MedDCR: Learning to Design Agentic Workflows for Medical Coding

2025-11-17

AI Summary

Manual workflow design in medical coding is rigid and fails to adapt to the inherent variability of real-world clinical text. Method: This paper proposes the first agent framework that models workflow construction as a learnable task. It introduces a Designer-Coder-Reflector closed-loop architecture, integrating large language model–driven multi-step reasoning, clinical guideline–constrained execution, cross-document consistency modeling, and a dynamic memory mechanism to enable automatic workflow evolution and optimization. Contribution/Results: It achieves the first end-to-end learnable, interpretable, and reusable coding workflow generation. Evaluated on multiple benchmark datasets, it significantly outperforms state-of-the-art methods, delivering higher coding accuracy, strong generalization across diverse clinical narratives, and enhanced decision transparency and reliability—establishing a novel paradigm for clinical natural language processing.

Technology Category

• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Multiagent Systems: Mechanism Design

Application Category

• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Medical coding converts free-text clinical notes into standardized diagnostic and procedural codes, which are essential for billing, hospital operations, and medical research. Unlike ordinary text classification, it requires multi-step reasoning: extracting diagnostic concepts, applying guideline constraints, mapping to hierarchical codebooks, and ensuring cross-document consistency. Recent advances leverage agentic LLMs, but most rely on rigid, manually crafted workflows that fail to capture the nuance and variability of real-world documentation, leaving open the question of how to systematically learn effective workflows. We present MedDCR, a closed-loop framework that treats workflow design as a learning problem. A Designer proposes workflows, a Coder executes them, and a Reflector evaluates predictions and provides constructive feedback, while a memory archive preserves prior designs for reuse and iterative refinement. On benchmark datasets, MedDCR outperforms state-of-the-art baselines and produces interpretable, adaptable workflows that better reflect real coding practice, improving both the reliability and trustworthiness of automated systems.

Problem

• Automating multi-step reasoning for medical coding from clinical notes
• Learning adaptable workflows instead of using rigid manual designs
• Improving reliability and interpretability of automated medical coding systems

Innovation

• Closed-loop framework for workflow design learning
• Designer-Coder-Reflector architecture with memory archive
• Generates interpretable adaptable workflows from feedback