Gunther Eysenbach1. 1. Centre for Global eHealth Innovation, University Health Network, Toronto M5G2C4 and Department of Health Policy, Management and Evaluation, University of Toronto, Canada. geysenba@uhnres.utoronto.ca
Abstract
BACKGROUND: Syndromic surveillance uses health-related data that precede diagnosis and signal a sufficient probability of a case or an outbreak to warrant further public health response. OBJECTIVE: While most syndromic surveillance systems rely on data from clinical encounters with health professionals, I started to explore in 2004 whether analysis of trends in Internet searches can be useful to predict outbreaks such as influenza epidemics and prospectively gathered data on Internet search trends for this purpose. RESULTS: There is an excellent correlation between the number of clicks on a keyword-triggered link in Google with epidemiological data from the flu season 2004/2005 in Canada (Pearson correlation coefficient of current week clicks with the following week influenza cases r=.91). The "Google ad sentinel method" proved to be more timely, more accurate and - with a total cost of Can$365.64 for the entire flu-season - considerably cheaper than the traditional method of reports on influenza-like illnesses observed in clinics by sentinel physicians. CONCLUSION: Systematically collecting and analyzing health information demand data from the Internet has considerable potential to be used for syndromic surveillance. Tracking web searches on the Internet has the potential to predict population-based events relevant for public health purposes, such as real outbreaks, but may also be confounded by "epidemics of fear". Data from such "infodemiology studies" should also include longitudinal data on health information supply.
BACKGROUND: Syndromic surveillance uses health-related data that precede diagnosis and signal a sufficient probability of a case or an outbreak to warrant further public health response. OBJECTIVE: While most syndromic surveillance systems rely on data from clinical encounters with health professionals, I started to explore in 2004 whether analysis of trends in Internet searches can be useful to predict outbreaks such as influenza epidemics and prospectively gathered data on Internet search trends for this purpose. RESULTS: There is an excellent correlation between the number of clicks on a keyword-triggered link in Google with epidemiological data from the flu season 2004/2005 in Canada (Pearson correlation coefficient of current week clicks with the following week influenza cases r=.91). The "Google ad sentinel method" proved to be more timely, more accurate and - with a total cost of Can$365.64 for the entire flu-season - considerably cheaper than the traditional method of reports on influenza-like illnesses observed in clinics by sentinel physicians. CONCLUSION: Systematically collecting and analyzing health information demand data from the Internet has considerable potential to be used for syndromic surveillance. Tracking web searches on the Internet has the potential to predict population-based events relevant for public health purposes, such as real outbreaks, but may also be confounded by "epidemics of fear". Data from such "infodemiology studies" should also include longitudinal data on health information supply.
Authors: Dena M Bravata; Kathryn M McDonald; Wendy M Smith; Chara Rydzak; Herbert Szeto; David L Buckeridge; Corinna Haberland; Douglas K Owens Journal: Ann Intern Med Date: 2004-06-01 Impact factor: 25.391
Authors: Kenneth D Mandl; J Marc Overhage; Michael M Wagner; William B Lober; Paola Sebastiani; Farzad Mostashari; Julie A Pavlin; Per H Gesteland; Tracee Treadwell; Eileen Koski; Lori Hutwagner; David L Buckeridge; Raymond D Aller; Shaun Grannis Journal: J Am Med Inform Assoc Date: 2003-11-21 Impact factor: 4.497
Authors: Sharad Goel; Jake M Hofman; Sébastien Lahaie; David M Pennock; Duncan J Watts Journal: Proc Natl Acad Sci U S A Date: 2010-09-27 Impact factor: 11.205
Authors: Jason S Shapiro; Farzad Mostashari; George Hripcsak; Nicholas Soulakis; Gilad Kuperman Journal: Am J Public Health Date: 2011-02-17 Impact factor: 9.308
Authors: Natasha L Tilston; Ken T D Eames; Daniela Paolotti; Toby Ealden; W John Edmunds Journal: BMC Public Health Date: 2010-10-27 Impact factor: 3.295