Elsevier

Public Health

Volume 148, July 2017, Pages 120-128
Public Health

Original Research
Social media indicators of the food environment and state health outcomes

https://doi.org/10.1016/j.puhe.2017.03.013Get rights and content

Highlights

  • We constructed indicators of food type and caloric density of foods from tweets.

  • We utilized Yelp data to track popularity of cuisines.

  • High caloric foods were associated with higher mortality and chronic conditions.

  • Cafes were associated with lower mortality and lower chronic conditions.

  • Alcohol tweets and bars were linked with lower mortality but higher excessive drinking.

Abstract

Objectives

Contextual factors can influence health through exposures to health-promoting and risk-inducing factors. The aim of this study was to (1) build, from geotagged Twitter and Yelp data, a national food environment database and (2) to test associations between state food environment indicators and health outcomes.

Study design

This is a cross-sectional study based upon secondary analyses of publicly available data.

Methods

Using Twitter's Streaming Application Programming Interface (API), we collected and processed 4,041,521 food-related, geotagged tweets between April 2015 and March 2016. Using Yelp's Search API, we collected data on 505,554 unique food-related businesses. In linear regression models, we examined associations between food environment characteristics and state-level health outcomes, controlling for state-level differences in age, percent non-Hispanic white, and median household income.

Results

A one standard deviation increase in caloric density of food tweets was related to higher all-cause mortality (+46.50 per 100,000), diabetes (+0.75%), obesity (+1.78%), high cholesterol (+1.40%), and fair/poor self-rated health (2.01%). More burger Yelp listings were related to higher prevalence of diabetes (+0.55%), obesity (1.35%), and fair/poor self-rated health (1.12%). More alcohol tweets and Yelp bars and pub listings were related to higher state-level binge drinking and heavy drinking, but lower mortality and lower percent reporting fair/poor self-rated health. Supplemental analyses with county-level social media indicators and county health outcomes resulted in finding similar but slightly attenuated associations compared to those found at the state level.

Conclusions

Social media can be utilized to create indicators of the food environment that are associated with area-level mortality, health behaviors, and chronic conditions.

Introduction

Food environment characteristics are critical contextual factors affecting how people access food.1 Varying sociocultural conditions and physical features of the environment influence food choices.2 For example, people are concerned about food quality and availability, locations of stores and restaurants, prices, customer service, and operating hours.1, 2 Food environments, which can be characterized by risk factors (such as exposure to high caloric foods) as well as health-promoting factors (such as availability of healthy food stores), can impact health. Higher prevalence of fast food restaurants have been found to be related to higher obesity rates at the state level.3 Another state-level analysis found that higher per capita number of fast food and full-service restaurants and reduced price of meals correlated with higher obesity rates.4 Conversely, areas with prominent access to healthy food outlets enable diets with fresh and healthy food. Studies have documented increased fruit and vegetable consumption5 and lower body mass index6 with more supermarket availability.7 Poor and minority neighborhoods have fewer large supermarkets than wealthy and majority white neighborhoods,8 which may increase health disparities.

Social media, such as Twitter, are online forms of communication where people create content, share information, and engage in social networking. Twitter can be used as a tool to examine individuals' food decision-making and how that is patterned by their built food environment.9 Chen and Yang found that higher numbers of green retailer (grocery stores chains and local fruit and vegetable stores) within a buffered distance of the Twitter user's geotagged location was associated with more healthful food tweets. This significant difference may indicate that people living in healthier food environments may engage in healthier eating behaviors.7 Ghosh and Guha found a strong positive correlation between tweets about high calorie foods/obesity and locations of McDonalds.10 Widener and Li found that disadvantaged areas had fewer positive Twitter references for fruits and vegetables.11

Moreover, social processes may influence health behaviors. Social processes can affect health via (1) the maintenance of norms around healthy behaviors, (2) stimulation of interest in new activities, (3) emotional support for making healthy choices, (4) the dispersal of knowledge about health promotion practices, and (5) political advocacy and collective action around health.12, 13, 14, 15, 16 Ghosh and Guha found obesity-prevention-themed tweets positively correlated with the number of policies related to obesity, nutrition, and physical activity at the state level,10 possibly indicating higher levels of health advocacy in certain areas. Children who live in states with weaker competitive food and beverage laws are at greater risk of being overweight or obese than their peers who live in states with strong laws.17 The social environment can not only offer opportunities for social control, in regulating unhealthy behaviors and facilitating the social learning of healthy behaviors but can also promote risky behaviors. The spread of health behaviors such as food consumption, health screening, smoking, alcohol consumption, drug use, and sleep has been observed to spread through social networks.18, 19, 20, 21

Social media data have also been analyzed to understand how individuals communicate health topics, the popularity of topics, and sentiment towards current health topics (e.g. vaccines).22 For instance, Myslín et al. analyzed tweets to examine sentiment towards various tobacco products and found that hookah and electronic cigarettes were characterized by more positive sentiment than references to traditional tobacco products.23 Social learning theory posits that learning is a cognitive process that occurs in a social context. Views and activities described via social media can help shape perceived norms, attitudes, beliefs, and subsequently behaviors of people. Liking or following alcohol marketing social media pages has been found to be associated with early age at first alcohol use and heavier alcohol consumption among youth.24 Social media have been utilized for health education and behavioral change interventions such as those aimed to increase physical activity and decrease smoking. Social media can be used for health promotion campaigns to provide health information and social support.25, 26 In addition, user-driven websites and applications such as Yelp have emerged to provide a platform for people to post reviews and testimonies of local businesses and services. In 2016, Yelp's mobile app averaged 65 million users per month.27 Yelp reviews have been leveraged to understand patient experiences at health facilities—information which can be utilized to improve quality of care.28 Yelp data can be used to understand the types of food businesses in a community and the popularity of various foods.

In this study, we examine factors related to the food environment. From Twitter data, we obtain indicators of socially modeled eating and drinking behaviors, possibly capturing prevalent norms and preferences around food. From Yelp data, we assess the availability and popularity of cuisines as perceived by visitors to restaurants. The widespread use of the internet and the abundance of openly shared personal opinions with geotagged check-ins at various locations enable researchers to understand area characteristics, which are unique strengths of utilizing social media data over traditional means of data collection.

The present study constructs a national database of food environment indicators from publicly available Twitter and Yelp data. We then test associations between state-level food environment indicators and health outcomes, accounting for differences in state demographic characteristics via census data that may act as potential confounders related to both food environment indicators and health outcomes.

Section snippets

Twitter data collection and spatial join

For approximately one year, from April 2015–March 2016, we utilized Twitter's Streaming Application Programming Interface (API) to continuously collect a random 1% sample of publicly available tweets with latitude and longitude coordinates. In total, we collected 79,848,992 geotagged tweets from 603,363 unique Twitter users in the contiguous United States (excluding Alaska and Hawaii). The median number of tweets per user was four. Each geotagged tweet was assigned to its corresponding state

Results

Table 1 presents descriptive statistics of our social media database. About 5% of tweets were food-related. The average caloric density of food tweets was 234 (per 100 g). About one in six (16%) food tweets included healthy food (lean proteins, fruits, vegetables) references. About 28% of healthy food tweets and 35% of alcohol tweets were categorized as happy (Table 1) when compared to 19% of tweets overall. Regarding Yelp data, about 27% of popular Yelp listings were for international

Discussion

In this study, we utilize innovative data collection and processing techniques to characterize the food environments of states with data from Twitter and Yelp. We do not expect these data to perfectly capture dietary behaviors—because for instance, not everyone tweets all the foods they consume. Nonetheless, these data may be useful indicators of area-level social norms and preferences. Moreover, public posts about certain foods and behaviors may influence the opinions and behaviors of others,

Acknowledgments

The authors would like to thank Drs. Ming Wen, James VanDerslice, and Ken Smith for their guidance and input on this project.

Ethical approval

The study was approved by the University of Utah Institutional Review Board.

Funding

This study is supported by a grant from NIH's Big Data to Knowledge Initiative, 5K01ES025433.

Competing interests

None declared.

References (54)

  • Q.C. Nguyen et al.

    Leveraging geotagged Twitter data to examine neighborhood happiness, diet, and physical activity

    Appl Geogr

    (2016)
  • S. Greenwood et al.

    Social media update 2016

    (2016)
  • X. Chen et al.

    Contextual uncertainties, human mobility, and perceived food environment: the uncertain geographic context problem in food access research

    Am J Public Health

    (2015)
  • J. Maddock

    The relationship between obesity and the prevalence of fast food restaurants: state-level analysis

    Am J Health Promot

    (2004)
  • K. Morland et al.

    The contextual effect of the local food environment on residents' diets: the atherosclerosis risk in communities study

    Am J Public Health

    (2002)
  • A.F. Brown et al.

    The neighborhood food resource environment and the health of residents with chronic conditions

    J General Intern Med

    (2008)
  • K. Morland et al.

    Neighborhood characteristics associated with the location of food stores and food service places

    Am J Prev Med

    (2001)
  • D. Ghosh et al.

    What are we ‘tweeting’ about obesity? Mapping tweets with Topic Modeling and Geographic Information System

    Cartogr Geogr Inf Sci

    (2013)
  • M.M. Ali et al.

    Weight-related behavior among adolescents: the role of peer effects

    PLoS One

    (2011)
  • L.R. Vartanian et al.

    Social models provide a norm of appropriate food intake for young women

    PLoS One

    (2013)
  • L. Berkman et al.

    Social networks, host resistance, and mortality: a nine-year follow-up study of Alameda County residents

    Am J Educ

    (1979)
  • E. Hennessy et al.

    State-level school competitive food and beverage laws are associated with children's weight status

    J Sch Health

    (2014)
  • M.A. Pachucki et al.

    Social network concordance in food choice among spouses, friends, and siblings

    Am J Public Health

    (2011)
  • N.L. Keating et al.

    Minimal social network effects evident in cancer screening behavior

    Cancer

    (2011)
  • J.N. Rosenquist et al.

    The spread of alcohol consumption behavior in a large social network

    Ann Intern Med

    (2010)
  • S.C. Mednick et al.

    The spread of sleep loss influences drug use in adolescent social networks

    PLoS One

    (2010)
  • C.Y. Bahk et al.

    Publicly available online tool facilitates real-time monitoring of vaccine conversations and sentiments

    Health Aff

    (2016)
  • Cited by (34)

    • The public health community's use of social media for policy advocacy: a scoping review and suggestions to advance the field

      2021, Public Health
      Citation Excerpt :

      Social media platforms provide avenues to engage stakeholders and affect change.19,20 Existing research acknowledges that digital and social media can contribute to public health initiatives in a variety of ways,21–23 even if their potential is not yet fully realised.24 Specific to advocacy and policy change, social media's potential has been recognised in health and other domains 15,25,26 and described as increasingly salient.16,20,27

    • A scoping review on the use of machine learning in research on social determinants of health: Trends and research prospects

      2021, SSM - Population Health
      Citation Excerpt :

      A minority of the studies explicitly discussed different reasons the addition of SDH variables improved the prediction (Boerstler & de Figueiredo, 1991; Darvishi et al., 2017; Engchuan et al., 2019; Fu et al., 2007; Li et al., 2019). Several studies used ML for data curation of text (Crossley et al., 2020; Meng et al., 2017; Nguyen et al., 2016a, 2016b, 2017a, 2017b, 2017c; Prayaga et al., 2019; Robson and Boray, 2019), surveys (Adeyinka, Olakunde, & Muhajarine, 2019), maps or street images (Hu et al., 2009; Larkin & Hystad, 2019; Nguyen et al., 2017a; Suel et al., 2019), and accelerometer data (Brondeel, Pannier, & Chaix, 2016). These studies highlighted practical ML applications, e.g., estimating health literacy from the text at the individual patient level (Crossley et al., 2020) or reducing socioeconomic inequalities in medication refills (Prayaga et al., 2019).

    • Describing and mapping diversity and accessibility of the urban food environment with open data and tools

      2020, Applied Geography
      Citation Excerpt :

      Open, crowdsourced food outlet data available on sites such as OpenStreetMap (OSM) and Yelp are current, free, easily accessible with basic coding skills, and geo-located. These data have been used in a variety of urban research projects (Gomez-Lopez et al. 2017; Jiang & Thill, 2015; Martí et al. 2019) including to create indicators of the food environment (Nguyen et al. 2017). They help characterise urban food access using fewer resources than traditional data collection and without the cost associated with commercial business directories (Martí et al. 2019).

    • Model for prediction of the carbonyl value of frying oil from the initial composition

      2020, LWT
      Citation Excerpt :

      Deep frying is a common and popular food processing method. With the popularity of fried foods around the world, the quality and safety of fried food has received widespread attention (Gadiraju, Patel, Gaziano, & Djoussé, 2015; Nguyen et al., 2017). The quality of fried food depends heavily on the quality of the frying oil because the original fats in the food are almost completely replaced by frying oil (Gertz, Aladedunye, & Matthäus, 2017).

    View all citing articles on Scopus
    View full text