Constructing event graph and mining critical paths of Ebola virus disease: a public health emergency analysis based on WHO disease outbreak news

Most Ebola virus disease research focuses on clinical diagnosis or static knowledge networks, limiting dynamic spatiotemporal analysis of epidemic evolution. We examined whether an event graph derived from World Health Organization (WHO) Disease Outbreak News (DONs) could reconstruct outbreak evolution across reports and identify response-timing patterns, including delayed interventions, unactioned risk assessments, and cross-border escalation.

We retrospectively analysed 346 Ebola-related DON reports published through August 2025. We used a large language model-assisted framework to extract events and their relationships. We built an event graph in Neo4j and applied a bounded depth-first search procedure that removed cycles, generated candidate paths, and selected critical paths across six dimensions: case burden, spatial spread, international risk signals, intervention measures, path length, and graph structural importance.

The graph comprised 2233 event nodes and 10,479 relations across 155 event groups; 47 representative critical paths were selected. Laboratory confirmation often marked the shift from uncertain signals to an organised response. Healthcare-facility exposure and community deaths served as sentinel events for infection-control review and transmission-chain investigation. Preparedness, surveillance, and contact tracing generally followed spatial spread and international-risk signals rather than general border restrictions. Repeated burden updates without documented interventions indicated potential response or reporting gaps. Although established control measures aligned with WHO advice, the report defined explicit triggers, deadlines, inter-agency handoffs, and completion indicators less clearly.

The event graph reconstructed longitudinal outbreak and response trajectories that are difficult to capture through descriptive summaries or entity-centred knowledge graphs. Effective Ebola response depends not only on available control measures but also on activating them at appropriate transition points. This framework complements, rather than replaces, WHO guidance, real-time surveillance, and formal epidemiological causal inference.