Influence of spatial connectivity on the spread of Mpox in Maï-Ndombe province, Democratic Republic of the Congo, 2025.

Abstract

Background: Human mobility is a major determinant of the spatial spread of emerging infectious diseases. In Maï-Ndombe province, Democratic Republic of the Congo, dependence on waterways as the primary transport network, combined with a degraded road infrastructure and marked environmental constraints, creates profound heterogeneity in spatial connectivity between health zones. In this context, the spread of Mpox shows an irregular spatial distribution whose structural mechanisms remain poorly understood, partly due to the scarcity of mobility data and the fragility of surveillance systems.

Objective: To quantify the intra-provincial spatial connectivity of Maï-Ndombe by simultaneously integrating road and river networks, and to assess its influence on the spread of Mpox between 2022 and 2025.

Methods: A spatial connectivity analysis based on an enhanced gravity model integrating demographic attractiveness and a distance-cost factor accounting for slope and land use via the Fuzzy-AHP method was combined with a spatialized metapopulation SEIR model. Inter-zone flows, centrality, and accessibility were quantified. The performance of SEIR models with and without connectivity was compared using RMSE, MAE, and precision gain per zone.

Results: Flows are strongly concentrated around pivotal zones (Bokoro, Nioki, Mushie), while peripheral areas (Mimia, Oshwe, Kiri) remain structurally isolated, with accessibility provided primarily by river corridors (Gini index = 0.65; CV = 1.52). The integration of connectivity degrades the overall model fit (ΔRMSE = -0.468), reflecting the predominance of local transmission at the provincial scale. However, this aggregate degradation conceals marked spatial heterogeneity: connectivity significantly improves predictions in highly connected zones (Bokoro: +59.8%; Mushie: +33.3%; Nioki: +17.3%), while degrading them in poorly connected zones (Kiri: -46.1%; Mimia: -27.4%; Oshwe: -22.4%).

Conclusion: Spatial connectivity does not strengthen the overall prediction of incidence, but structures the spatial redistribution of epidemic risk. The aggregate model degradation reflects both the limitations of the surveillance system and structural constraints of the gravity model, particularly the use of static populations as denominators and the absence of temporal exposure duration at destination zones. These results highlight the importance of targeting surveillance interventions in high-centrality areas, while developing metapopulation approaches that integrate the temporal dimension of mobility to improve preparedness for future Mpox outbreaks in tropical forest environments.