Seasonal influenza vaccine effectiveness among persons with COPD in Thailand: a retrospective national cohort study

Abstract

The effectiveness of trivalent inactivated influenza vaccine against respiratory illness outpatient visits and hospitalizations among patients with chronic obstructive pulmonary disease (COPD) is less established in Thailand. Using the Thailand National Health Security Office’s database, we assembled a retrospective cohort of Thai adults aged ≥18 years with COPD prior to each of the 2011 to 2013 influenza seasons. We used ICD-10 codes (J40, J41, and J44.1) to identify participants from all public COPD clinics in Thailand, and abstracted data on COPD history, hospital visits and admissions, and influenza vaccination status. Influenza vaccine effectiveness (VE) was estimated by comparing ICD-10 coded discharge diagnosis for influenza and pneumonia (J9-J18) and COPD exacerbations (J44.1) in each season between vaccinated and unvaccinated COPD patients using Poisson regression analysis. Propensity scores were used to match vaccinated with unvaccinated COPD patients by age, sex, smoking status, severity of COPD, and numbers of COPD exacerbations and hospitalizations in the previous year. Overall, 117,894 COPD patients were included in the cohort; the median age was 70 years (inter-quartile range 61-77) and 90,161 (77%) were male. Influenza vaccination coverage was 35% in 2011 and 41% in 2012 and 2013. The predominant circulating strain in 2011 and 2013 was A(H3N2), well-matched by the vaccine strain. In 2012, the predominant circulating strain was influenza B, but the vaccine strain had a poor antigenic match. Overall, the pooled VE for three years against ICD-coded influenza, pneumonia, and COPD exacerbations were estimated at 20% (95% confidence interval [CI], 6%-34%), 35% (95% CI, 34%-36%), and 38.8% (95% CI, 38.3%-39.3%), respectively. Overall, influenza vaccination was associated with a modest protection among people with COPD against ICD-10 coded influenza, pneumonia, and COPD exacerbations. However, despite propensity score matching, our results may be affected by residual confounding and some measurement bias.