e. calendar weeks 40–20, for seasons 2003/04–2008/09, were collected BI-6727 for the 20–39 years age group. This laboratory surveillance data was collected from the Swedish Institute for Communicable Disease Control and linked to the weekly patient data. Data by age group was only available from calendar week 46, 2003 and onwards, and data beyond calendar week 20, 2009 were excluded to avoid the inclusion of the pandemic influenza A(H1N1)pdm09. The estimated proportions were multiplied with the weekly number of laboratory influenza cases, resulting in the weekly number of RIRI hospitalizations

attributed to influenza among pregnant women. The weekly numbers were then aggregated per season. For each season, 2003/04–2008/09 we also extracted the total number of main diagnoses of influenza in the register data during the extended season, defined

as the time between calendar week 27 one year to calendar week 26 the following year. In 2009 the last included week was week 20. There were no influenza diagnoses outside the surveillance season. We then added the influenza diagnoses in each extended season to the estimated RIRI hospitalizations attributed to influenza, calculated from the model, and thereby obtained an estimate of the total number of influenza hospitalizations of pregnant women per season. As part of our main analysis we also calculated the NNV per season [23] equation(1) NNVi=1VEicasesink,where VE = vaccine effectiveness against influenza, cases = total number of influenza hospitalizations per season, n = number of unvaccinated pregnant women, selleck chemical i = season and k = year the

season turned into. We assumed that all pregnant women were unvaccinated, below and thus n was the number of pregnant women between 2003 and 2009. The VE was allowed to vary in order to carry out a sensitivity analysis: 40–80%. This wide range of VE was chosen since estimations of the VE and its confidence intervals have varied widely between studies [24] and [25] and the match to the circulating subtype of influenza may vary. We also calculated the mean NNV using the average n and the average cases. To create the possible worst and best case scenarios of NNV, we first calculated the 95% confidence intervals of number of hospitalizations attributable to influenza for each season. For the worst possible scenario, the most severe season, we substituted the cases parameter for the maximum of all confidence interval limits; and for the best possible scenario, the mildest season, the minimum of all limits. Each scenario included the previously described range of VE. As subanalyses we calculated the total number of influenza hospitalizations by the first, second and third trimesters. For our analysis we used STATA IC 10 and R 2.15.0 with package mgcv 1.7–22. During 2000–2009 the yearly incidence of pregnant women who delivered a child ranged from 87,866–109,594.