Original Research Analysing dengue fever spread in Kenya using the Zero-Inflated Poisson model

Lameck Ondieki Agasa, Faith Thuita, Thomas Achia, Antony Karanja

Journal of Public Health in Africa | Vol 16, No 1 | a781 | DOI: https://doi.org/10.4102/jphia.v16i1.781 | © 2025 Lameck Ondieki Agasa, Faith Thuita, Thomas Achia, Antony Karanja | This work is licensed under CC Attribution 4.0
Submitted: 02 September 2024 | Published: 28 February 2025

About the author(s) Lameck Ondieki Agasa, Department of Public and Global Health, Faculty of Health Sciences, University of Nairobi, Nairobi, Kenya; and, Department of Community Health and Behavioral Sciences, School of Health Sciences, Kisii University, Kisii, Kenya
Faith Thuita, Department of Public and Global Health, Faculty of Health Sciences, University of Nairobi, Nairobi, Kenya
Thomas Achia, Institute of Mathematical Sciences, Strathmore University, Nairobi, Kenya
Antony Karanja, Department of Mathematics, Faculty of Science and Technology, Multimedia University, Nairobi, Kenya

Abstract

Background: Dengue fever (DF), transmitted by Aedes mosquitoes, remains a major public health concern in tropical and subtropical regions. Understanding the influence of climatic variables on DF incidence is essential for improving outbreak prediction and control measures.

Aim: This study analysed the impact of climatic factors on DF incidence in Kenya using a Zero-Inflated Poisson (ZIP) model.

Setting: The study focused on DF cases in Kenya from 2019 to 2021.

Methods: A ZIP model was applied to monthly dengue case data and associated climatic variables, such as temperature, rainfall, and humidity. The model addresses over-dispersion and excess zeros in the data, providing a more accurate depiction of DF dynamics.

Results: The ZIP model revealed significant associations between climatic variables and DF incidence. Humidity (β = 0.0578, standard error [s.e.] = 0.0024, z = 24.157, p < 2e-16) and temperature (β = 0.0558, s.e. = 0.0053, z = 10.497, p < 0.01) showed a positive relationship with dengue cases, while rainfall (β = –0.0045, s.e. = 0.0003, z = –16.523, p < 0.01) had a significant negative effect. The over-dispersion test confirmed excess variability in the data (O statistic = 456.3, p = 0.004), and the Vuong test supported the use of the ZIP model over a standard Poisson model. Model comparison indicated superior fit for the ZIP model (akaike information criterion [AIC] = 5230.959 vs. 27061.367 for Poisson), effectively accounting for zero-inflation.

Conclusion: The results suggest that higher humidity and temperature favor dengue transmission, while heavy rainfall may disrupt mosquito breeding, reducing cases. These findings provide a basis for targeted public health interventions.

Contribution: This study enhances understanding of DF-climate interactions in Kenya, supporting the application of ZIP modelling for improved disease surveillance and control strategies.

Keywords

dengue fever; Zero-Inflated Poisson model; climatic factors; epidemiology; Kenya.

Sustainable Development Goal

Goal 3: Good health and well-being

Metrics Total abstract views: 205
Total article views: 224

Crossref Citations

No related citations found.