Too Hot, Too Wet: Bayesian Spatial Modelling of Climate-Driven Salmonella Risk in New South Wales, Australia, 1991-2022

Abstract

Salmonella infections contribute significantly to gastrointestinal-related hospitalisations in Australia and remain a major global public health concern. Although seasonal patterns in Salmonella incidence have been documented globally, there is limited evidence on the influence of climatic factors, particularly rainfall, humidity, flooding, and temperature, in the Australian context. This study investigated the relationship between climatic extremes and Salmonella infections across Local Health Districts (LHDs) in New South Wales (NSW), Australia, using a Spatial Bayesian Distributed Lag Non-Linear Model (SB-DLNM). Spatial modelling revealed a marked geographical heterogeneity in the risk of Salmonella related to climate in NSW. High ambient temperatures consistently increased risk, with 99th-percentile contrasts typically yielding relative risks (RR) of 2.4 - 4.8 across Local Health Districts (LHDs). Monthly rainfall showed the opposite direction statewide: very dry months were associated with a higher risk, whereas very wet months were generally protective (RR < 1). In contrast, discrete flooding events were strongly and positively associated with risk (99th-percentile flood index RR ∼ 18 - 23.5), with the greatest effects in some LHDs of the metropolitan/coastal region. Humidity displayed modest but consistent positive associations (99th-percentile RR ∼ 1.1 - 1.5). Temperature and humidity exhibited J-shaped exposure-response relationships, where the lowest risk occurred at moderate values. This contrasts with rainfall, which demonstrated an inverse (protective) association, and flooding, which showed a monotonic increase in risk with intensity. These results have important public-health implications under a warming, flood-prone climate.

Keywords: Salmonella, foodborne disease, climate variability, environmental drivers, climate change and health, spatial Bayesian