Flu surveillance comparison finds flaws in Google tool
A comparison of traditional surveillance and Google Flu Trends (GFT) over the past decade found that GFT isn't a reliable tool for tracking seasonal and pandemic influenza, having missed the first 2009 H1N1 pandemic wave and overestimated the 2012-13 H3N2 epidemic, according to a study released yesterday.
A research team from New York, the National Institutes of Health, and George Washington University published the findings in PLoS Computational Biology. GFT monitors Internet searches for flu information as a way to estimate flu activity.
The team's comparison of weekly flu-like illness and GFT data spanned 10 flu seasons, from Jun 1, 2003, through Mar 30, 2013. GFT officially launched in November 2008. For the weekly flu surveillance component the group used national and regional data from the US Centers for Disease Control and Prevention (CDC). The local data they used came from New York City's emergency department syndromic surveillance system.
The traditional surveillance and GFT estimates showed similar seasonal and epidemic patterns over the three geographic scales, but the group found notable differences between the retrospective and prospective periods. They found that the original GFT model severely underestimated the first pandemic wave, and the updated GFT model greatly exaggerated the intensity of the 2012-13 season, which suggested substantial flaws in the tool.
The researchers noted that their study isn't the first to flag problems with GFT estimates, but their work is the first to compare the system across a decade of flu activity. They wrote that the shortcomings are serious, given that an initial pandemic wave is a time when accurate data are most needed.
The team concluded that there is a place for search query monitoring in disease surveillance and that it could benefit from integrating near real-term surveillance data and improved computational and modeling methods. Such improvements could make the systems more useful for public health decision-making, they wrote.
Oct 17 PLoS Comput Biol study
CDC flu report shows little activity, one H3N2v infection
The US CDC today resumed its weekly flu surveillance reports, noting little flu activity so far but reporting another variant H3N2 (H3N2v) case, in a patient from Iowa who had contact with pigs.
Weekly flu reports were suspended during the 16-day government shutdown, which ended the night of Oct 16. The CDC originally had planned to issue its first full flu surveillance report of the 2013-14 season today, but the agency said today's report is an abbreviated version. The next report, on Oct 25, will be a full version
So far flu markers are below epidemic thresholds. For example, the CDC said the percentage of doctor's visits for flulike illness was 1.1% last week, putting it below the national baseline of 2.0% and only slightly higher than the 1.0% reported in the CDC's last report before the shutdown. It said all 10 regions are below their baselines.
Puerto Rico is the only area reporting regional spread of flu, while Alabama, South Carolina, and Texas reported local activity. Roughly half of the states, plus the District of Columbia, Guam, and the US Virgin Islands, reported sporadic flu cases.
The additional H3N2v case was reported in a patient who had contact with swine the week before he or she got sick. The patient has recovered, and no further cases were detected during follow-up of close contacts. The CDC said the case will be reported as the first H3N2v case of the 2013-14 season. For the season that just ended, 18 cases were reported.
Oct 18 CDC weekly flu update