Influenza vaccine effectiveness by test-negative design - Comparison of inpatient and outpatient settings

Background: Observational studies of influenza vaccine effectiveness (VE) are increasingly using the test negative design. Studies are typically based in outpatient or inpatient settings, but these two approaches are rarely compared directly. The aim of our study was to assess whether influenza VE estimates differ between inpatient and outpatient settings. Methods: We searched the literature from Medline, PubMed and Web of Science using a combination of keywords to identify published studies of influenza VE using the test-negative design. Studies assessing any type of influenza vaccine among any population in any setting were considered, while interim studies or re-analyses were excluded. Retrieved articles were reviewed, screened and categorized based on study setting, location and influenza season. We searched for parallel studies in inpatient and outpatient settings that were done in the same influenza season, in the same location, and in the same or similar age groups. For each of the pairs identified, we estimated the difference in VE estimates between settings, and we tested whether the average difference was significant using a paired t-test. Results: In total 25 pairs of estimates were identified that permitted comparisons between VE estimates in inpatient and outpatient study settings. Within pairs, the prevalence of influenza was generally higher among patients enrolled in the outpatient studies, while influenza vaccination coverage among the test negative control groups was generally higher in the inpatient studies. There was no heterogeneity in the paired differences in VE, and the pooled difference in VE between inpatient and outpatient studies was -2% (95% confidence interval: -12%, 10%). Conclusions: We found no differences in VE estimates between inpatient and outpatient settings by studies using the test-negative design. Further research involving direct comparisons of VE estimates from the two settings in the same populations and years would be valuable. (C) 2016 Elsevier Ltd. All rights reserved.