Implementation of large language models in electronic health records
6/10 importance · ai assisted · clinical documentation · 2025 · validator confidence 80% · deployed
The problem
Electronic health records have increased clinicians' documentation workload, and real-world LLM deployments in live hospital systems remain limited.
Why it matters
Documentation burden takes clinicians' time away from patient care in hospitals.
What the AI did
A secure, GDPR-compliant, on-premises LLM assistant with retrieval augmented generation was integrated into the Epic EHR to give context-aware answers from patient data, clinical documents, and medical literature.
Results
One-month pilot with 28 physicians across nine specialties: 64% used it daily, 482 multi-turn conversations; summarization, information retrieval, and note drafting were over 70% of interactions; hospital-wide deployment followed.
Impact
The pilot showed high physician engagement and the system was subsequently deployed hospital-wide.
How important is this problem?
6/10 — significant — thousands to millions affected
Documentation burden takes clinician time from patient care; abstract quantifies deployment (1,028 users, 14,910 conversations) but not problem scale. Hospital-wide deployment shows real, significant benefit at one hospital.
inferred — an explicit judgement from the source text.
Scale stated in source: not stated in source
Ratings use a published rubric and are AI-assisted, not measured or peer-reviewed. Method
Sources & verification
- Primary source
- https://doi.org/10.1371/journal.pdig.0001141
- Model / method
- Qwen3-235B with Retrieval Augmented Generation
- Organization
- not stated in source
- Model openness
- open-source model
- Validator notes
- glm-5.3: ok (confidence 80) | Real deployment with engagement metrics; documentation-burden reduction itself not measured.
- Keywords
- LLM, EHR, RAG, clinical documentation, hospital deployment, Qwen3
A second-stage review judged this entry against the primary source; status “verified”. “Verified” means the evidence supported the claim — it does not mean the underlying result was independently reproduced.
Related
-
Multimodal deep learning of fundus abnormalities and traditional risk factors for cardiovascular risk prediction
Cardiovascular disease, the leading global cause of death, needed better risk identification from accessible data.
Medicine & Health · 2023 · ai assisted · 10/10 importance
-
Reliable Tuberculosis Detection Using Chest X-Ray With Deep Learning, Segmentation and Visualization
Detecting tuberculosis accurately and early from chest X-ray images, since late diagnosis can be life-threatening.
Medicine & Health · 2020 · ai assisted · 10/10 importance
-
A comparative analysis of machine learning classifiers for stroke prediction: A predictive analytics approach
Accurate early prediction of stroke is needed to enable timely treatment, but medical data is imbalanced.
Medicine & Health · 2022 · partially solved · 10/10 importance
-
Cardiovascular Disease Diagnosis from DXA Scan and Retinal Images Using Deep Learning
Cardiovascular disease often goes undiagnosed until serious events; non-invasive diagnosis from retinal images and DXA scans was lacking in Qatar.
Medicine & Health · 2022 · partially solved · 10/10 importance
-
Proteomics and Machine Learning Approaches Reveal a Set of Prognostic Markers for COVID-19 Severity With Drug Repurposing Potential
Predicting and managing COVID-19 disease severity was difficult, and mechanisms underlying severity were unclear.
Medicine & Health · 2021 · partially solved · 10/10 importance