BACKGROUND: Flu Near You (FNY), a near real-time participatory surveillance system for tracking symptoms in the U.S. and Canada, provides a novel source for influenza-like illness (ILI) data. This study aimed to evaluate whether FNY provided Maricopa County Department of Public Health (MCDPH) with more timely or additional information during the 2015-2016 influenza season compared to the National Syndromic Surveillance Program BioSense Platform.
METHODS: We filtered national FNY data to include only Maricopa County zip codes. Using CDC’s “Updated Guidelines for Evaluating Public Health Surveillance Systems”, we evaluated three major attributes of the system. Timeliness was assessed by describing the time intervals between symptom onset, symptom reporting, and data analysis. We measured representativeness by comparing the demographics of FNY users to Census population data from Maricopa County. To assess data quality, we analyzed the correlation of ILI trends (defined as fever AND [cough OR sore throat]) from FNY and BioSense by calculating the Pearson correlation statistic and root mean square error (RMSE). To account for potential bias from those who used FNY only once, we removed one-time users from the dataset and reexamined the correlation to BioSense.
RESULTS: Maricopa County had 557 unique FNY users (175 [31%] one-time users) from 10/05/15 – 10/02/16, with 19 to 232 users per week (mean=130). Each Monday, users received surveys to report symptoms experienced during the previous 7 days. The survey for that timeframe was open to symptom reporting for one week before the dataset was finalized and posted to a dashboard for use by public health agencies. Therefore, symptom data were received by MCDPH 8 – 14 days after onset. Compared to the general population, Maricopa County FNY users included significantly more females (72.3% vs. 50.57%, p<0.01) and adults over 45 (p<0.01). Of the 162 users who reported ILI symptoms, 74 (45.7%) did not seek healthcare. FNY rates of ILI over time were significantly correlated with BioSense rates (Pearson=0.65, p<0.01, RMSE=1.60%). The correlation remained significant but did not improve when one-time users were removed from the dataset (Pearson=0.55, p-value<0.01, RMSE=1.51%). During the 2015-16 season, FNY detected a peak in ILI three weeks prior to BioSense.
CONCLUSIONS: FNY captured data from persons who did not seek healthcare and identified a peak in ILI three weeks earlier than BioSense. These timely, unique data may complement existing influenza surveillance. In an effort to increase representativeness and regular use of FNY, MCDPH continues to encourage enrollment in Maricopa County.