Air pollutants and health outcomes: Assessment of confounding by influenza

We assessed confounding of associations between short-term effects of air pollution and health outcomes by influenza using Hong Kong mortality and hospitalization data for 1996-2002. Three measures of influenza were defined: (i) intensity: weekly proportion of positive influenza viruses, (ii) epidemic: weekly number of positive influenza viruses ≥4% of the annual number for ≥2 consecutive weeks, and (iii) predominance: an epidemic period with co-circulation of respiratory syncytial virus <2% of the annual positive isolates for ≥2 consecutive weeks. We examined effects of influenza on associations between nitrogen dioxide (NO2), sulfur dioxide (SO2), particulate matter with aerodynamic diameter ≤10 μm (PM10) and ozone (O3) and health outcomes including all natural causes mortality, cardiorespiratory mortality and hospitalization. Generalized additive Poisson regression model with natural cubic splines was fitted to control for time-varying covariates to estimate air pollution health effects. Confounding with influenza was assessed using an absolute difference of >0.1% between unadjusted and adjusted excess risks (ER%). Without adjustment, pollutants were associated with positive ER% for all health outcomes except asthma and stroke hospitalization with SO2and stroke hospitalization with O3. Following adjustment, changes in ER% for all pollutants were <0.1% for all natural causes mortality, but >0.1% for mortality from stroke with NO2and SO2, cardiac or heart disease with NO2, PM10and O3, lower respiratory infections with NO2and O3and mortality from chronic obstructive pulmonary disease with all pollutants. Changes >0.1% were seen for acute respiratory disease hospitalization with NO2, SO2and O3and acute lower respiratory infections hospitalization with PM10. Generally, influenza does not confound the observed associations of air pollutants with all natural causes mortality and cardiovascular hospitalization, but for some pollutants and subgroups of cardiorespiratory mortality and respiratory hospitalization there was evidence to suggest confounding by influenza.