Respiratory Virus Surveillance Data

We accessed influenza virus surveillance data from 3 sources in the study area: University of Washington Clinical Virology Laboratory, Public Health–Seattle & King County, and Seattle Children’s Hospital. Each laboratory was in King County and participated in the United States Influenza Virologic Surveillance System during the study period. Influenza testing data were available for September 30, 2001, through December 29, 2012, except for the third quarter of 2002 (25). Clinical specimens collected as part of routine care were tested in laboratories for evidence of influenza virus, and results were reported to local and the state health departments and CDC. The 3 sites used viral culture or reverse transcription PCR (RT-PCR), with an increasing use of RT-PCR over the study period. We did not include influenza data from Tacoma and Snohomish counties. Public health respiratory virus surveillance was not conducted in the counties during the study period. We reviewed limited influenza testing data from the largest hospital systems in each county. Total influenza tests from Tacoma (23,741) and Snohomish counties (<3,000) were very low compared with those from King County (372,022) and were available for only part of our study period (2007–2008 and 2008–2012 for Tacoma and 2010–2013 for Snohomish). Influenza seasonality and peak seasons were similar in all 3 sites. Laboratory reports did not consistently distinguish between influenza A subtypes or influenza B lineages; therefore, we included only influenza A and B as exposure variables. The seasonality and temporal peaks of the proportion positive of influenza A and influenza B data among these sites were similar, so we aggregated each across all 3 counties for analyses.

RSV laboratory data were collected as part of routine care by the University of Washington Clinical Virology Laboratory and reported to the National Respiratory and Enteric Virus Surveillance System; these data were available for the period September 30, 2007–December 29, 2012 (26). RSV tests used antigen detection, viral culture, and RT-PCR testing, with RT-PCR use increasing over the period. RSV subtypes were not available.

We used weekly surveillance data for our model. We divided the weekly number of influenza A and influenza B detections by the weekly number of influenza tests performed and multiplied the result by 100 to calculate a weekly percentage of positive tests. We calculated the weekly percentage of positive RSV tests similarly.