D2MDT: Department-aware Multidisciplinary Team Consultation with Deliberation for Efficient Clinical Prediction
Researchers have developed D2MDT, a novel multi-agent system designed to enhance clinical prediction using electronic health records. This system structures EHR evidence and assigns department-specific perspectives to doctor agents, facilitating collaborative consultation. D2MDT improves efficiency by employing residual deliberation, which updates only unresolved consensus, and has demonstrated improved predictive performance in mortality prediction experiments. AI
IMPACT Introduces a more efficient multi-agent approach for clinical prediction, potentially improving diagnostic accuracy and resource allocation in healthcare.