For reasons explained later our

modelling and NNV estimation subsequently required restriction to calendar week 46, 2003–calendar week 20, 2009. Since an influenza diagnosis may not have been established for all admitted with influenza, we combined hospitalizations with a main ICD-10 diagnosis of influenza and hospitalizations with a main diagnosis of a respiratory infection that can possibly be related to influenza (RIRI) (Table 1). Regardless of the number of times the diagnosed individuals were admitted and discharged during a calendar week, a maximum of one hospitalization episode per week and person was included. There is no register on all pregnancies Bosutinib in Sweden, but there is a Medical Birth Register. Therefore only pregnant women who had given birth in Sweden were eligible for our study. The register includes women who delivered a living child, or a deceased child after 27 weeks (before June 2008) or after 21 weeks (thereafter). The national registration numbers of the women who had given birth during the study period were collected from the Swedish Medical Birth Register and linked to the National Patient Register. Both registers are kept by the National Board of Health and Welfare (NBHW). Identified cases with a main diagnosis

Modulators belonging to either influenza or RIRI were categorized as such. Nearly all pregnant women in Sweden regularly attend prenatal care [20]. Nonetheless 3–8% of the women lacked a registered date of their last period, or an ultrasound estimated date of beginning of their pregnancy, EGFR inhibitor and were excluded from the study. Based on the date of the beginning of the pregnancy trimesters were approximated (1st: ≤84 days, 2nd: 85–182 days, 3rd: ≥183 days). Finally, the number of pregnant women was aggregated by calendar week, year and trimester. The data was extracted and aggregated old by the NBHW and thereafter delivered to the investigators. Since the study was carried out with aggregated data it did not require a review by an Ethics Review Board. To estimate the number of hospitalizations

with RIRI that could be attributed to influenza but for which the main diagnosis was not influenza, we fitted a generalized additive (GAM) quasi-Poisson regression model with identity link [21] to the RIRI hospitalizations. The model included: calendar week, which modelled the baseline with a cyclic penalized cubic regression spline function; and the weekly number of laboratory influenza reports with one parameter for each season, which modelled hospitalizations above the baseline that could be attributed to influenza. By using identity link we could assume that these hospitalizations were proportional to the laboratory influenza cases. We also calculated Wald confidence intervals for the proportions. During the included time period, 94–95% of all pregnant women were 20–39 years old [22].