6-005-1430-Eyenbach

850 views

Published on

Published in: Business, Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
850
On SlideShare
0
From Embeds
0
Number of Embeds
3
Actions
Shares
0
Downloads
13
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

6-005-1430-Eyenbach

  1. 1. Gunther Eysenbach MD MPH<br />Gunther Eysenbach MD MPH<br />Pandemics in the Age of Twitter: A Case Study of Infodemiology and Infoveillanceof Social Media as New Methods for Knowledge Translation Research and Syndromic Surveillance<br />Director, Consumer Health & Public Health Informatics Lab<br />Professor Department ofHealthPolicy, Management and Evaluation, University of Toronto;<br />Senior Scientist, Centrefor Global eHealth Innovation,Division of Medical Decision Making andHealth Care Research; <br />Toronto General Research Institute ofthe UHN, Toronto General Hospital, Canada<br />geysenba@gmail.com<br />
  2. 2. Economists have something public health practitioners don’t have: Real-time indices to track behavior and emotions<br />S&P 500 Stock Index<br />Fear Index<br />
  3. 3. The premise<br />“The Internet has made measurable what was previously immeasurable: The distribution of health information in a population, tracking (in real time) health information trends over time, and identifying gaps between information supply and demand. “<br />Eysenbach G. Infodemiology: Tracking Flu-Related Searches on the Web for Syndromic Surveillance. Proc AMIA Fall Symp2006: 244-249<br />
  4. 4. Eysenbach G. Infodemiology: Tracking Flu-Related Searches on the Web for Syndromic Surveillance. Proc AMIA Fall Symp2006: 244-249<br />
  5. 5. Infodemiology: Tracking Flu-Related Searches on the Web for Syndromic Surveillance(Eysenbach G, 2006) <br />Google searches vsFlu Cases 2004/05<br />Sentinel Physician ILI reports<br />vs Flu Cases 2004/05<br />Eysenbach G. Infodemiology: Tracking Flu-Related Searches on the Web for Syndromic Surveillance. Proc AMIA Fall Symp2006: 244-249<br />
  6. 6. Idea was 3 yrs later confirmed (Ginsberg, 2009) and commercialized by Google<br />
  7. 7. Infodemiology / infoveillance – related terms<br /><ul><li>“pulse” (Douglas Hubbard, Wiley 2011)
  8. 8. “epidemic intelligence” (Brownstein)
  9. 9. Infometrics, webometrics</li></ul>http://www.youtube.com/watch?v=Xe27-mrQibQ<br />
  10. 10. Infodemiology * / InfoveillanceApplication Areas<br />Syndromic surveillance<br />detecting and quantifying disparities in health information availability<br />identifying and monitoring of public health relevant publications on the Internet (e.g. anti-vaccination sites)<br />measure information diffusion and knowledge translation, and tracking the effectiveness of health marketing campaigns. <br />Monitoring health-related behavior<br />* science of distribution and determinants of information in an electronic medium, specifically the Internet, or in a population, with the ultimate aim to inform public health and public policy.<br />
  11. 11. Infodemiology Research Program @ Centre for Global eHealth Innovation, Eysenbach Lab<br />Developing innovative tools & methods to measure/track health-related attitudes, knowledge, emotions, public attention, behavior in real time for public healthusing textual data from the Internet & Social Media<br />Investigate how the public is using social media during a pandemic, and how social media can be used to engage the public<br />
  12. 12. Studying information patterns in the era of user-generated information (Web 2.0) <br />enables us to measure user attitudes, behavior, awareness, knowledge, attention, information needs etc.<br />Image Source: http://web2.wsj2.com/<br />Gunther Eysenbach MD, MPH, www.medicine20congress.com<br />
  13. 13. Infovigilan open source infoveillance prototypeCentre for Global eHealth Innovation, Toronto<br />Filter<br />Keywords / Concepts of Interest<br />User-submitted <br />information<br />Infovigil<br />Aggregator/<br />Datamining/<br />Vizualisation<br />Websites<br />Public,<br />Clinicians,<br />Epidemiologists<br />
  14. 14.
  15. 15. “H1N1 marks the first instance in which a global pandemic has occurred in the age of Web 2.0 and presents a unique opportunity to investigate the potential role of these technologies in public health emergencies.”<br />Chew C, Eysenbach G (2010) Pandemics in the Age of Twitter: Content Analysis of Tweets during the 2009 H1N1 Outbreak. PLoS ONE 5(11): e14118<br />
  16. 16.
  17. 17. Infoveillance<br />Predicting/tracking public-health relevant events (syndromic surveillance for outbreaks, “all-hazards” surveillance), <br />Tracking changes in behavior, attitudes, knowledge (e.g. as a result of public health messages or interventions)<br />Situational awareness regarding current concerns, issues, questions, emotions, of the public<br />Eysenbach G. Infodemiology and Infoveillance: New Methods…. J Med Internet Res 2009: e11http://www.jmir.org/2009/1/e11<br />
  18. 18. Absolute and relative number of H1N1/Swine-Flu Tweets<br />% both<br />% H1N1<br />% swine flu<br />The relative proportion of tweets using “H1N1” increased from 8.8% to 40.5% in an almost linear fashion (R2= .788; p < .001), indicating a gradual adoption of the WHO-recommended H1N1 terminologyas opposed to “Swine Flu”<br />
  19. 19. 53 %<br />23 %<br />14 %<br />8 %<br />1 %<br />2 %<br />Chew C, Eysenbach G (2010) Pandemics in the Age of Twitter: Content Analysis of Tweets during the 2009 H1N1 Outbreak. PLoS ONE 5(11): e14118<br />
  20. 20. Media Resonance Analysis<br />Absolute number of tweets<br />(Blue: swine flu, red: h1n1)<br />spikes mainly due to major news events e.g <br /><ul><li> [A] WHO declares pandemic,
  21. 21. [P] Obama declares national emergency
  22. 22. [B] Harry Potter actor Rupert Grint has Swine Flu</li></ul>Chew C, Eysenbach G (2010) Pandemics in the Age of Twitter: Content Analysis of Tweets during the 2009 H1N1 Outbreak. PLoS ONE 5(11): e14118<br />
  23. 23. Relative usage of terminology:<br />H1N1:SwineFlu Ratio<br />#oink<br />campaign<br /><ul><li>The relative proportion of tweets using “H1N1” increased in an almost linear fashion (R2= .788; p < .001), indicating a gradual adoption of the WHO-recommended H1N1 terminologyas opposed to “Swine Flu”
  24. 24. also social media campaigns show some effect ([G] #oink campaign of farmers)</li></ul>Chew C, Eysenbach G (2010) Pandemics in the Age of Twitter: Content Analysis of Tweets during the 2009 H1N1 Outbreak. PLoS ONE 5(11): e14118<br />
  25. 25. “Happiness / Humor / Mood Index”:<br />Smileys : Frowneys Ratio<br />
  26. 26. Question IndexNumber of tweets with ? : Total Tweets<br />Chew C, Eysenbach G (2010) Pandemics in the Age of Twitter: Content Analysis of Tweets during the 2009 H1N1 Outbreak. PLoS ONE 5(11): e14118<br />
  27. 27. Prayer IndexNumber of tweets with “pray” : Total Tweets<br />H1N1 Hospitalizations / Deaths<br />
  28. 28. Personal Experiences<br />Chew C, Eysenbach G (2010) Pandemics in the Age of Twitter: Content Analysis of Tweets during the 2009 H1N1 Outbreak. PLoS ONE 5(11): e14118<br />
  29. 29. Number of tweets with “personal experiences” correlates to H1N1 incidence<br />Chew C, Eysenbach G (2010) Pandemics in the Age of Twitter: Content Analysis of Tweets during the 2009 H1N1 Outbreak. PLoS ONE 5(11): e14118<br />
  30. 30. Some newer findings on vaccination sentiment<br />
  31. 31. Vaccine / Vaccination Mentionings<br />Chew C, Eysenbach G (2010) Pandemics in the Age of Twitter: Content Analysis of Tweets during the 2009 H1N1 Outbreak. PLoS ONE 5(11): e14118<br />
  32. 32. Sentiment Categories<br />
  33. 33. Manual Analysis: Anti-Vaccination Concerns<br />
  34. 34. Manual Sentiment Analysis<br />Green = Pro-vaccination<br />Red = Anti-vaccination<br />
  35. 35. % of H1N1 Pro- and Anti-Vaccination Tweets over 2009(keyword based)<br />Red = anti<br />Green = pro<br />
  36. 36. Pro-Vaccination Tweets (keyword-based)<br />Subconcepts: Green = support/promotion; Red = positive intent/behaviour; Blue = positive emotion; <br />Yellow = rollout satisfaction<br />1 = Jun 18: “Vaccinate kids to control H1N1 flu: researchers”<br />2 = Jul 13: “WHO says health workers priority for H1N1 vaccine”<br />3 = Aug 22: “Swine flu vaccine seems safe in early trials”<br />4 = Sep 16: “FDA approves H1N1 vaccines”<br />
  37. 37. Anti-Vaccination Tweets (keyword-based)<br />Subconcepts: Green = vaccine health risks; Blue = conspiracy; Red = rollout dissatisfaction; Yellow = negative intent/behavior; Purple = negative emotion; Navy Blue = vaccine downplay/dissuasion<br />1 = Jul 8: “Canadian Doctor: H1N1 vaccination a eugenics weapon for mass <br /> extermination”<br />2 = Aug 16: “Swine flu jab linked to killer nerve disease: leaked letter reveals <br /> concerns of neurologists”<br />3 = Sep 27: “Doctor admits vaccine is more deadly than the H1N1 virus and <br /> wouldn’t give it to his children” <br />
  38. 38. Another analysis: Number of retweets as measure of influence<br />
  39. 39. Who was the most RT’ed?<br />Retweets (RT) may be a measure of influence, reach, or interest<br />
  40. 40. Other Heavily RT’ed Accounts<br />Surprising Finding: With few exceptions, these are mainly the traditional “intermediaries”<br />
  41. 41. Dr. Mercola, author of “The Great Bird Flu Hoax”<br />
  42. 42. Conclusions<br />Infoveillance: New methodology, offers wealth of quantitative + qualitative data, complementary to traditional survey methods, more timely and inexpensive<br />Twitter is a rich source of opinions and experiences, which can be used for near real-time content and sentiment analysis, knowledge translation research, and potentially as a syndromic surveillance tool, allowing health authorities to become aware of and respond to real or perceived concerns/issues raised by the public<br />Social media appeared underused by Canadian public health authorities during the H1N1 pandemic <br />
  43. 43. Future work<br />Natural language processing / machine learning approaches<br />Grant proposal pending – re-engineer Infovigil to make it usable for other groups<br />
  44. 44. “In the era of the 24-hour news cycle, the traditional once-a-day press conference featuring talking heads with a bunch of fancy titles has to be revamped and supplemented with Twitter posts, YouTube videos and the like. The public needs to be engaged in conversations and debate about issues of public health, they don’t need to be lectured to.”-Andre Picard <br />Picard A (2010) Lessons of H1N1: Preach less, reveal more. Globe and Mail. <br />Available: http://www.webcitation.org/5qYZly99e.<br />
  45. 45. Principal Investigator:Gunther Eysenbach MD MPHDirector, Consumer Health & Public Health Informatics LabCentre for Global eHealth Innovationgeysenba@gmail.com<br />Acknowledgements<br />Thanks to CIHR & Reviewers<br />Cynthia Chew (MSc Student): Coding & Qualitative Analysis of Tweets<br />Latifa Mnyusiwalla (MHI Student): Vaccination Sentiment Analysis<br />Marina Sokolova PhD, CHEO Ottawa: Natural Language Processing<br />Phil Cairns: Developer<br />
  46. 46. Appendix<br />For Q & A<br />
  47. 47. Challenges<br />Data are noisy - advanced NLP (natural language processing) algorithms required<br />not necessarily representative for the general population (young + educated)<br />Spamming / twitter bots<br />Geographic origin difficult to determine<br />Privacy issues?<br />
  48. 48. http://www.jmir.org/2009/1/e11<br />
  49. 49. The science of distribution and determinants of disease in populations Epidemiology,<br />Polls, Focus groups<br />Public Health Interventions<br />Policy Decisions<br />Public Health Professionals<br />Policy Makers<br />Traditional Knowledge Translation Circle<br />Population Behaviour, <br />Attitudes, <br />Health Status<br />PR / Media Campaigns<br />
  50. 50. Information &<br />Communication<br />patterns<br />The science of distribution and determinants of disease in populations Epidemiology,<br />Polls, Focus groups<br />Public Health Interventions<br />Policy Decisions<br />Public Health Professionals<br />Policy Makers<br />Traditional Knowledge Translation Circle<br />Population Behaviour, <br />Attitudes, <br />Health Status<br />PR / Media Campaigns<br />Web 1.0: Webpages, News<br />Web 2.0: User generated content, social media <br />Searches, Navigation, Clicks<br />
  51. 51. Information &<br />Communication<br />patterns<br />The science of distribution and determinants of disease in populations Epidemiology,<br />Polls, Focus groups<br />Public Health Interventions<br />Policy Decisions<br />Public Health Professionals<br />Policy Makers<br />Traditional Knowledge Translation Circle<br />Population Behaviour, <br />Attitudes, <br />Health Status<br />Infoveillance<br />PR / Media Campaigns<br />“Infodemiology”the epidemiology of informationAnalyzing information & communication patterns (on the web)<br />Web 1.0: Webpages, News<br />Web 2.0: User generated content, social media <br />Searches, Navigation, Clicks<br />Metrics<br />

×