Maintenance Notice

Due to necessary scheduled maintenance, the JMIR Publications website will be unavailable from Monday, March 11, 2019 at 4:00 PM to 4:30 PM EST. We apologize in advance for any inconvenience this may cause you.

Who will be affected?

Advertisement

Citing this Article

Right click to copy or hit: ctrl+c (cmd+c on mac)

Published on 01.04.19 in Vol 5, No 2 (2019): Apr-Jun

Preprints (earlier versions) of this paper are available at http://preprints.jmir.org/preprint/11477, first published Jul 03, 2018.

This paper is in the following e-collection/theme issue:

    Original Paper

    Wet Markets and Food Safety: TripAdvisor for Improved Global Digital Surveillance

    1Massachusetts Institute of Technology, Cambridge, MA, United States

    2Institute of Global Health, Faculty of Medicine, University of Geneva, Geneva, Switzerland

    3Institute for Environmental Sciences, University of Geneva, Geneva, Switzerland

    4Division of Tropical and Humanitarian Medicine, University Hospitals of Geneva, Geneva, Switzerland

    5Citizen Cyberlab, Centre Universitaire d’Informatique, University of Geneva, Carouge, Switzerland

    6Digital Epidemiology Lab, École Polytechnique Fédérale de Lausanne, Lausanne, Switzerland

    Corresponding Author:

    Rafael Ruiz de Castañeda, PhD

    Institute of Global Health

    Faculty of Medicine

    University of Geneva

    Campus Biotech

    Chemin des Mines 9

    Geneva, 1202

    Switzerland

    Phone: 41 223790951

    Email:


    ABSTRACT

    Background: Wet markets are markets selling fresh meat and produce. Wet markets are critical for food security and sustainable development in their respective regions. Due to their cultural significance, they attract numerous visitors and consequently generate tourist-geared information on the Web (ie, on social networks such as TripAdvisor). These data can be used to create a novel, international wet market inventory to support epidemiological surveillance and control in such settings, which are often associated with negative health outcomes.

    Objective: Using social network data, we aimed to assess the level of wet markets’ touristic importance on the Web, produce the first distribution map of wet markets of touristic interest, and identify common diseases facing visitors in these settings.

    Methods: A Google search was performed on 31 food market–related keywords, with the first 150 results for each keyword evaluated based on their relevance to tourism. Of all these queries, wet market had the highest number of tourism-related Google Search results; among these, TripAdvisor was the most frequently-occurring travel information aggregator, prompting its selection as the data source for this study. A Web scraping tool (ParseHub) was used to extract wet market names, locations, and reviews from TripAdvisor. The latter were searched for disease-related content, which enabled assignment of GeoSentinel diagnosis codes to each. This syndromic categorization was overlaid onto a mapping of wet market locations. Regional prevalence of the most commonly occurring symptom group - food poisoning - was then determined (ie, by dividing the number of wet markets per continent with more than or equal to 1 review containing this syndrome by the total number of wet markets on that continent with syndromic information).

    Results: Of the 1090 hits on TripAdvisor for wet market, 36.06% (393/1090) conformed to the query’s definition; wet markets were heterogeneously distributed: Asia concentrated 62.6% (246/393) of them, Europe 19.3% (76/393), North America 7.9% (31/393), Oceania 5.1% (20/393), Africa 3.1% (12/393), and South America 2.0% (8/393). Syndromic information was available for 14.5% (57/393) of wet markets. The most frequently occurring syndrome among visitors to these wet markets was food poisoning, accounting for 54% (51/95) of diagnoses. Cases of this syndrome were identified in 56% (22/39) of wet markets with syndromic information in Asia, 71% (5/7) in Europe, and 71% (5/7) in North America. All wet markets in South America and Oceania reported food poisoning cases, but the number of reviews with syndromic information was very limited in these regions (n=2).

    Conclusions: The map produced illustrates the potential role of touristically relevant social network data to support global epidemiological surveillance. This includes the possibility to approximate the global distribution of wet markets and to identify diseases (ie, food poisoning) that are most prevalent in such settings.

    JMIR Public Health Surveill 2019;5(2):e11477

    doi:10.2196/11477

    KEYWORDS



    Introduction

    Background

    Traditional food markets (ie, wet markets) play important roles in food security and local development [1]; however, they also have negative health implications. In 2003, severe acute respiratory syndrome spread globally from a Chinese wet market, causing hundreds of deaths [2] and major economic losses [3]. Avian influenza has also been repeatedly associated with wet markets [1,4]. Foodborne Campylobacter, Salmonella, Giardia, and Escherichia are most common in these settings, leading to 18 million disability-adjusted life years annually [5]. These are particularly important in low- and middle-income countries (LMICs), but their true impact is unknown given that many episodes go unreported [6].

    Due to wet markets’ cultural importance, there exists extensive tourist-geared information on the Web, often on websites such as Yelp and TripAdvisor, which serve as forums to share experiences. Although the use of these data remains unexploited for wet markets, it has yielded compelling results for restaurants. For example, iwaspoisoned.com [7] serves as a platform where individuals can report symptoms of food poisoning alongside the offending eatery. Through citizen participation, the website has identified several foodborne disease outbreaks before traditional epidemiological methods [8].

    Objectives

    This approach shaped the objectives of this study, which were to show the link between wet markets and tourism as well as to exploit tourist-generated social network data to create the first map of the distribution of wet markets of touristic interest and their associated adverse health events. Rather than function as an epidemiological analysis in which foodborne disease incidence related to wet market visits is calculated, our approach aims to showcase the potential of Web-based social networks to pick up potentially overlooked instances of disease.


    Methods

    Establishing the Data Source

    A Google search was performed on different food market types (see Multimedia Appendix 1). For each keyword, the first 150 results were scraped and characterized based on their relevance to tourism (ie, presence of tourist-geared content and promotion of an area to potential visitors). Of all food market types, wet market was linked to the highest proportion of tourism-related websites—59.3% (89/150) of Google Search results. This connection motivated its use in this study. Any social networking websites were flagged, and their touristic importance was assessed. Of these, TripAdvisor appeared most frequently and had the most comprehensive wet market–related information, prompting its selection as this study’s data source.

    Web Scraping

    The term wet market was inputted into TripAdvisor (July 2017). Wet market names and locations were harvested using ParseHub [9]. Irrelevant results (eg, waterparks with wet in their names) were removed manually. A Python 2.7 script integrating geocoding library GeoPy [10] was developed to convert wet market locations into geographic coordinates for mapping.

    Text Mining

    For each wet market, the TripAdvisorReviews section was parsed for mentions of keywords most often associated with foodborne disease, a list corroborated by the National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases Web page on foodborne illness [11]: diarrhea, vomit/vomiting/vomited, foodpoisoning, stomach ache, headache, nausea/nauseous, upset stomach, sick, ill, and dizzy. Comments containing at least one of these were manually extracted. When possible, comments were assigned a GeoSentinel diagnosis code [12] through medical expert analysis using syndrome keywords and indicators of symptom duration. Wet market and syndromic distribution were analyzed using descriptive statistics.

    The study represents a passive analysis of information on TripAdvisor. The investigators did not participate in nor were involved in TripAdvisor communications (ie, there was no posting), so the analysis should not be considered active internet-based research requiring human subject consent.


    Results

    Global Wet Market Distribution

    Wet market yielded 1090 attractions on TripAdvisor, 36.06% (393/1090) of which aligned with the term’s definition. Mapping revealed Asia as the region with the greatest wet market density, accounting for 62.6% (246/393) of wet markets reviewed (Figure 1). The second-most wet market–dense region was Europe with 19.3% (76/393) of wet markets reviewed, followed by North America with 7.9% (31/393), Oceania with 5.1% (20/393), Africa with 3.1% (12/393), and South America with 2.0% (8/393) .

    Figure 1. Locations of TripAdvisor-sourced wet markets, with zoomed inset in Southeast Asia. Red circles represent wet markets where visitors reported adverse health events, whereas empty circles denote locations where such reports were lacking. The background country border is sourced from Natural Earth vector data projected in the World Robinson coordinate reference system using Quantum Geographic Information System (QGIS) 2.18.2.
    View this figure

    Assessment of Syndromic Information

    All 393 wet markets were reviewed on TripAdvisor. Moreover, 14.5% (57/393) of these contained at least one review with syndromic information and were broken down regionally as 68% (39/57) Asian, 12% (7/57) European, 12% North American (7/57), 3% South American (2/57), and 3% Oceanic (2/57). In total, these yielded 98 reviews with syndromic information, of which Asian markets accounted for 74% (73/98). Acute gastroenteritis <12 hrs, food poisoning was the most common diagnosis in wet markets globally 54% (51/95) syndrome references, after discarding 3 reviews corresponding to unascertainable illness; Figure 2). Asia, where the review volume was particularly noticeable, had 56% (22/39) of wet markets with this specific syndrome. The syndromes Acute gastroenteritis > 12 hrs, Acute gastroenteritis <12 hrs, and Diarrhea, acute unspecified comprised 95% (90/95) non-N/A diagnoses from the 57 wet markets.

    Figure 2. Frequency distribution of 98 TripAdvisor reviewer diagnoses following GeoSentinel encoding. ac=acute; gastro=gastroenteritis; IBS=irritable bowel syndrome; N/A=symptoms present but illness unascertainable; resp tract inf=respiratory tract infection; unsp=unspecified.
    View this figure

    Discussion

    Principal Findings

    This study shows the touristic importance of wet markets and provides, to our knowledge, the first global map of wet markets’ locations and related syndromic information communicated on TripAdvisor by their visitors. Our map reveals that wet markets are heterogeneously distributed across continents. Asia is a hotspot, both in the number of markets and the absolute number of adverse associated health events. When controlling by the number of continental wet markets with syndromic information, South America and Oceania exhibited the highest proportion of GeoSentinel acute gastroenteritis <12 hrs, food poisoning diagnoses. However, both these regions were wet market–sparse (n=2 wet markets with syndromic information in each) compared with a continent like Asia.

    Use of region-specific terminology for designating markets could have yielded denser mapping and should be used in future work. Furthermore, the choice to look only at TripAdvisor reviews in English may have impacted the wet market distribution shown in the map. An example is the density of wet markets, which seems to be in preferred destinations of English-speaking tourists; another example is that the concentration of wet markets in Africa is largely in English-speaking countries. This could prompt search on the platform using other languages to enhance the results.

    Little is known about the wet market health threats to locals and travelers partly because of underreporting of adverse health events experienced by visitors [13]. Although social networking platforms do not paint a complete picture of wet markets (eg, total number of regional wet markets and how many yearly visitors), they can provide opportunities to instantaneously report and detect these elusive cases—primarily traveler-related ones—and improve epidemiological monitoring within these settings.

    Limitations and Strengths

    Reviews from Yelp, a TripAdvisor-like platform connecting citizens with local businesses, have been leveraged by New York public health authorities to detect restaurant-related foodborne disease events [14]; interestingly, fewer than 2% of individuals with an alleged illness explicitly mention reporting their case to a medical professional [14]. Twitter has been implemented in St. Louis, Missouri, to detect food poisoning cases [15], but its potential is restricted by the short post length. The website iwaspoisoned.com is equally notable in its effort to crowdsource information, though it is limited in scope outside the United States. Wet markets are more challenging establishments because of their typical location in LMICs and their operation under often-limited regulations. This necessitates tapping into other Web-based communities—social platforms such as TripAdvisor— to glean information on visitor health.

    Traveler’s diarrhea is a common disorder affecting tourists visiting developing countries [16] and is generally associated with consumption of foods prepared under unhygienic conditions (common in wet markets). Our analysis shows that acute gastroenteritis (food poisoning) and diarrhea were the most frequent illnesses among wet market visitors. However, this must be carefully considered from medical and epidemiological perspectives as the result is based on rarely corroborated, Web-based descriptions. We also cannot exclude that reported symptoms could have been caused by an event before or after (but unrelated to) a wet market visit.

    Conclusions

    As use of TripAdvisor in isolation has accounted for only a fraction of wet markets of touristic importance and their associated health risks, larger datasets are needed to confirm the results presented here and to explore others. Even so, the pipeline we present is significant for travel medicine and epidemiology. It could ultimately contribute to predictive models for improved epidemic forecasting and to the development of diagnostic tools based on syndromic surveillance and artificial intelligence. TripAdvisor could crosstalk with other social networks (eg, Yelp) for maximal information coverage and partner with other initiatives for a more structured collection of wet market–related health information in near-real time. In this way, we can gain an improved understanding of global wet markets and their associated health risks while also ensuring their safer promotion.

    Acknowledgments

    NEK was supported by the MIT International Science and Technology Initiatives and by a Summer School Grant from the Zeno Karl Schindler Foundation. RRdC was supported by Fondation Louis-Jeantet, which the authors thank very much. The authors also thank Prof A Flahault for supporting this work at the Institute of Global Health. The authors thank 2 anonymous reviewers for their constructive comments.

    Authors' Contributions

    RRdC, IB, and NEK designed the study. NR, GA, JLFM, MMM, and SPM guided the implementation. NR and NEK developed the spatial analyses. GA paired GeoSentinel diagnosis codes with TripAdvisor reviews. NEK wrote the initial draft of the manuscript, and all coauthors reviewed and improved it.

    Conflicts of Interest

    None declared.

    Multimedia Appendix 1

    List of market types whose relevance to tourism was assessed in an initial Google Search screen.

    DOCX File, 13KB

    References

    1. World Health Organization. Switzerland: WHO Press; 2006. A guide to healthy food markets   URL: http://www.who.int/foodsafety/publications/capacity/healthymarket_guide.pdf?ua=1 [WebCite Cache]
    2. Wu T, Perrings C, Kinzig A, Collins JP, Minteer BA, Daszak P. Economic growth, urbanization, globalization, and the risks of emerging infectious diseases in China: a review. Ambio 2017 Feb;46(1):18-29 [FREE Full text] [CrossRef] [Medline]
    3. Lee JW, McKibbin WJ. Estimating the global economic costs of SARS. In: Knobler S, Mahmoud A, Lemon S, Mack A, Sivitz L, Oberholtzer K, editors. Learning From SARS: Preparing for the Next Disease Outbreak: Workshop Summary. Washington DC: National Academies Press; 2004.
    4. Wu J, Lu J, Faria NR, Zeng X, Song Y, Zou L, et al. Effect of live poultry market interventions on influenza A(H7N9) virus, Guangdong, China. Emerg Infect Dis 2016 Dec;22(12):2104-2112 [FREE Full text] [CrossRef] [Medline]
    5. Grace D. Food safety in low and middle income countries. Int J Environ Res Public Health 2015 Aug 27;12(9):10490-10507 [FREE Full text] [CrossRef] [Medline]
    6. Bernardo TM, Rajic A, Young I, Robiadek K, Pham MT, Funk JA. Scoping review on search queries and social media for disease surveillance: a chronology of innovation. J Med Internet Res 2013 Jul 18;15(7):e147 [FREE Full text] [CrossRef] [Medline]
    7. Quade P. IWasPoisoned.   URL: https://iwaspoisoned.com/ [accessed 2018-06-27] [WebCite Cache]
    8. Neimark J. NPR. 2017 Nov 30. 'I Was Poisoned': Can Crowdsourcing Food Illnesses Help Stop Outbreaks?   URL: https:/​/www.​npr.org/​sections/​thesalt/​2017/​11/​30/​565769194/​i-was-poisoned-can-crowdsourcing-food-illnesses-help-stop-outbreaks [WebCite Cache]
    9. Fomina A, Prelich P, Toarca S. ParseHub.   URL: https://www.parsehub.com/ [accessed 2018-07-02] [WebCite Cache]
    10. Esmukov K. GitHub. 2006. geopy   URL: https://github.com/geopy/geopy [accessed 2018-06-27] [WebCite Cache]
    11. National Institute of Diabetes and Digestive and Kidney Diseases. 2014. Foodborne Illnesses   URL: https://www.niddk.nih.gov/health-information/digestive-diseases/foodborne-illnesses [accessed 2018-12-17] [WebCite Cache]
    12. International Society of Travel Medicine. GeoSentinel Document Center (Sites Only)   URL: http://www.istm.org/geosentineldocumentcente2 [accessed 2018-06-29] [WebCite Cache]
    13. Nsoesie EO, Kluberg SA, Brownstein JS. Online reports of foodborne illness capture foods implicated in official foodborne outbreak reports. Prev Med 2014 Oct;67:264-269 [FREE Full text] [CrossRef] [Medline]
    14. Ostrov BF. USA Today. 2015 Oct 24. Using Yelp to track food poisoning outbreaks   URL: https:/​/www.​usatoday.com/​story/​news/​2015/​10/​24/​kaiser-using-yelp-to-track-food-poisoning-outbreaks/​74469136/​ [WebCite Cache]
    15. Harris JK, Hinyard L, Beatty K, Hawkins JB, Nsoesie EO, Mansour R, et al. Evaluating the implementation of a Twitter-based foodborne illness reporting tool in the city of St. Louis Department of Health. Int J Environ Res Public Health 2018 Apr 24;15(5):- [FREE Full text] [CrossRef] [Medline]
    16. Porter CK, Olson S, Hall A, Riddle MS. Travelers' diarrhea: an update on the incidence, etiology, and risk in military deployments and similar travel populations. Mil Med 2017 Sep;182(S2):4-10. [CrossRef] [Medline]


    Abbreviations

    LMIC: low- and middle-income country


    Edited by T Sanchez; submitted 03.07.18; peer-reviewed by S Ghosh, ICH Fung; comments to author 26.11.18; revised version received 18.01.19; accepted 18.01.19; published 01.04.19

    ©Nicole E Kogan, Isabelle Bolon, Nicolas Ray, Gabriel Alcoba, Jose L Fernandez-Marquez, Martin M Müller, Sharada P Mohanty, Rafael Ruiz de Castañeda. Originally published in JMIR Public Health and Surveillance (http://publichealth.jmir.org), 01.04.2019.

    This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on http://publichealth.jmir.org, as well as this copyright and license information must be included.