D2MDT: Department-aware Multidisciplinary Team Consultation with Deliberation for Efficient Clinical Prediction

2026-06-02

AI Summary

This work addresses the limitations of existing clinical prediction methods in supporting multidisciplinary collaborative reasoning and the prevalent issues of low evidence discriminability and redundant interactions in multi-agent systems. The authors propose a department-aware multidisciplinary consultation framework that leverages structured electronic health record modeling and semantic evidence retrieval to assign specialty-specific perspectives to physician agents and extract complementary evidence. A residual deliberation mechanism is introduced to iteratively refine only those aspects lacking consensus, while integrating consensus reports with structured representations for efficient prediction. Evaluated on mortality prediction tasks, the approach significantly improves accuracy while substantially reducing redundant inter-agent communication, thereby achieving a favorable balance between predictive performance and consultation efficiency.

Abstract

Electronic health records (EHRs) are central to clinical prediction, but existing methods either rely on correlation-driven deep models or use single large language models (LLMs), making it difficult to support multidisciplinary clinical reasoning. Recent multi-agent systems (MAS) provide a promising alternative, yet current EHR-grounded MAS methods still suffer from weak evidence differentiation across agents and redundant multi-round interaction. We propose D2MDT, a Department-aware MultiDisciplinary Team Consultation with Deliberation for Efficient clinical prediction. D2MDT first constructs structured EHR evidence and consultation-ready semantic evidence for multi-agent consultation. It then assigns patient-specific department perspectives to doctor agents and retrieves complementary evidence for collaborative consultation. To improve efficiency, D2MDT further introduces residual deliberation, which updates only unresolved consensus rather than replaying the full discussion history. Finally, D2MDT fuses the refined consensus report with structured EHR representations for prediction. Experiments on mortality prediction show that D2MDT improves both predictive performance and consultation efficiency. We release the code online to ease the reproducibility of this paper.

Problem

• clinical prediction
• electronic health records
• multidisciplinary reasoning
• multi-agent systems
• evidence differentiation

Innovation

• multi-agent system
• clinical prediction
• electronic health records
• residual deliberation
• department-aware consultation