Observing what Internet users type into Web search engines could be a unique way to conduct infectious disease surveillance, according to researchers at the Swedish Institute for Infectious Disease Control in Solna, Sweden.
In a letter to the editor in the August issue of Emerging Infectious Diseases, "Eye-Opening Approach to Norovirus Surveillance," researchers Anette Hulth, Yvonne Andersson, Kjell-Olof Hedlund, and Mikael Andersson assert that norovirus infections "are assumed to greatly affect society, but little is known about the prevalence of the disease in the community. Samples sent to laboratories usually originate from hospitalized persons and thus give a good view of the situation in healthcare settings. We suspect, however, that these numbers do not depict the true prevalence of norovirus infections in society. We therefore present a new approach to estimate the number of cases and spread of norovirus infections in the community."
Hulth and colleagues plotted the number of Web queries for "vomit" submitted to the search engine on a medical Web site in Sweden (www.vardguiden.se). This number was normalized to account for the increasing use of the Web site over time and aggregated by week, starting with week 40 in 2005. They also plotted the number of norovirus findings per week from 16 regional laboratories, as recorded by the Swedish Institute for Infectious Disease Control.
For the time series on Web search queries and laboratory findings, the researchers report that they fitted harmonic functions on the half-year with no or little activity, defining baselines for each series, and explain, "By performing this procedure, we can identify the onset of each activity that is assumed to occur when the level rises above the 99 percent prediction interval of the baseline. By analyzing the figure and investigating the statistical outcomes, we glimpse the prevalence of norovirus infections in society, as estimated by the search pattern".
The researchers say they uncovered several insights: "First, the onset of vomiting in the community precedes the onset of confirmed norovirus infections in healthcare settings. In three of the four full seasons investigated, this precedence was one to four weeks. Second, the curve for the Web queries shows much sharper increases and decreases than does the curve on the number of reported norovirus findings. Third, neither search behavior nor reporting of positive tests is driven by media for the winter vomiting disease (confirmed by a linear regression)."
The researchers continue, "In the 200506 season, the laboratory reporting raised above the defined prediction interval in week 13, much later than the Web queries. This season had no new variants of norovirus genotype GII.4. This season still showed community infections, even though few reports came from institutions. For the current season (2009-2010), the interval between onset of Web queries and onset of norovirus infections in hospitals (week 46 and week 1, respectively) was eight weeks. In comparison with previous seasons, this delay could mean a low total number of reported cases. However, in late December, a new variant of GII.4 affected healthcare settings in southern Sweden with increasing norovirus infections, while the rest of the country still showed relatively low virus activity. Other pathogens such as rotavirus, Salmonella spp., Staphylococcus aureus, and Bacillus cereus can cause vomiting. Usually in Sweden, rotavirus infections peak in late winter, and bacterial diseases have a minor incidence compared with norovirus. In our opinion, these other pathogens would not interfere with the interpretation of the results. In our routine surveillance of Web queries, we also include other query terms, such as diarrhea and stomach flu. However, searches for vomiting show the most distinct pattern, and vomiting is the most pronounced symptom of a norovirus infection. The use of harmonic functions for describing baseline Web searches and laboratory reporting is a simple model, especially because the parameters are estimated by using the half-year with the least activity. Nonetheless, it is a direct approach, and we believe that the method still captures the time of onsets well."
Hulth and colleagues emphasize that "Web queries indicate the presence of norovirus infections in communities. Predictions of the onset of the norovirus laboratory reporting should also be possible, but further studies are needed to confirm that theory. Web queries have previously been correlated with influenza3-7 and have been explored retrospectively for listeriosis,8 Salmonella spp.,9 West Nile virus, and respiratory syncytial virus.10 With the Web queries, we get an additional surveillance system for the time of the year when few norovirus tests are conducted. In addition, knowing more about the impact of norovirus in the community means that we could provide more adequate information and advocate wiser measures for prevention and control."
Reference: Hulth A, Andersson Y, Hedlund K-O, Andersson M. Eye-opening approach to norovirus surveillance [letter]. Emerg Infect Dis. August 2010. http://www.cdc.gov/EID/content/16/8/1319.htm
1. Serfling RE. Methods for current statistical analysis of excess pneumoniainfluenza deaths. Public Health Rep. 1963;78:494506. PubMed
2. Simonsen L, Clarke MJ, Williamson GD, Stroup DF, Arden NH, Schonberger LB. The impact of influenza epidemics on mortality: introducing a severity index. Am J Public Health. 1997;87:194450. PubMed DOI: 10.2105/AJPH.87.12.1944
3. Eysenbach G. Infodemiology: tracking flu-related searches on the web for syndromic surveillance. AMIA Annu Symp Proc 2006:2448.
4. Polgreen PM, Chen Y, Pennock DM, Forrest ND. Using Internet searches for influenza surveillance. Clin Infect Dis. 2008;47:14438. PubMed DOI: 10.1086/593098
5. Ginsberg J, Mohebbi MH, Patel RS, Brammer L, Smolinski MS, Brilliant L. Detecting influenza epidemics using search engine query data. Nature. 2009;457:10124. PubMed DOI: 10.1038/nature07634
6. Hulth A, Rydevik G, Linde A. Web queries as a source for syndromic surveillance. PLoS One. 2009;4:e4378. PubMed DOI: 10.1371/journal.pone.0004378
7. Wilson N, Mason K, Tobias M, Peacey M, Huang QS, Baker M. Interpreting "Google Flu Trends" data for pandemic H1N1 influenza: the New Zealand experience. Euro Surveill. 2009;14(44):pii=19386.
8. Wilson K, Brownstein JS. Early detection of disease outbreaks using the Internet. CMAJ. 2009;180:82931. PubMed DOI: 10.1503/cmaj.090215
9. Brownstein JS, Freifeld CC, Madoff LC. Digital disease detectionharnessing the Web for public health surveillance. N Engl J Med. 2009;360:21535, 2157. PubMed DOI: 10.1056/NEJMp0900702
10. Carneiro HA, Mylonakis E. Google trends: a web-based tool for real-time surveillance of disease outbreaks. Clin Infect Dis. 2009;49:155764. PubMed DOI: 10.1086/630200