Abstract Influenza remains a significant and recurrent public health burden in temperate regions. Meteorological factors such as temperature, humidity, and rainfall are recognised as associated with influenza transmission patterns, exhibiting complex, nonlinear, temporally lagged, and spatially heterogeneous effects. This study employed a Spatial Bayesian Distributed Lag Non-Linear Model (SB-DLNM) to investigate the associations between meteorological factors and influenza incidence across 15 Local Health Districts, New South Wales, Australiathe short-term meteorological variables on influenza incidence across multiple Local Health Districts within New South Wales, Australia. The method incorporates (i) cross-basis functions to model delayed and non-linear meteorological impacts; (ii) a comparative analysis of case-crossover and time-series designs to distinguish monthly-lag associations from broader temporal trends; and (iii) spatial partial pooling to enhance the stability of estimates, particularly in data-sparse regions. Temperature demonstrated the strongest associations with influenza risk (Relative Risk (RR) range: 1.16–3.90), with elevated risks observed predominantly at cold temperature extremes. While exposure-response curves suggest minimum risk at moderate temperatures ( <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(18-22^{\circ }\hbox {C}\)</EquationSource> </InlineEquation>), the available data primarily capture cold-related effects; warm-temperature associations remain uncertain due to limited extreme heat observations. Humidity showed marked spatial heterogeneity with variable effects across districts (RR range: 1.32–5.69), while rainfall demonstrated minimal associations (RR typically 1.03–1.42). Exceedance probabilities for RR>1 were moderate across all variables, ranging from 17.5% to 58%, with no extreme hot spots observed. Partial pooling effectively stabilised estimates in sparse datasets, improving the robustness of spatial risk assessment. These findings underscore the importance of cold temperatures in influenza transmission patterns, providing a robust framework for public health surveillance. Our use of monthly aggregated data captures population-level seasonal associations rather than acute exposure-infection dynamics, which represents an important interpretive constraint.Among the meteorological variables, temperature emerged as the strongest predictor of influenza risk, with peak incidence observed within moderate temperature ranges ( <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(20-22^{\circ }\hbox {C}\)</EquationSource> </InlineEquation>) Graphical Abstract A schematic overview of the workflow from merging meteorological and influenza data, evaluating four modelling approaches (with Model 3 highlighted as the best), to generating spatial risk maps and relative risk estimates for influenza in NSW.

Meteorological Drivers of Influenza Spread in New South Wales: A Spatial Bayesian Distributed Lag Non-linear Model Approach

期刊论文
出版模式：

开放获取
发表日期：

2026年2月25日

查看全文

Mohammad Khan,
Oyelola Adegboye,
Shiyang Lyu,
Kiki Adhinugraha,
Theophilus Emeto,
David Taniar

摘要

Abstract

Influenza remains a significant and recurrent public health burden in temperate regions. Meteorological factors such as temperature, humidity, and rainfall are recognised as associated with influenza transmission patterns, exhibiting complex, nonlinear, temporally lagged, and spatially heterogeneous effects. This study employed a Spatial Bayesian Distributed Lag Non-Linear Model (SB-DLNM) to investigate the associations between meteorological factors and influenza incidence across 15 Local Health Districts, New South Wales, Australiathe short-term meteorological variables on influenza incidence across multiple Local Health Districts within New South Wales, Australia. The method incorporates (i) cross-basis functions to model delayed and non-linear meteorological impacts; (ii) a comparative analysis of case-crossover and time-series designs to distinguish monthly-lag associations from broader temporal trends; and (iii) spatial partial pooling to enhance the stability of estimates, particularly in data-sparse regions. Temperature demonstrated the strongest associations with influenza risk (Relative Risk (RR) range: 1.16–3.90), with elevated risks observed predominantly at cold temperature extremes. While exposure-response curves suggest minimum risk at moderate temperatures ( \(18-22^{\circ }\hbox {C}\) ), the available data primarily capture cold-related effects; warm-temperature associations remain uncertain due to limited extreme heat observations. Humidity showed marked spatial heterogeneity with variable effects across districts (RR range: 1.32–5.69), while rainfall demonstrated minimal associations (RR typically 1.03–1.42). Exceedance probabilities for RR>1 were moderate across all variables, ranging from 17.5% to 58%, with no extreme hot spots observed. Partial pooling effectively stabilised estimates in sparse datasets, improving the robustness of spatial risk assessment. These findings underscore the importance of cold temperatures in influenza transmission patterns, providing a robust framework for public health surveillance. Our use of monthly aggregated data captures population-level seasonal associations rather than acute exposure-infection dynamics, which represents an important interpretive constraint.Among the meteorological variables, temperature emerged as the strongest predictor of influenza risk, with peak incidence observed within moderate temperature ranges ( \(20-22^{\circ }\hbox {C}\) )

Graphical Abstract

A schematic overview of the workflow from merging meteorological and influenza data, evaluating four modelling approaches (with Model 3 highlighted as the best), to generating spatial risk maps and relative risk estimates for influenza in NSW.