Evaluation of vaccine effectiveness in older adults using routinely collected data: a quasi-experimental approach

Abstract

Vaccination of older adults is a key component of public health policy, but further evidence is required to understand its effectiveness in practice. Electronic health records (EHRs) present a potential alternative to the gold-standard evidence of clinical trials, particularly for populations, such as older adults, who may be under-represented in trials due to ethical and practical constraints in recruitment. Importantly, EHRs also allow the real-world study of an intervention as it is delivered in practice, and its effect in clinically important sub-groups. However, EHRs are not purposed to collect informaton on confounders, which may bias results from the analyisis of routinely-collected data. This motivated my review of quasi-experimental (QE) methods as a means of indirectly adjusting for confounding. My published methodological review found that the longitudinal information available in EHRs offer many opportunities for mitigating against confounding bias, but many methods may be under-utilised. The prior event rate ratio (PERR) and its alternative formulation, described under the Pairwise framework, is a recently developed method that utilises longitudinal information. This before-and-after approach can be applied to rate and survival data, allowing an easy comparison to many trial results. The data on vaccination in UK older adults was also the basis for further study of the performance and limitations of the method beyond exisiting simulation studies. Through comparison to weighted regression, I demonstrated how the source of confounding and robustness of the results could be explored. In a novel application of the PERR and Pairwise methods to interactions, I investigated the effectiveness of the pneumococcal vaccine in older patients, and found evidence for an increase in effectiveness with age across the years of policy implementation, 2003-2005. In my investigation of the influenza vaccine in annual cohorts from 1997 to 2011, I found consistent evidence of a moderately protective effect against myocardial infarction, but that this may decrease with age. The evidence also indicated a protective effect against influenza itself, but no age trend in its effectiveness was detected.