A Modular Multi-Agent Framework for Clinical Documentation and Hospital Operations

Clinical onboarding, documentation, and discharge workflows are still fragmented and labor-intensive, restraining efficiency and consistency in hospital workflows. Here, we introduce a modular, memory-extended multi-agent architecture to automate and manage these workflows through three specialist agents. The Onboarding Agent combines demographic and historical information to pre-populate intake documents and uses an ensemble of ARIMA, LSTM, and XGBoost models for wait-time estimation (MAE = 14.9 min), boosting timeliness in operations. The Clinical Notes Agent, based on transformer-based language architectures (ClinicalBERT), conducts biomedical named entity recognition and structured SOAP note composition, reaching F1-scores of 0.66 (TREATMENT), 0.62 (PROBLEM), and 0.55 (TEST), outclassing rule-based baselines. The Discharge Agent generates treatment summaries and follow-up intentions, ensuring accuracy and completeness of patient instructions. Coordinated through a shared memory substrate, agentic orchestration reduces normalized edit distance to reference notes to 0.250 (from 0.385) (Δ = 0.135; p = 0.0002) and boosts discharge accuracy by 10.5 percentage points (p = 0.0006). Tested offline on public datasets, the architecture shows that context-aware, interoperable agentic AI can efficiently diminish documentation workload, boost information fidelity, and enable enhanced workflow coordination under clinician oversight, and provides a reproducible platform compliant with risk protocols for large-scale, safe, and data-driven healthcare operations.

• Publication year

  2025

• Venue

  International Conference on Advances in Computing and Artificial Intelligence

• Publication date

  2025-12-26

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/ACAI68217.2025.11406271
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Publicly Available Clinical BERT Embeddings

  2019cited by this paper
• Medical error—the third leading cause of death in the US

  2016cited by this paper
• Changes in Burnout and Satisfaction With Work-Life Balance in Physicians and the General US Working Population Between 2011 and 2014.

  2015cited by this paper
• THE WORLD HEALTH ORGANIZATION

  1954cited by this paper

• No citing papers are available for this paper.