Search

Google Flu Trends was developed to estimate US influenza-like illness (ILI) rates from internet searches; however ILI does not necessarily correlate with actual influenza virus infections.Influenza activity data from 2003-04 through 2007-08 were obtained from three US surveillance systems: Google Flu Trends, CDC Outpatient ILI Surveillance Network (CDC ILI Surveillance), and US Influenza Virologic Surveillance System (CDC Virus Surveillance). Pearson's correlation coefficients with 95% confidence intervals (95% CI) were calculated to compare surveillance data. An analysis was performed to investigate outlier observations and determine the extent to which they affected the correlations between surveillance data. Pearson's correlation coefficient describing Google Flu Trends and CDC Virus Surveillance over the study period was 0.72 (95% CI: 0.64, 0.79). The correlation between CDC ILI Surveillance and CDC Virus Surveillance over the same period was 0.85 (95% CI: 0.81, 0.89). Most of the outlier observations in both comparisons were from the 2003-04 influenza season. Exclusion of the outlier observations did not substantially improve the correlation between Google Flu Trends and CDC Virus Surveillance (0.82; 95% CI: 0.76, 0.87) or CDC ILI Surveillance and CDC Virus Surveillance (0.86; 95%CI: 0.82, 0.90).This analysis demonstrates that while Google Flu Trends is highly correlated with rates of ILI, it has a lower correlation with surveillance for laboratory-confirmed influenza. Most of the outlier observations occurred during the 2003-04 influenza season that was characterized by early and intense influenza activity, which potentially altered health care seeking behavior, physician testing practices, and internet search behavior.

BackgroundGoogle Flu Trends (GFT) uses anonymized, aggregated internet search activity to provide near-real time estimates of influenza activity. GFT estimates have shown a strong correlation with official influenza surveillance data. The 2009 influenza virus A (H1N1) pandemic [pH1N1] provided the first opportunity to evaluate GFT during a non-seasonal influenza outbreak. In September 2009, an updated United States GFT model was developed using data from the beginning of pH1N1.Methodology/principal findingsWe evaluated the accuracy of each U.S. GFT model by comparing weekly estimates of ILI (influenza-like illness) activity with the U.S. Outpatient Influenza-like Illness Surveillance Network (ILINet). For each GFT model we calculated the correlation and RMSE (root mean square error) between model estimates and ILINet for four time periods: pre-H1N1, Summer H1N1, Winter H1N1, and H1N1 overall (Mar 2009-Dec 2009). We also compared the number of queries, query volume, and types of queries (e.g., influenza symptoms, influenza complications) in each model. Both models' estimates were highly correlated with ILINet pre-H1N1 and over the entire surveillance period, although the original model underestimated the magnitude of ILI activity during pH1N1. The updated model was more correlated with ILINet than the original model during Summer H1N1 (r = 0.95 and 0.29, respectively). The updated model included more search query terms than the original model, with more queries directly related to influenza infection, whereas the original model contained more queries related to influenza complications.ConclusionsInternet search behavior changed during pH1N1, particularly in the categories "influenza complications" and "term for influenza." The complications associated with pH1N1, the fact that pH1N1 began in the summer rather than winter, and changes in health-seeking behavior each may have played a part. Both GFT models performed well prior to and during pH1N1, although the updated model performed better during pH1N1, especially during the summer months.