Background: Dengue is a mosquito-borne viral disease transmitted by infected Aedes mosquitoes and is highly prevalent in tropical and subtropical regions, with Indonesia among the countries significantly affected. This study aims to assess temporal and spatial patterns in dengue haemorrhagic fever incidence rates and develop predictive models that could support early warning of dengue outbreaks.

Methods: This study utilised monthly district-level data on dengue haemorrhagic fever from January 2015 to December 2024. Temporal patterns were analysed using Seasonal-Trend decomposition using Loess (STL), while spatial structure was assessed using semivariogram. Spatial clustering patterns and hotspots were identified using Local Indicators of Spatial Association (LISA). For predictive modelling, oceanographic, climatic, socioeconomic, and demographic variables were included as candidate predictors. Oceanographic and climatic variables were modelled using the Distributed Lag Non-linear Model (DLNM) and incorporated into Generalized Linear Mixed Models (GLMM) and Generalized Additive Mixed Models (GAMM) with Negative Binomial and Poisson distributions. The dataset was divided into training (2015–2021) and testing (2022–2024) periods for model fitting and evaluation.

Results: Dengue haemorrhagic fever incidence rates in Indonesia showed a consistent seasonal pattern, typically peaking between January and June. Furthermore, spatial analysis indicated spatial dependence, suggesting that districts in closer geographic proximity exhibited more similar incidence rates than those situated farther apart, and the presence of clustering patterns. GLMM and GAMM, incorporating distributed lag structures, enabled one-month-ahead prediction of dengue haemorrhagic fever incidence across districts.

Conclusion: Dengue haemorrhagic fever incidence in Indonesia showed distinct spatial and temporal patterns. GLMM and GAMM, incorporating distributed lag structures, may provide valuable information to support early warning systems for dengue outbreaks.

Keywords: Dengue, temporal patterns, spatial analysis, mixed-effects models, DLNM