emtelligent Medical Language Engine Improves LLM Coding Accuracy
Solution outperforms LLMs alone with greater clinical reliability and no hallucinations
MINNEAPOLIS, July 7, 2026 /PRNewswire/ -- emtelligent®, a leader in clinical-grade AI, today announced that testing shows that LLMs using its next-generation Medical Language Engine significantly outperform LLMs alone in medical coding.
The findings are based on a head-to-head evaluation using 272 real hospital discharge summaries and a rigorously validated dataset of nearly 79,000 labeled medical concepts. The study evaluated how leading frontier models mapped clinical text to standardized medical ontologies using SNOMED CT, one of the most comprehensive and granular clinical terminologies.
In head-to-head testing against the leading medical LLMs, emtelligent's clinical AI achieved an 89.85% F1 score with zero hallucinations and identified 4 of 4 relevant clinical concepts. The F1 score measures how accurately and completely a model identifies relevant information. By contrast, the best-performing LLM achieved only a 55% F1 score while fabricating nearly 1 in 3 medical codes.
Many healthcare organizations use retrieval-augmented generation (RAG) to reduce hallucinations, but the study found that RAG did not improve their performance. RAG-assisted LLMs reduced hallucinations but lowered accuracy to a 22.64% F1 score. Fewer than 1 in 4 relevant clinical concepts were identified.
"Understanding medical ontologies and terminology is beyond the ability of general-purpose LLMs," said Tim O'Connell, M.D., CEO and co-founder of emtelligent. "And since LLMs don't answer 'I don't know' to queries, they can fabricate answers without checking them against accepted medical ontologies. Those coding hallucinations put patients, revenue, and compliance at risk."
Clinical coding is foundational to nearly every downstream healthcare function, from clinical decision-making and risk modeling to reimbursement, quality, compliance, and real-world evidence generation. When clinical language is captured incorrectly, missed entirely, or mapped to invalid codes, those errors can compound across millions of records and create operational, financial, and regulatory risk.
The study shows that general-purpose LLMs are not reliable enough to manage the coding layer on their own. They can generate medical codes, but they do not consistently produce outputs that are valid, precise, and supported by the source text.
emtelligent's Medical Language Engine addresses that gap by identifying clinical entities, coding them to accepted medical ontologies, and capturing the assertions and relations that give those entities context. This gives LLMs structured clinical data they can use to improve accuracy and reduce hallucinations.
emtelligent's platform includes three complementary capabilities:
- Document AI automates document intake and processing by splitting, classifying, extracting metadata, and applying OCR to bundled PDFs, faxes, TIFs, scanned records, and other complex medical documents.
- Medical Language Engine extracts and structures clinical meaning from text, including entity linking, coding, assertions, relations, and clinical data extraction across medical ontologies such as SNOMED, ICD-10, RxNorm, LOINC, and others.
- AI Chart Review provides a human-in-the-loop workflow interface for clinicians, coders, quality teams, and administrative users to search, summarize, review, and validate charts with full traceability to source documentation.
These capabilities can be deployed together as a complete end-to-end solution or used separately depending on the organization's needs. New Model Context Protocol (MCP) server functionality also allows healthcare organizations to integrate their own LLMs, adding accurate medical coding to the LLM context using emtelligent's extraction and medical language accuracy.
"Healthcare has been unable to fully trust AI in coding because of concerns about accuracy," O'Connell said. "Our Medical Language Engine eliminates those worries and provides organizations with a complete and scalable medical AI solution that can accelerate insights and drive immediate efficiency across their organization."
These study results point to a larger shift in healthcare AI: the next wave will be led by organizations that get the data layer right. General-purpose models can be powerful tools, but clinical-grade performance depends on purpose-built infrastructure that can deliver accurate, traceable clinical data at enterprise scale.
About emtelligent:
emtelligent® is dedicated to building AI-powered solutions that transform unstructured data into actionable insights at enterprise scale. Engineered by medical and data science experts, our AI-powered solutions help payers, pharma, health systems, and health tech innovators extract, structure, and integrate their clinical data with unmatched accuracy and speed.
Generic AI solutions fall short of healthcare's exacting standards. emtelligent's Medical Language Engine represents the most accurate and feature-rich medical AI on the market, helping healthcare organizations improve operational efficiency, reduce administrative burden, and enhance quality of care across the healthcare industry.
Leverage data like never before and transform your potential into performance. Learn more at emtelligent.com.
Contact:
Supreme Communications for emtelligent
SOURCE emtelligent
Share this article