177 Analysis of Flu Near You As a Flu-Tracking Tool for Multiple Spatial Scales in the US

Monday, June 20, 2016: 10:00 AM-10:30 AM
Exhibit Hall Section 1, Dena'ina Convention Center
Kristin Baltrusaitis , HealthMap, Boston Children's Hospital, Boston, MA
John Brownstein , Harvard Medical School, Boston, MA
Adam Wade Crawley , Skoll Global Threats Fund, San Francisco, CA
Giuseppe Conidi , Boston Public Health Commission, Boston, MA
Julia Gunn , Boston Public Health Commission, Boston, MA
Mauricio Santillana , Harvard Medical School, Boston, MA

BACKGROUND:  Flu Near You (FNY), a participatory disease surveillance system, allows volunteers to report if they experience influenza-like symptoms using a brief weekly survey. We compare FNY influenza-like illnesses (ILI) with ILI estimates from governmental agencies over three influenza seasons in the United States at four different geographical levels.

METHODS:  Raw and noise-filtered FNY ILI rates were compared to ILI rates from the Centers for Disease Control and Prevention ILINet surveillance system at the national and HHS-defined regional levels for a time period containing three flu-seasons, 2012-2013, 2013-2014, and 2014-2015. We also present comparisons of FNY ILI for a select number of states and for the Greater Boston area. The Pearson correlation and root mean squared error (RMSE) were calculated at each geographical level.

RESULTS:  Overall, correlation values were observed to decrease as the geographic resolution increased. At the national level, the Pearson correlation between FNY noise-filtered and CDC ILI was 0.957, whereas for the 10 HHS regions, correlation values of about 0.8-0.9 were observed. State-level correlation values ranged from 0.12-0.74, and finally, a correlation of about 0.670 was observed for the greater Boston Area.   

CONCLUSIONS:  Our findings suggest that FNY ILI values track official ILI estimates in geographic areas with a significant enough number of FNY participants. As the FNY user base is increased throughout the US,  FNY may become a complementary way to timely track flu activity, especially in populations who do not access health care systems or in areas with limited surveillance data.