines were searched in March 2019. The inclusion criteria were experimental studies, studies focusing on microblogs, studies focusing on healthy adult populations, and studies focusing on either source credibility or authenticity. Exclusion criteria were studies involving participants aged under 18 years and clinical populations, gray literature, blogs, WeChat conversations, web-based reviews, non-English papers, and studies not involving participants’ perceptions.

Results

Overall, 22 eligible papers were included, giving a total of 25 research studies. Among these studies, Facebook and Twitter were the most common SM platforms investigated. The most effective communication style differed depending on the SM platform. Factors reported to impact credibility included language used online, expertise heuristics, and bandwagon heuristics. No papers were found that assessed authenticity.

Conclusions

Credibility and authenticity are important concepts studied extensively in the marketing and communications disciplines; however, further research is required in a health context. Instagram is a less-researched platform in comparison with Facebook and Twitter.

Strategies to Improve Health Communication: Can Health Professionals be Heroes?

Communicating evidence-based nutrition messages to the public is challenging and is often in conflict with popular opinions, particularly from social media influencers (SMIs). In order to increase engagement with nutrition professionals (NPs) on social media, we aimed to explore young adults' perceptions of the authenticity and trustworthiness of SMIs and NPs Instagram posts. A cross-sectional questionnaire was administered to students (n = 149) from an Australian University. Participants viewed a real-life Instagram profile and one post from both a NP and a SMI. Main outcomes were post authenticity and trustworthiness, and emotional message appeals measured on five-point Likert scales. Regression models were developed to assess whose post (the NP or SMI) was perceived to be more authentic and trustworthy. Participants were young adults (median age (25th, 75th percentiles): 20 (19,21)), with approximately half identifying as female. A high heroic message appeal (+1SD above mean) significantly increased the perceived authenticity of the NPs post only (p = 0.01). Post authenticity enhanced post trustworthiness, but only when a heroic message appeal was used by the NP. When appropriate, NPs should convey positive emotions such as bravery and success to enhance the authenticity and trustworthiness of their posts.

Changes in bone turnover following gonadotropin-releasing hormone (GnRH) agonist administration and estrogen treatment in cynomolgus monkeys: a short-term model for evaluation of antiresorptive therapy

In this study we determine the early time course of estrogen deficiency-induced bone loss in the cynomolgus monkey and examine the potential of this method for evaluating antiresorptive therapies. In two groups of animals, estrogen deficiency was induced by the administration of a gonadotropin-releasing hormone agonist (GnRHa) and bone turnover was measured using biochemical markers. Two weeks after receiving GnRHa, serum estradiol decreased to below the detection limit in most animals and remained there through 6 months or until estrogen replacement started (months 4-6). Relative to untreated animals, urinary deoxypyridinoline (dPyr), as well as C- and N-telopeptides of type I collagen, were significantly elevated 4 weeks after receiving GnRHa. Serum osteocalcin increased in GnRHa-treated animals as early as week 4 and the level was significantly higher than in untreated control animals from weeks 8-24. Estradiol treatment returned all measures of bone turnover to control levels within 2 weeks. The use of biochemical markers as surrogates of bone turnover and loss was validated by measurement of bone mineral density (BMD), which showed a significant reduction at 6 months in estrogen-deficient animals. However, lumbar BMD in animals that received GnRHa and estradiol was similar to that in animals that had not received GnRHa. In conclusion, a monthly depot injection of GnRHa resulted in increased bone turnover due to estrogen deficiency, as early as 4 weeks after treatment. Estrogen administration returned bone turnover to control levels in 2 weeks. This method represents a valid model for evaluating antiresorptive agents in the short term in a nonhuman primate. Furthermore, the data suggest that changes in biochemical markers in response to antiresorptive therapy in humans may be detectable at much earlier timepoints than commonly used.

Development of a women's wellness center in Almaty, Kazakhstan

Twelve Women's Wellness Centers were established by American International Health Alliance (AIHA) partnerships in the New Independent States of the former Soviet Union. The services offered by these centers focused on health promotion, disease prevention, and other health and social issues that affect women's health. The innovations in the model include patient-centered care that emphasizes patient choice and the introduction of the role of the nurse educator.

The effect of high-efficiency and standard vacuum-cleaners on mite, cat and dog allergen levels and clinical progress

The major triggers for allergic asthma are exposure to allergens of the house dust mite, Dermatophagoides pteronyssinus, and of pets. Unfortunately studies of techniques designed to reduce house dust mite and pet allergens have had mixed results. However, new so-called 'improved' products continue to appear on the market and require subjective evaluation. The homes of 60 house dust mite-allergic patients were studied to compare the effects of high-efficiency and standard vacuum-cleaners on allergen concentration. Der p 1 (house dust mite), Fel d 1 (cat) and Can f 1 (dog) allergens were measured in four separate locations in each home. Clinical analysis was by lung function, bronchodilator usage and histamine challenge techniques. There was a significant reduction in Fel d 1 (ng/m2) in dust samples from the living-room carpet (p = 0.046), bedroom carpet (p = 0.003) and mattress (p = 0.013) and living-room sofa (p = 0.005) after 12 months of using the high-efficiency cleaners, but only in the mattress sample using the standard cleaners (p = 0.014). Can f 1 (ng/g dust) was reduced in the mattress sample after using the high-efficiency vacuum-cleaners (p = 0.028), but not at other sites. Der p 1 levels were not significantly changed over this period. Clinically, patients in the high-efficiency group showed improvements in peak expiratory flow rate (PEFR) (p = 0.004), FEV1 (p = 0.026) and bronchodilator usage (p = 0.005) after 12 months. When the cat-sensitive patients were analyzed separately, improvements in histamine PC20 (p = 0.039) were also seen. Reducing Fel d 1 concentrations, in the absence of any change in Der p 1 concentrations, can produce significant improvements in the lung function of atopic, asthmatic patients. This effect was primarily achieved in those patients with cat sensitivity, but who did not possess a cat themselves.

Homicide against women in the workplace

OBJECTIVES

To examine the nature and magnitude of violence against women in the workplace.

METHOD

Data from the National Traumatic Occupational Fatalities surveillance system were analyzed for the 13-year period, 1980 to 1992. Employment information was coded from industry and occupation narratives taken from the death certificates. Rates were calculated using annual average employment data published by the Bureau of Labor Statistics.

RESULTS

From 1980 to 1992, 2,001 women and 7,935 men were victims of work-related homicides, for rates of 0.32 and 1.01 per 100,000 female and male workers, respectively. The majority of female homicide victims were employed in two industries-retail trade (46%) and services (22%). The highest risk industry was grocery stores (1.74 per 100,000 workers), followed by eating and drinking establishments, hotels and motels, and justice and public order.

CONCLUSIONS

The findings presented here are consistent with previous research indicating that homicide is the leading cause of occupational injury death for women, but that women are at lower risk than their male counterparts. Homicide played a major role in sales and service, sectors that have increased over the last several decades and in which future growth is predicted. Feasible and effective prevention strategies to reduce the risk of work-related homicides must be developed and implemented.

Workplace homicide: industries and occupations at high risk

Homicide is to blame for 20 workplace deaths each week. Although no single intervention strategy will be appropriate in all situations, the author points out that interventions cannot be designed without knowledge of the demographic characteristics of victims and the distribution of workplace violence across industries and occupations. Such data are presented by gender, age, race, geographic distribution, method of homicide, and industry and occupation.

Occupational injury mortality rates in the United States: changes from 1980 to 1989

Changes in occupational injury mortality rates over the 1980s were examined through analysis of the National Traumatic Occupational Fatalities surveillance system. The US occupational injury mortality rate decreased 37% over the decade, with decreases seen in nearly every demographic and employment sector. Greater declines were among men, Blacks, and younger workers, as well as among agricultural, trade, and service workers. Electrocutions, machine-related incidents, and homicides showed the greatest decreases. Changes in occupational mortality rates by demography, industry, and cause of death indicate the areas in which the most progress has been made and those that are prime targets for prevention efforts.

Workplace-related homicide among health care workers in the United States, 1980 through 1990

OBJECTIVE

To improve understanding of the epidemiology of fatal violence directed toward physicians and other health care workers (HCWs) in health care settings.

DESIGN

Analyses of data for 1980 through 1990 from the National Traumatic Occupational Fatalities surveillance system.

MAIN OUTCOME MEASURES

Overall occurrence of occupational injury deaths and occurrence of workplace-related homicides among HCWs.

RESULTS

From 1980 through 1990, a total of 522 HCWs died from injuries sustained while working. The most common causes of death were motor vehicle crashes (122 [23.4%]), homicide (106 [20.3%]), and suicide (88 [16.9%]). Firearms were used in the greatest number (78 [73.6%]) of workplace-related homicides among HCWs.

CONCLUSIONS

These findings highlight the need for strengthened surveillance and more accurate estimates of the risks of workplace-related violent injury for HCWs in the United States.

Occupational injury deaths among females. The US experience for the decade 1980 to 1989

From 1980 through 1989, females accounted for 44% of the total employed population. Thus, occupational safety and health issues specific to the experience of women merit consideration. Research has demonstrated that the occupational fatality experience of females is not adequately described by the group of all workers. The leading cause of death for all workers is motor vehicle incidents, while the leading cause of occupational injury death of females is homicide. The National Institute for Occupational Safety and Health (NIOSH) has compiled a decade of data on the fatal occupational injury experience of US workers, providing a sufficient number of female cases to allow separate analyses. Over the decade, 3821 females died as a result of injuries sustained at work, with an average annual fatality rate of 0.82/100,000 female workers. Among industries, retail trade and services accounted for nearly half of all occupational injury deaths to females. The detailed occupations with the highest rates of work-related injury death were airplane pilots and navigators, drivers of heavy trucks, construction laborers, and police and detectives. Information on the causes of work-related injury death by occupation is fundamental to the prevention of these deaths. The causes of death in the highest-risk occupations included aircraft crashes, motor vehicle collisions, pedestrians struck by motor vehicles, and homicides by firearms. These data provide a foundation for the prevention of occupational injury deaths among females in the United States.

Industries and occupations at high risk for work-related homicide

Homicide is the third leading cause of injury death in the workplace. The death certificate-based National Traumatic Occupational Fatalities surveillance system and estimates of annual employment were used to calculate average annual rates of work-related homicide for detailed industries and occupations for the nation for 1980 to 1989. Workers in the taxicab industry had the highest rate of work-related homicide (26.9 per 100,000 workers). High rates were also identified for workers providing public and private security, and in a number of retail trade and service industries. For many high-risk industries, the risk was excessive for male workers only. Differences between rates for black and nonblack workers varied across industries and occupations. Immediate efforts to protect workers, and long-term efforts to describe and study work-related homicide thoroughly and to evaluate interventions are needed.

Elderly persons with mental retardation: an exceptional population with special needs

The passage of recent legislation has ensured special service provisions for a previously overlooked population, elderly persons with mental retardation. This article presents a broad overview of current research studies on this topic and focuses on identification of the population, residential options, and service needs of elderly persons with mental retardation. Similarities and differences between older persons with developmental disabilities and the non-retarded elderly, and service need requirements of this particular population are also included. Finally, the importance of developing collaborative services of developmental disability agencies and aging services is discussed as well as recommendations for future research needs.

Homicide in the workplace: the U.S. experience, 1980-1988

1. Homicide was the third leading cause of occupational injury death in the United States from 1980 to 1988. 2. Of all occupational injury deaths to females from 1980 to 1988, 40% were homicides. 3. The highest age specific rates of occupational homicide were found among workers 65 years and older (2.04 per 100,000 workers). 4. Protection of workers will require the collaborative work of public health and safety professionals, including occupational health nurses, as well as law enforcement, engineering, and architecture professionals and behavioral scientists.

Capillary column gas chromatography with nitrogen-phosphorus detection for determination of nitrogen- and phosphorus-containing pesticides in finished drinking waters: collaborative study

A joint U.S. Environmental Protection Agency/AOAC interlaboratory method validation study was conducted on EPA Method 507, Determination of Nitrogen- and Phosphorus-Containing Pesticides in Finished Drinking Water by Gas Chromatography with a Nitrogen-Phosphorus Detector, to determine the mean recovery and precision for analyses of 45 nitrogen- or phosphorus-containing pesticides in reagent water and finished drinking waters. The study design was based on Youden's nonreplicate plan for collaborative tests of analytical methods. The waters were spiked with 45 nitrogen- or phosphorus-containing pesticides at 6 concentration levels, prepared as 3 Youden pairs. Ten volunteer laboratories extracted the spiked test waters with methylene chloride, performed a solvent exchange with methyl tert-butyl ether, and analyzed an aliquot of each extract by gas chromatography using a nitrogen-phosphorus detector. Results were analyzed using an EPA computer program, which measured recovery and precision for each of the 45 pesticides and compared the performance of the method between water types. Method 507 was judged acceptable for all analytes tested except merphos, which thermally decomposed in the injection port of the gas chromatograph. Five compounds (carboxin, disulfoton, metolachlor, pronamide, and simazine) exhibited statistically significant matrix effects for the finished drinking water. The method has been adopted official first action by AOAC.