Cohort studies of influenza vaccine effectiveness: Based on Finnish register data

Abstract

This thesis is motivated by the Finnish policy of estimating inuenza vaccine effectiveness, i.e., the vaccine-attributable reduction in inuenza incidence, each season from medical and demographic register data. It presents and examines methods that enable effcient use of such routinely collected data,including the assessment and control of confounding and information bias.

Register-based cohort studies are conducted to estimate the inuenza vaccine effectiveness in two-year-old children and elderly aged 65 years and older. With estimates ranging from less than 0% to 90 %, the results concerning children suggest high variability in vaccine effectiveness across different seasons, vaccines and virus types. As the cohorts of children are fairly homogeneous, confounding is deemed negligible. By contrast, the elderly cohorts are less homogeneous and confounders are thus considered. The presence of confounding is conffrmed by using off-season hospitalisation for acute respiratory infection as a negative control outcome. The confounder-adjusted analysis suggests that the inuenza vaccine effectiveness in elderly was greater than 0% but did not exceed 50% in recent seasons.

A weighted partial likelihood approach with probabilistic deletion of false positives is proposed to correct for information bias. This novel method allows unbiased estimation of vaccine effectiveness if the sensitivity of the outcome measurement and the rate at which false positives occur are known. If these parameters are unknown, the magnitude of information bias can be assessed for a range of plausible parameter values.

This thesis demonstrates the potential of the examined cohort study design to enable timely and population-based estimation of inuenza vaccine effectiveness in Finland. It also calls for validation data to ensure study validity.