Open-source large language model-based on-premises pipeline for automated data extraction from unstructured electronic health records: a pilot study - Takeaways - MDSpire

On-Premises Pipeline Utilizing Open-Source Large Language Models for Automated Extraction of Data from Unstructured Electronic Health Records: A Pilot Investigation

  • By

  • Vasileios Ntinopoulos

  • Hector Rodriguez Cetina Biefer

  • Laura Rings

  • Rodney Alexander Rosalia

  • Omer Dzemali

  • June 1, 2026

Share

  • 1

    The study evaluated an on-premises, open-source LLM-based data extraction pipeline for unstructured electronic health records.

  • 2

    Fifty medical texts in German were processed using 14 mid-sized LLMs, assessing their performance in various classification tasks.

  • 3

    Qwen3-30b-a3b-q8 achieved the highest overall accuracy of 0.954, with 13 LLMs exceeding 0.90 in overall accuracy.

  • 4

    Twelve LLMs demonstrated perfect accuracy in information extraction, while nine LLMs showed perfect response consistency.

  • 5

    The pilot study indicates the feasibility of LLM-based automated EHR data extraction, warranting further validation in healthcare.

Original Source(s)

Open-source large language model-based on-premises pipeline for automated data extraction from unstructured electronic health records: a pilot study

  • by Vasileios Ntinopoulos, Hector Rodriguez Cetina Biefer, Laura Rings, Rodney Alexander Rosalia, Omer Dzemali

  • June 1, 2026

Related Content

Prevention of Surgical Site Contamination With Cutibacterium acnes During Cardiac Surgery: A Randomized

First HIV-Positive-to-HIV-Positive Lung Transplant Reported

The procedure was performed under a HOPE Act research protocol at an NYU Langone Health center the institution said is among the limited number of US transplant centers equipped and approved to perform HOPE lung transplants.

Delayed splenic rupture following video-assisted thoracoscopic left upper lobectomy: a case report and literature review