AI Model Improves Outbreak Transmission Reconstruction Accuracy

A new research paper introduces a novel approach to improve the accuracy of reconstructing disease transmission pathways during outbreaks. The method utilizes a learned temporal prior, trained on various disease families, which can be applied without refitting to new outbreak data. This prior demonstrated significantly higher accuracy in identifying parent-child transmission relationships compared to existing methods. Beyond improving accuracy, the study also highlights a critical issue: the inherent uncertainty in current epidemiological transmission labels. An audit of mpox cases revealed that over half of the inter-host pairs were genomically unresolved or unsupported, indicating a need for more robust methods to assess label reliability. The research emphasizes that incorporating this measurable uncertainty into transmission graphs can alter the prioritization of source cases for intervention, suggesting a more nuanced approach to public health responses.