A time series analysis of sulphur dioxide, temperature, and influenza incidence in 1976-1987

Patterns in the occurrence of influenza-like symptoms after increases in ambient sulphur dioxide (SO2) concentrations, have been investigated. Three routinely collected sets of data were used: (a) air pollution, including SO2, as measured continuously by the National Air Monitoring System; (b) ambient outdoor air temperature as measured hourly by the Royal Dutch Meteorological Institute and (c) weekly influenza incidence as monitored by the Dutch Sentinel Practice Network. The region investigated has been confined to the three western provinces of the Netherlands. The data were expressed as weekly average values. To correct for autocorrelation the 'Box-Jenkins approach' to model time series was used. After correction for seasonal patterns and a log transformation the ARMA (autoregressive and moving average) models were identified. The relation between the time series has been assessed by a transfer function model. After correction for temperature and with a delay of two weeks, a significant association could be ascertained between the hourly SO2 concentration and influenza incidence. An association between lower temperature and increased influenza incidence one week later was established as well. A relation of air pollution and influenza with a delay of about two weeks is quite possible. However, 12 years may have been too short a period to correct for secular patterns and SO2 was probably only an indicator of air pollution. Further investigation seems warranted.