We assessed the association between monthly El Niño/Southern Oscillation (ENSO) and hepatitis A in Australia between 1991 and 2002. In order to identify the appropriate model for this particular data set, three conventional regression models were performed using the time-series data of the occurrence of hepatitis A and ENSO in Australia between 1 January 1991 and 31 December 2001. Model I was an Ordinary Least-Squares Regression model; Model II was a Poisson auto-regression time series model; and Model III was an Auto-Regressive Integrated Moving Average (ARIMA) time series model. Information on the hepatitis A cases was obtained from the Commonwealth Department of Health and Ageing. ENSO and population data were supplied by the Australian Bureau of Meteorology and the Australia Bureau of Statistics, respectively. The results show that model I (β = −1.42, p < 0.05) and model II (β = −0.01, p < 0.01) had a negative relationship between ENSO and hepatitis A, but the residuals analysis revealed that the assumptions underlying these statistics were violated (Ljung-Box p < 0.01). Model III (β = 1.72, p < 0.05) had a positive relationship between ENSO and hepatitis A and the residual analysis indicate the model fitted the data well (Ljung-Box p>0.05). The estimates from ARIMA time series model were much likely to represent the true relationships because the model took the auto-correlated errors into account.
School of Public Health, Queensland University of Technology