#hayfever; A longitudinal study into hay fever related tweets in the UK

This paper describes a longitudinal study that has collected and analysed over 512,000 UK geolocated tweets over 2 years from June 2012 that contained instances of the words “hayfever” and “hay fever”. The results indicate that the temporal distribution of the tweets collected in 2014 correlates strongly (r=0.97, p<0.01) with incidents of hay fever reported by the Royal College of General Practitioners (RCGP) in the same year. An analysis of the content of the tweets indicates that users are self-reporting common, often severe symptoms as well as the uses of medication. We conclude that hay fever related tweets provide a real-time, free and easily accessible source of data at a finer level of granularity than currently available data sets. The implications for researchers, health professionals and sufferers are also discussed.

Transcript

1. #hayfever; A
   longitudinal study
   into hay fever related
   tweets in the UK
   Dr Ed de Quincey, Keele University
   Dr Theo Kyriacou, Keele University
   Dr Thomas Pantin, Macclesfield Hospital
   Photo by Maja Dumat
   

   View Slide

2. Dr Ed de Quincey @eddequincey
   Postgraduate Course Director, Lecturer in Computer Science
   School of Computing and Mathematics, KeeleUniversity
   Lead of the Software Engineering Research Cluster
   instagram.com/eddequincey
   

   View Slide

3. The current UK hay
   fever prevalence is
   between 20-25%
   of the
   population,
   projected to rise to
   39% by 2030
   (Emberlin, 2010).
   Photo from “Hayfever hotspots: As pollen counts rise, our
   unique British map tells you where to avoid”
   http://www.dailymail.co.uk/health/article-1015294/Hayfever-hotspots-As-
   pollen-counts-rise-unique-British-map-reveals-avoid.html
   

   View Slide

4. Surges in incidence of hay fever (allergic
   rhinitis) in spring and summer are
   commonly known as the hay fever
   season, with the main pollens in the
   UK being birch pollen, March to
   mid May, and grass pollen, late
   May to August (Emberlin, 2010).
   

   View Slide

5. The Meteorological
   Office (Met Office)
   provide weekly pollen
   forecasts and the
   Royal College of
   General
   Practitioners (RCGP)
   produce weekly
   service reports.
   http://www.metoffice.gov.uk/public/weather/forecast/
   

   View Slide

6. Photo by “A Guy Taking Pictures”
   For researchers and sufferers of hay fever,
   there is currently no method for identifying
   real-time, geolocated hay fever incidence.
   

   View Slide

7. A promising approach in the related field
   of Epidemiological Intelligence to detect
   seasonal illnesses is the use of Social Media
   (de Quincey & Kostkova, 2010). By collecting
   incidences of users self reporting illnesses
   on twitter, it has been shown that
   outbreaks can be predicted 1-2 weeks
   before RCGP data indicates (Szomszor et al, 2012).
   

   View Slide

8. View Slide

9. “1486 following, 283 followers and 134
   updates and 36 direct messages to
   hayfever sufferers on Twitter”
   http://www.figarodigital.co.uk/case-study/Kleenex.aspx
   

   View Slide

10. BENADRYL® Social Pollen Count
    

    View Slide

11. 2nd GRF One Health Summit 2013
    http://www.socialslurp.co.uk/benadryl-pollen-hotspot-goes-tits-up/
    

    View Slide

12. “hayfever” and “hay fever”
    

    View Slide

13. 512,198 tweets
    294,010 people
    Since 20th June 2012 and 31st July 2014
    69.4% “Hayfever”
    30.6% “Hay fever”
    

    View Slide

14. View Slide

15. Distribution similar to Pollen Calendars produced by the National Pollen
    and Aerobiology Research Unit, with peaks in June/July, reductions
    through August/September, no pollen from October to January and
    then a rise in March
    

    View Slide

16. The highest number of tweets posted was 5,826 on the 26th of June
    2012. 52% (3,002) of these were retweets of a tweet from a user,
    @carolineflack1, who currently has 1.7 million followers. A related
    phenomenon was seen when a tweet by @GemmaAnneStyles, who
    currently has 3.16 million followers, was retweeted 1,353 times.
    Photo by “Pop Sugar”
    “The Harry Styles Effect”
    

    View Slide

17. Frequent words included “my” (120,119);
    “like” (34,017); “today” (33,826); “eyes”
    (32,494); “hate” (30,964); “I’m” (28,436);
    “bad” (25,770); “tablets” (24,502);
    “f@ck” (22,052); “nose” (19,859); “cold”
    (15,381); “summer” (15,248) and
    “sneezing” (14,429).
    Content of tweets
    

    View Slide

18. #sneeze;
    #cantstopsneezing;
    #sneezing; #achoo;
    #itchyeyes; #soreeyes;
    #puffyeyes; #sneezy; #sniff
    and #sniffles
    #hashtags relating to symptoms
    

    View Slide

19. “I have hayfever” (4,804)
    “I have hay fever” (2,672)
    “my hayfever” (32,304)
    “my hay fever” (11,386)
    Self-reporting phrases
    

    View Slide

20. Symptom Word Number of tweets
    Eyes (itchy, red, watery etc.) 32,494 (6.3%)
    Sneezing 26,391 (5.1%)
    Nose (runny, blocked etc.) 19,859 (3.8%)
    Itch (y,ing) 9,033 (1.7%)
    Red 5,844 (1.1%)
    Block (ed) 5,655 (1.1%)
    Throat 5,025 (0.9%)
    Water (y,ing,ed) 3,822 (0.7%)
    Runn (y,ing) 3,449 (0.7%)
    Cough 1,421 (0.3%)
    Mouth 800 (0.2%)
    Ears 574 (0.1%)
    Number of tweets containing symptom related terms
    

    View Slide

21. Photo by e-Magine Art
    27,553 tweets related to
    medication were found
    (containing terms such as “medicine”; “tablets”;
    “meds”; “medication”; “pill’; “spray” and “drugs”).
    670 tweets related to
    drug efficacy
    (“tablets don’t work” and “the pills don’t work”)
    

    View Slide

22. Utilising a profanity
    dictionary, the number of
    tweets containing
    expletives was found to be
    around 11% (55,515).
    “Hayfever is totally kicking my **** today. The *******.”
    

    View Slide

23. Top 5 locations for number of tweets
    Location No of tweets No of tweets/Population
    London 86,460 1.03%
    Manchester 21,489 4.18%
    Birmingham 10,458 0.96%
    Liverpool 10,206 2.17%
    Bristol 8,244 1.88%
    

    View Slide

24. Map Source Professor Jean Emberlin, PollenUK
    

    View Slide

25. “Thank you for sight of your map. It matches what I would
    expect to see, broadly fewer instances of pollen being
    registered in Scotland, more in the south and southeast of
    England.” - Dr Peter Burt, Biometeorologist
    

    View Slide

26. 0"
    0.2"
    0.4"
    0.6"
    0.8"
    1"
    0" 5" 10" 15" 20" 25" 30" 35"
    Normalised+Occurrence+per+Week+
    Normalised"RCGP"(all)"
    Normalised"Tweets"
    Comparison between the number of tweets with the GP
    reported data for the first 31 weeks of 2014
    (r=0.97, p<0.01)
    

    View Slide

27. Daily peaks of hay fever incidence
    can be identified as they happen in
    specific places in the UK.
    

    View Slide

28. No content filtering needed (?)
    

    View Slide

29. For comparison the
    Met Office
    offered to provide
    Daily stats from
    March 2011 up until
    May 2013 for 18
    UK locations at a
    cost of £12,100 +
    Vat
    

    View Slide