Estimation of influenza vaccine effectiveness from routine surveillance data.

Abstract

BACKGROUND: Influenza vaccines are reviewed each year, and often changed, in an effort to maintain their effectiveness against drifted influenza viruses. There is however no regular review of influenza vaccine effectiveness during, or at the end of, Australian influenza seasons. It is possible to use a case control method to estimate vaccine effectiveness from surveillance data when all patients in a surveillance system are tested for influenza and their vaccination status is known. METHODOLOGY/PRINCIPAL FINDINGS: Influenza-like illness (ILI) surveillance is conducted during the influenza season in sentinel general practices scattered throughout Victoria, Australia. Over five seasons 2003-7, data on age, sex and vaccination status were collected and nose and throat swabs were offered to patients presenting within three days of the onset of their symptoms. Swabs were tested using a reverse transcriptase polymerase chain reaction (RT-PCR) test. Those positive for influenza were sent to the World Health Organization (WHO) Collaborating Centre for Reference and Research on Influenza where influenza virus culture and strain identification was attempted. We used a retrospective case control design in five consecutive influenza seasons, and estimated influenza vaccine effectiveness (VE) for patients of all ages to be 53% (95% CI 38-64), but 41% (95% CI 19-57) adjusted for age group and year. The adjusted VE for all adults aged at least 20 years, the age groups for whom a benefit of vaccination could be shown, was 51% (95% CI 34-63). Comparison of VE estimates with vaccine and circulating strain matches across the years did not reveal any significant differences. CONCLUSIONS/SIGNIFICANCE: These estimates support other field studies of influenza vaccine effectiveness, given that theoretical considerations suggest that these values may underestimate true effectiveness, depending on test specificity and the ratio of the influenza ILI attack rate to the non-influenza ILI attack rate. Incomplete recording of vaccination status and under-representation of children in patients from whom a swab was collected limit the data. Improvements have been implemented for prospective studies.