BACKGROUND: Population-based surveillance for respiratory illness helps inform disease burden estimation for influenza and other respiratory viruses. While broadly defined acute respiratory illness (ARI) is a common standard of surveillance, influenza-like illness (ILI) is a predictive and cost-effective case definition used for influenza surveillance. The Influenza Incidence Surveillance Project (IISP) conducts surveillance for ILI through a network of outpatient healthcare providers (HCPs) representing state and local health department jurisdictions, a subset of which also report ARI. We describe a preliminary analysis to model a ratio of ARI to ILI visits that allows for ARI rate estimation in sites conducting surveillance only for ILI.
METHODS: From 2010 through 2015, 8-12 health departments each recruited 4-7 HCPs to report weekly counts of all-cause and ILI visits (reported fever with cough or sore throat); ARI visit counts (two respiratory symptoms without meeting ILI criteria) were collected from 11 sites in 2010-11 and 2-4 sites in remaining years. Using data during October through May from sites conducting ARI and ILI surveillance, we developed a linear regression model to predict the ratio of ARI to ILI by HCP and week, controlling for HCP characteristics, surveillance season, census region, weekly patient all-cause and ILI visits, and weekly proportion of a patient subset testing positive for influenza. Models were evaluated based on goodness of fit and correlation with observed ARI rates. Patient population was used to calculate estimated ARI rates for all IISP sites.
RESULTS: A total of 780,881 patient visits occurred, including 17,236 and 43,233 ILI and ARI visits, respectively. Significant covariates included in the final model were surveillance season, weekly all-cause visit counts, weekly percent ILI, HCP’s percent of patients aged <18 years, and the weekly percent influenza test-positive. The median predicted ARI to ILI ratio was 2.3 (CI 1.9-2.9), and was highest among student health, urgent and large primary care centers as compared with small primary care clinics. Observed rates of ARI visits ranged by season from 17 to 41 per 1000 population, and were consistent with estimated rates obtained by applying the model ratio to both the ARI sites (range 15 to 41 per 1000 population) and all IISP sites (range 32 to 57 per 1000 population).
CONCLUSIONS: Modeling the ARI to ILI ratio allows for estimation of ARI visit rates from the IISP, providing a comparison to ARI-based respiratory surveillance networks and a possible future source for outpatient ARI burden estimates.