OBJECTIVES: We analyzed the possibility of using sales data on over-the-counter (OTC) medicine in syndromic surveillance for early detection and/or prediction of influenza pandemics. METHODS: Data who collected on daily November 2003-April 2004 sales of common cold medication commercially obtained from 600 pharmacies in Japan enrolled in reporting throughout the 2003/04 influenza season. OTC sales data were compared to daily influenza incidence data to determine correlations and predictability. Data included Sentinel Surveillance of influenza, National Surveillance of Dally Influenza Outpatients (Daily Case Reporting: DCR), and Mailing-List-Based Influenza Epidemic Database (MLflu). RESULTS: The seasonal influenza epidemic peak for 2003-2004 was observed on January 30, February 1, and February 2 for sentinel surveillance, DCR, and MLflu. Sales peaks of OTC medicines occured twice in 2003-2004 season, once at the end of the year and once on January 27. Peak OTC sales for common cold medication per pharmacy were preceded by 18, 21, and 13 days for sentinel surveillance, DCR, and MLflu. OTC sales successfully explained 74-85% in the variation of influenza incidence which is interpretable as sufficient power of explanation. CONCLUSIONS: OTC sales proved to be a good indicator of real-time surveillance for influenza epidemics. Further analysis on multiyear data is needed to demonstrate the robustness of results. To confirm the advantage in the system of real-time syndromic surveillance, it is also necessary to check explanatory power of OTC sales on the variation of influenza incidence with prospective datasets.
OBJECTIVES: We analyzed the possibility of using sales data on over-the-counter (OTC) medicine in syndromic surveillance for early detection and/or prediction of influenza pandemics. METHODS: Data who collected on daily November 2003-April 2004 sales of common cold medication commercially obtained from 600 pharmacies in Japan enrolled in reporting throughout the 2003/04 influenza season. OTC sales data were compared to daily influenza incidence data to determine correlations and predictability. Data included Sentinel Surveillance of influenza, National Surveillance of Dally Influenza Outpatients (Daily Case Reporting: DCR), and Mailing-List-Based Influenza Epidemic Database (MLflu). RESULTS: The seasonal influenza epidemic peak for 2003-2004 was observed on January 30, February 1, and February 2 for sentinel surveillance, DCR, and MLflu. Sales peaks of OTC medicines occured twice in 2003-2004 season, once at the end of the year and once on January 27. Peak OTC sales for common cold medication per pharmacy were preceded by 18, 21, and 13 days for sentinel surveillance, DCR, and MLflu. OTC sales successfully explained 74-85% in the variation of influenza incidence which is interpretable as sufficient power of explanation. CONCLUSIONS: OTC sales proved to be a good indicator of real-time surveillance for influenza epidemics. Further analysis on multiyear data is needed to demonstrate the robustness of results. To confirm the advantage in the system of real-time syndromic surveillance, it is also necessary to check explanatory power of OTC sales on the variation of influenza incidence with prospective datasets.