Skip to main content
AAO-HNSF 2024 Annual Meeting & OTO EXPO Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

Late-Breaking Scientific Oral Presentations

Care Equity, Quality, and Safety

Session: Late-Breaking Scientific Orals: The Otolaryngology Community and Systems

A Novel Machine Learning Model for Triaging Head and Neck Patient Referrals

Monday, September 30, 2024
9:36 AM – 9:42 AM EDT
Location: MBCC, 222
Introduction: Inaccurate patient triage leads to poor clinic resource utilization and delays in patient care, which may significantly increase morbidity and mortality in cancer patients. We developed a natural language processing (NLP) model to streamline triage workflows for head and neck practice coordinators. This study assesses the model’s ability to categorize and triage patient referrals based on their referring medical records.

Methods: We performed a retrospective cohort study of 80 new patients at a tertiary head and neck cancer center between January and April 2024. Referral documents (e.g. outside clinic notes, imaging, pathology reports) were collected and entered into the NLP model which subsequently determined patient pathology type and severity level. Pathologies were divided into non-endocrine head and neck neoplasm, thyroid pathology, parathyroid pathology, and benign lesions. Disease severity was stratified into low (benign lesions/conditions), moderate (indeterminate lesions without a clear diagnosis), and high (high suspicion or biopsy-proven malignancy). The ground truths were attending physician’s clinic note and final pathology report.

Results: The NLP model exhibited high accuracy (83.8%) in predicting pathology from referring medical records. Sensitivity for non-endocrine head and neck neoplasm, thyroid pathology, parathyroid pathology, and benign lesions were 91.5%, 88.9%, 100%, and 65.2%, respectively. Specificity for non-endocrine head and neck neoplasm, thyroid pathology, parathyroid pathology, and benign lesions were 87.9%, 91.6%, 97.5%, and 98.2%, respectively. The model also showed high accuracy (88%) in assessing disease severity with a sensitivity of 96.6% and specificity of 66.7%.

Conclusions: We demonstrated a novel NLP model that predicts underlying pathology based on referring documents with high levels of specificity and sensitivity. The model especially excelled at identifying patients requiring urgent care based on disease severity. This tool may help head and neck practice coordinators screen referrals accurately, thereby optimizing patient care.

    Presenting Author(s)

  • Stefanie Seo, BA

    Medical Student
    The Johns Hopkins University School of Medicine, United States

    Disclosure information not submitted.

  • AD

    Andy S. Ding, MD, MSE

    Resident Physician
    The Johns Hopkins University School of Medicine
    Baltimore, MD, United States

    Disclosure information not submitted.

    • Co-Author(s)

  • AA

    Ameen Ahmad, BS

    Medical Student
    The Johns Hopkins University School of Medicine, United States

    Disclosure information not submitted.

  • KX

    Kevin Xin, MD

    Resident Physician
    University of California, Irvine, United States

    Disclosure information not submitted.

  • MJ

    Max Jiam, BS

    IIAM Corporation, United States

    Disclosure information not submitted.

    • Senior Author(s)

  • NJ

    Nicole T. Jiam, MD

    Otology, Neurotology & Skull Base Surgery
    Harvard Medical School & Massachusetts Eye and Ear
    Boston, Massachusetts, United States

    Disclosure information not submitted.

  • CF

    Carole Fakhry, MD, MPH

    Professor
    Johns Hopkins Medical Institute
    Baltimore, Maryland, United States

    Disclosure information not submitted.


Back to Top