Josephine van Dongen

6 Non-specific effects of human rotavirus vaccination 145 Data collection Data collection for cohor t par ticipants included monthly parental questionnaires until 18 months of age. Parents answered yes/no questions concerning acute respiratory illness (ARI) symptoms, presence of eczema symptoms and hospital admission in the past month. ARI symptoms included fever with or without nasal congestion/runny nose, cough and earache. In addition the number of days with ARI symptoms in each month was recorded. For hospitalization, the reason and number of hospital days was also collected. Non-specific effect outcomes We defined non-specific effect outcomes of rotavirus vaccination as the relative change in incidence of parent-repor ted acute hospitalization,ARI or eczema between rotavirus vaccinated and unvaccinated infants. Acute hospitalization was defined as any hospital admission following initial post-natal discharge, excluding hospital admissions for acute gastroenteritis. Hospital admissions for scheduled medical or surgical interventions were also excluded. Analyses Descriptive statistics were used to compare patient characteristics and outcomes between vaccinated and unvaccinated infants.We calculated the cumulative incidence and incidence rate for each non-specific effect outcome by group status. As secondary outcomes, we calculated the cumulative acute hospitalization days and days with ARI symptoms for rotavirus vaccinated and unvaccinated infants. Next, we used Cox regression to model the effect of at least one dose of rotavirus vaccination on time to first hospital admission, with age (between two and 18 months) as time axis. Infants were censored when lost to follow up, dropped-out, deceased or hospitalized, whichever came first.The propor tional hazard assumption was visualized graphically and tested with Schoenfelds residuals. The final model with covariates was derived using log-likelihood ratio test, hazard ratios (HR) and their 95% confidence intervals (CI) were provided. Infants were categorized as rotavirus vaccinated from 28 days post-dose one onwards. For ARI, we used a negative binomial model with offset for persontime of observation to compare the number of months with ARI between rotavirus vaccinated and unvaccinated groups. Incidence rate ratios (IRR) were obtained with their 95% CIs. We used Akaike Information Criterion to select the final model. The analysis was repeated for the outcome months with eczema. For the secondary outcome of cumulative days hospitalized or days with ARI complaints we used a Poisson model. Because of overdispersion for hospitalization days we used a negative binomial model. In sensitivity analyses, we separately analysed effects up to six months of age, covering the