El Niño Southern Oscillation and the transmission of hepatitis A virus in Australia

To the Editor: We examined the possible association between the Southern Oscillation Index (SOI) and the occurrence of hepatitis A in Australia using a Seasonal AutoRegressive Integrated Moving Average (SARIMA) regression model.1 Our results indicate that the SOI is statistically significantly associated with the transmission of hepatitis A.

We obtained data on the monthly counts of hepatitis A cases in Australia and the monthly SOI between 1 January 1991 and 31 December 2000 from the Commonwealth Department of Health and Ageing and the Australian Bureau of Meteorology, respectively. Data on population sizes were obtained from the Australian Bureau of Statistics. Cross-correlations were used to compute a series of correlations between SOI and the incidence of hepatitis A over a range of time lags (defined as the time span between the SOI and the incidence of hepatitis A).

A SARIMA model was used to estimate the independent contribution of SOI in our study. We adjusted for seasonality by “seasonally differencing” (ie, replacing each observation by the difference between it and the observation from the previous year). In the modelling process, attention was paid to observations well outside the main body of the data (outliers) and the only outlier was excluded in the final SARIMA model. We used the SARIMA [1,0,0] [1,1,0]12 model (ie, first-order autoregressive combined, first-order seasonal autoregressive, after adjustment for first-order seasonal integration) to assess the association between SOI and the incidence of hepatitis A.

We found that a decrease in the SOI (ie, warmer and drier conditions) was statistically significantly associated, at a lag of 1 month, with an increase in the monthly incidence of hepatitis A (β = − 0.01; P = 0.001). Two El Niño events (1991–92 and 1997–98) were also clearly associated with an increased incidence of hepatitis A (Box).

The results suggest that there was an increase of about 360 cases per year in Australia for an, on average, interquartile range decrease in the SOI. The residuals in the model fluctuated randomly around zero, and there was no apparent autocorrelation between residuals at different lag times (data are available from the corresponding author).

These results indicate that the model fitted the data well, with no violation of assumptions. The significant association between SOI and the incidence of hepatitis A remained when the outlier was included in the model (β = − 0.013; P = 0.001).

El Niño Southern Oscillation (ENSO) has been found to be related to various health outcomes, including waterborne disease, vectorborne disease, and natural disaster-related deaths (eg, floods, bushfires and cyclones).2-4 Our study adds further evidence of ENSO-related health effects. Infectious diseases are, in general, sensitive to climate variability, as climate can influence the development and transmissibility of pathogens, and can also affect people’s behaviour.3,5 If the relationship between ENSO and hepatitis A is confirmed by other studies, these findings may facilitate the development of early warning systems for controlling and preventing this widespread communicable disease.

Incidence of hepatitis A in Australia and the Southern Oscillation Index between January 1991 and December 2000

Negative values of the SOI are associated with El Niño conditions (dry and warm in Australia), and positive values with La Niña conditions (wet and less warm in Australia).