A regression-based approach to modelling recurrent outbreaks: a case study of dengue fever in Nepal (2006–2022)

Understanding the temporal and spatial dynamics of dengue fever outbreaks is essential for effective public health planning. In this study, we develop a novel, data-driven approach to characterise dengue outbreaks in Nepal from 2006 to 2022 by extending the standard Negative Binomial regression model to incorporate multiple outbreak intensity functions (OIFs). Each OIF is defined by three interpretable parameters: the peak timing, the scale regulating the duration of the outbreak, and a magnitude parameter that quantifies the contribution of the OIF to the overall epidemic. Unlike mechanistic transmission models, this framework makes minimal assumptions and offers greater flexibility when working with aggregated surveillance data. The application of this modelling approach to annual dengue case counts from all 77 districts of Nepal, revealed substantial heterogeneity in outbreak timing and intensity. Most districts experienced three major OIFs, with the third, typically in 2022, accounting for the largest share of the disease burden. The modelling approach is generalisable and can be applied to routinely reported data for other infectious diseases. We also outline how the framework could be extended for outbreak detection if higher-resolution temporal data were available.