Refining early detection of Marburg Virus Disease (MVD) in Rwanda: Leveraging predictive symptom clusters to enhance case definitions.

Abstract

Background: Marburg Virus Disease (MVD) poses a significant global health risk due to its high case fatality rates (24%-88%) and the diagnostic challenges posed by its non-specific early symptoms, which overlap with other febrile illnesses like malaria. This study analyzed symptom patterns from the 2024 MVD outbreak in Rwanda to refine case definitions and enhance early detection.

Methods: A retrospective analysis was conducted of 6,613 suspected MVD cases (66 positive, 6,547 negative) reported between September 27 and December 20, 2024. Symptom prevalence and predictive value were assessed using multiple logistic regression models with L1 and L2 regularization to identify the most predictive symptoms. Models were validated using 5-fold cross-validation, with performance assessed through ROC analysis and standard accuracy metrics.

Results: Fever (78.8%), fatigue (63.6%), and headache (57.6%) were identified as the most common early symptoms, while hemorrhagic signs were rare (3.0%). The model achieved high accuracy (99.04%) and an AUC-ROC of 0.824, identifying fever, fatigue, nausea/vomiting, joint pain, and sore throat as key predictors.

Conclusion: Early symptom clusters, especially constitutional and gastrointestinal signs outperformed hemorrhagic symptoms for MVD detection. Findings challenge current case definitions, emphasizing the need for revised public health messaging and healthcare worker training. Integrating symptom-based models into surveillance could enhance detection, especially in resource-limited settings.