Monitoring influenza vaccine effectiveness using the national influenza surveillance system

Background: Due to the annual reformulation of the influenza vaccine composition estimates of the vaccine effectiveness (VE) are required every season. A portuguese influenza surveillance system is in place, and data obtained by this system may be used to evaluate VE (compared to studies specifically designed to this purpose). This study intends to evaluate the feasibility of using the national influenza surveillance system for monitoring the influenza VE (MonitorEVA). Methods: Two different study designs were implemented to estimate VE in season 2004-05 to 2011-12: Test negative design (TND) and screening method (SM). TND used a case-control approach where laboratory confirmed incident influenza like illness patients (ILI+) were compared to laboratory negative influenza ILI patients (ILI). Eligible individuals to this study consisted on individuals from all ages that consult a general practitioner or the emergency room (ER) with ILI symptoms. VE was estimated as 1-odds ratio of being vaccinated in cases versus controls adjusted for age and month of onset by logistic regression. For the SM, VE was estimated by comparing the proportion of vaccinated cases (recruited within TND) to the vaccine coverage in the source population (obtained from a sample of households, using a telephonic survey - ECOS sample), using the Orenstein formula and the Farrington method to adjust for age group. Results: Considering TND, crude point estimates of VE were under 40% in 2004/05, 2005/06 and 2011/12 season; between 50% and 70% in 2006/07, 2008/09 and 2010/11 seasons, and above 70% in the 2007/08 season. After adjustment for age group and month of onset, the VE point estimates decreased. SM estimates were in accordance to the TND ones but for the majority of the SM VE was lower than the TND. Discussion/Conclusions: Sample size and data quality are sufficient to obtain crude VE estimates with statistical significance (if VE is higher than 50%), however allowing less precise estimates. The surveillance data allowed the VE monitoring indicating if the VE was higher than 70% and less than 50%. Improvement of data quality in the surveillance program seems a potential way of improving precision and closing the gap between the two methods proposed