Geriatric syndromes extraction from discharge summaries: a new dataset, annotation scheme and initial findings - Takeaways - MDSpire

Geriatric syndromes extraction from discharge summaries: a new dataset, annotation scheme and initial findings

  • By

  • Imane Guellil

  • Salomé Andres

  • Atul Anand

  • Bruce Guthrie

  • Fahrurrozi Rahman

  • Abul Hasan

  • Huayu Zhang

  • Honghan Wu

  • Beatrice Alex

  • July 13, 2026

Share

  • 1

    Geriatric syndromes (GS) significantly impact older adults' quality of life and are often poorly represented in structured electronic health records.

  • 2

    A manually annotated corpus of 2,040 discharge summaries was developed to enhance the detection and classification of GS using natural language processing.

  • 3

    BERT-cased achieved the highest F1-score of 0.897 for document-level labeling, while BioClinicalBERT excelled in negation detection with an F1-score of 0.888.

  • 4

    Common GS like frailty, falls, and delirium showed the best extraction performance, while rarer categories and complex annotations posed challenges.

  • 5

    The study highlights that data distribution, annotation complexity, and mention structure directly influence the performance of NLP models in GS extraction.

Original Source(s)

Geriatric syndromes extraction from discharge summaries: a new dataset, annotation scheme and initial findings

  • by Imane Guellil, Salomé Andres, Atul Anand, Bruce Guthrie, Fahrurrozi Rahman, Abul Hasan, Huayu Zhang, Honghan Wu, Beatrice Alex

  • July 13, 2026

Related Content

New CDC Report Tracks US Mortality Trends

The annual provisional report summarizes changes in mortality rates, demographic patterns, and leading causes of death during 2025.

Individualized osilodrostat treatment for patients with ACTH-dependent Cushing’s syndrome: real-world evidence

Correction: Unraveling the transcriptome-based network of Tfh cells in primary Sjogren syndrome: insights from a systems biology approach