Conference Abstract | Volume 8, Abstract ELIC2025155 (Poster 124) | Published: 08 Aug 2025
Jibrin Jaafaru1,2, &
¹Department of Computer Science, University of Jos, Jos, Nigeria, ²West African Centre for Emerging Infectious Diseases, Jos University Teaching Hospital, Jos, Nigeria
&Corresponding author: Jibrin Jaafaru, Department of Computer Science, University of Jos, & West African Centre for Emerging Infectious Diseases, Jos, Nigeria. Email: jibrinx@yahoo.com
Received: 24 Mar 2025, Accepted: 09 Jul 2025, Published: 08 Aug 2025
Domain: Infectious Disease Epidemiology
Keywords: Lassa fever, Bayesian inference, under-reporting, simulation-based modelling, SNPE, Forecast
©Jibrin Jaafaru et al. Journal of Interventional Epidemiology and Public Health (ISSN: 2664-2824). This is an Open Access article distributed under the terms of the Creative Commons Attribution International 4.0 License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Cite this article: Jibrin Jaafaru et al., Inferring underreporting in Lassa fever surveillance in Nigeria via simulation-based modelling with climatic covariates. Journal of Interventional Epidemiology and Public Health. 2025;8(Conf Proc 5):00268. https://doi.org/10.37432/JIEPH-CONFPRO5-00268
Lassa fever remains a persistent public health challenge in West Africa, particularly in Nigeria, where outbreaks recur annually with varying intensity. Effective control and timely response depend on accurate epidemiological modeling, yet the reliability of reported case data is hindered by substantial under-reporting and inconsistent surveillance. Environmental factors such as rainfall, temperature, and vegetation strongly influence both disease transmission and the likelihood of detection, yet their role in shaping reporting patterns remains underexplored. Bridging this gap is critical for interpreting observed trends and guiding targeted public health interventions.
We propose a Bayesian simulation-based inference (SBI) framework that extends the classical SEIR (Susceptible–Exposed–Infectious–Recovered) model with a covariate-conditioned, time-varying reporting function. This function is parameterized by a neural network that maps weekly environmental covariates including rainfall, temperature, and vegetation index (NDVI) to the probability of reported cases, thereby capturing seasonal and regional variation in surveillance performance. To estimate the joint posterior over epidemiological parameters and reporting dynamics, we use Sequential Neural Posterior Estimation (SNPE) on historical Lassa fever data from Nigeria Center for Disease Control.
Our method enables likelihood-free inference in the presence of latent, noisy, and partially observed data. Posterior predictive checks show strong agreement between observed and simulated case trajectories. Inferred reporting functions reveal geographic and temporal patterns of under-reporting.
By integrating climatic covariates into a neural network–parameterized reporting function within a Bayesian simulation-based framework, our approach disentangles true transmission dynamics from reporting noise. This allows us to infer spatiotemporal patterns of under-reporting and identify regions where surveillance is likely weakest. These findings highlight the dual value of environmental data in informing both transmission modeling and surveillance evaluation, with direct implications for improving epidemic forecasting, resource prioritization, and policy response to Lassa fever in Nigeria.
Menu