Can routinely collected laboratory and health administrative data be used to assess influenza vaccine effectiveness? Assessing the validity of the Flu and Other Respiratory Viruses Research (FOREVER) Cohort
Kwong JC, Buchan SA, Chung H, Campitelli MA, Schwartz KL, Crowcroft NS, Jackson ML, Karnauchow T, Katz K, McGeer AJ, McNally DJ, Richardson DC, Richardson SE, Rosella LC, Simor A, Smieja M, Zahariadis G, Campigotto A, Gubbay JB. Vaccine. 2019; 37(31):4392-400. Epub 2019 Jun 17. DOI: 10.1016/j.vaccine.2019.06.011.
Background — Linking data on laboratory specimens collected during clinical practice with health administrative data permits highly powered vaccine effectiveness (VE) studies to be conducted at relatively low cost, but bias from using convenience samples is a concern. We evaluated the validity of using such data for estimating VE.
Methods — We created the Flu and Other Respiratory Viruses Research (FOREVER) Cohort by linking individual-level data on respiratory virus laboratory tests, hospitalizations, emergency department visits, and physician services. For community-dwelling adults aged > 65 years, we assessed the presence and magnitude of information and selection biases, generated VE estimates under various conditions, and compared our VE estimates with those from other studies.
Results — We included 65,648 unique testing episodes obtained from 54,434 individuals during the 2010-11 to 2015-16 influenza seasons. To examine information bias, we found the proportion testing positive for influenza for patients with unknown interval from illness onset to specimen collection was more similar to patients for whom illness onset date was ≤ 7 days before specimen collection than to patients for whom illness onset was > 7 days before specimen collection. To assess the presence of selection bias, we found the likelihood of influenza testing was comparable between vaccinated and unvaccinated individuals, although the adjusted odds ratios were significantly greater than 1 for some healthcare settings and during some influenza seasons. Over 6 seasons, VE estimates ranged between 36% (95%CI, 27-44%) in 2010-11 and 5% (95%CI, -2, 11%) in 2014-15. VE estimates were similar under a range of conditions, but were consistently higher when accounting for misclassification of vaccination status through a quantitative sensitivity analysis. VE estimates from the FOREVER Cohort were comparable to those from other studies.
Conclusions — Routinely collected laboratory and health administrative data contained in the FOREVER Cohort can be used to estimate influenza VE in community-dwelling older adults.
View full text