Network-based modelling of Bundibugyo Ebola virus disease importation and spread in Uganda using Displacement Tracking Matrix flow data
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
The 2026 Bundibugyo Ebola outbreak in Uganda, linked to the ongoing epidemic in the Democratic Republic of the Congo (DRC), spread through human mobility across borders and within the country. We aimed to quantify importation and spread patterns to inform targeted interventions at subnational level.
Methods
We constructed a data-driven directed weighted mobility network using IOM Displacement Tracking Matrix (DTM) flows collected from 15–24 May 2026 (11,245 observed movements) and the 2024 Uganda census (45.9 million people). A stochastic metapopulation SEIR model, incorporating pre-symptomatic transmission and movement of exposed and infectious individuals, was simulated over 90 days across 135 Ugandan districts and two DRC provinces. The 7-1-7 agile response was explicitly modelled with dynamic contact tracing coverage (40%→70%→85%). We evaluated three compliance scenarios (20%, 40%, 60%) for non-pharmaceutical interventions. Sobol global sensitivity analysis (500 samples, 200 bootstraps) identified key parameters driving outbreak size.
Findings
The mobility network was sparse (density 0.11), highly unequal (Gini coefficient 0.67), and modular (modularity 0.5). Kisoro district had the highest import risk (in-strength 3,823) and export risk (out-strength 1,350), while Kampala showed substantial in-strength (1,290) but lower out-strength (150). Under the containment scenario (effective R = 0.35), the model projected a median of 22 cumulative cases (95% CrI: 22–24) and 2 deaths (95% CrI: 2–4) in Kampala over 90 days. All other 134 districts had a median of 0 cases. Non-pharmaceutical interventions at high, moderate, and low compliance produced no statistically significant reduction in cases (22 cases across all scenarios). Superspreading events occurred in 34.6–40.6% of simulations. Sobol sensitivity analysis identified the infectious period (first-order index 0.838), case fatality rate (0.738), and basic reproduction number (0.664) as the most influential parameters; mobility-related parameters had substantially lower total-order indices. The agile 7-1-7 response alone is projected to contain the Bundibugyo Ebola outbreak to approximately 22 cases and 2 deaths in Kampala, with additional non-pharmaceutical interventions providing no significant added benefit. Resources should focus on targeted surveillance at high-risk importation hubs (Kisoro for border screening) and inland epidemic centres (Kampala for response capacity), rather than untargeted nationwide interventions. Investment in rapid strain-specific characterisation of biological parameters could improve predictive accuracy.
Funding
None.
Research in Context
Evidence before this study
We searched PubMed and WHO Disease Outbreak News (May 19, 2026) for outbreak modelling studies of Bundibugyo virus (BDBV) using the terms “Bundibugyo”, “ebolavirus”, “spillover”, “cross-border”, AND “stochastic model”. Previous outbreaks in Uganda (2007–08; 131 confirmed cases, 42 deaths) and DR Congo (2012; 38 confirmed cases, 13 deaths) suggested lower transmissibility and case-fatality ratio than Zaire ebolavirus, with reproduction number estimates ranging from 1.2 to 2.6. The 2018–20 North Kivu–Ituri Zaire ebolavirus outbreak—the second-largest EVD outbreak on record, with 3,481 cases and 2,299 deaths—was severely complicated by armed conflict, community mistrust, poor infrastructure, and cross-border population movement. Chamba et al. (2026) estimated a 94.2% importation probability for Uganda using a stochastic ensemble model calibrated to laboratory-confirmed cases. However, no published study has integrated IOM Displacement Tracking Matrix flow data with epidemic projections to provide subnational importation and spread risk estimates. Real-time cross-border spillover modelling for BDBV has been scant, and the absence of a licensed vaccine against BDBV complicates preparedness and response strategies.
Concurrently, Bangelesa and colleagues conducted a population mobility mapping exercise in Ituri Province using monthly data from Flowminder (March 2025–March 2026) and computed a normalised mobility intensity index (MII) for all 340 health zones. 1 Their analysis revealed that Ituri Province had the second highest average MII nationally (25.3), behind only Kinshasa (26.6), and 2.4 times the national median (10.7). Critically, the three outbreak epicentre health zones—Rwampara (MII 68.6, rank 2), Bunia (MII 62.6, rank 3), and Mongbwalu (MII 49.9, rank 11)—all exceeded the 95th national percentile (41.4), underscoring the exceptional connectivity of the affected zones.
Added value of this study
While Bangelesa et al. established the macro-level mobility intensity of Ituri Province relative to the rest of DR Congo, our study extends this work by constructing a directed, weighted mobility network at the district level within Uganda using IOM Displacement Tracking Matrix flow data (11,245 observed movements). We demonstrate that the mobility network is extremely unequal (Gini 0.67) and modular (modularity 0.5), with Kisoro as the primary importation gateway (in-strength 3,823) and Kampala as the highest-burden district (22 median cases). We further show that the 7-1-7 agile response alone is projected to contain the outbreak to approximately 22 cases, with additional non-pharmaceutical interventions providing no statistically significant added benefit. The Sobol sensitivity analysis identifies the infectious period (first-order index 0.838), case fatality rate (0.738), and basic reproduction number (0.664) as the most influential parameters, informing targeted data collection priorities. This is the first study to provide district-level operational guidance for BDBV outbreak response using real-time mobility data.
Implications of all the available evidence
Preparedness planning should prioritise targeted surveillance at high-risk importation hubs (Kisoro for border screening) and inland epidemic centres (Kampala for response capacity), rather than untargeted nationwide interventions. The extreme connectivity of affected zones—as quantified by Bangelesa et al., with Ituri Province having the second highest MII nationally and the three epicentre health zones all exceeding the 95th percentile—makes contact tracing extraordinarily resource-intensive. The absence of a licensed vaccine against Bundibugyo virus eliminates the ring vaccination strategy that has been effective in all six preceding Ebola outbreaks in DR Congo caused by Zaire ebolavirus. Rapid strain-specific characterisation of biological parameters through contact tracing and serial interval analysis could yield greater returns than refining operational protocols. Enhanced cross-border surveillance under International Health Regulations 2005 and the Africa CDC Public Health Emergency of Continental Security framework is crucial to harmonise detection and response across the DR Congo–Uganda–South Sudan tripoint.
