arXiv · March 2026 · Preprint

PLACID

Privacy-preserving Large language models for Acronym Clinical Inference and Disambiguation

LLM Clinical NLP Privacy-Preserving AI EHR Acronym Disambiguation

The Problem

Clinical notes are riddled with ambiguous acronyms.

Electronic health records contain tens of thousands of acronyms. "PT" could mean patient, physical therapy, prothrombin time, or posterior tibial — depending on context. Misinterpretation contributes to clinical errors and hinders downstream NLP tasks.

Existing solutions require sending sensitive patient data to external APIs — a non-starter in healthcare. PLACID addresses both problems at once.

Acronym Ambiguity Example

PT
Patient
82%
Physical Therapy
10%
Prothrombin Time
6%
Posterior Tibial
2%

Distribution depends entirely on clinical context

The Approach

Privacy-first, from the ground up.

Local Inference

Runs entirely on-premises. Patient data never leaves the institution. No external API calls, no data sharing.

Context-Aware LLM

Uses surrounding clinical context — the full sentence, section headers, patient history — to infer the most likely acronym expansion.

EHR-Scale

Designed to operate at scale across millions of clinical notes, improving downstream NLP tasks including coding, summarisation, and QA.

Performance

Accurate. Fast. Private.

Disambiguation Accuracy by Acronym Frequency

High-frequency acronyms (top 100) 0%
Medium-frequency acronyms 0%
Low-frequency / rare acronyms 0%

* Figures are representative — see arXiv paper for full experimental details.

0

Acronyms in scope

0

% Local inference

0ms

External data transfer

0

Team members

Team

The people behind PLACID.

Manjushree B. Aithal

Manjushree B. Aithal

Lead Author · PhD

Privacy-preserving NLP and clinical language models

Alexander Kotz

Alexander Kotz

Co-author · CPBS PhD Student

Computational Bioscience · CU Anschutz

James Mitchell

James Mitchell

Senior Author · PI

Department of Biomedical Informatics · CU Anschutz

arXiv:2603.23678 · March 2026

Read the full paper.

Available now as a preprint on arXiv. Peer review in progress.

View on arXiv →