LLM pipeline struggles with FHIR R4 validation, highlighting data integrity issues

By PulseAugur Editorial · [1 sources] · 2026-08-03 22:33

A critical care physician developed OMAXI, an LLM pipeline designed to convert Spanish clinical documents into structured data without inferring information. During an experiment to map OMAXI's output to FHIR R4 standards and validate it against a Medplum server, two of 38 resources were rejected. One rejection involved a norepinephrine dosage due to missing patient weight, highlighting the system's adherence to data integrity over simple validation. The process revealed that the schema itself could discard data the LLM had already processed, and errors in terminology assignment and unit conversion were also identified. AI

IMPACT Highlights challenges in applying LLMs to structured clinical data and the importance of data integrity over simple format validation.

RANK_REASON The item describes the development and testing of an LLM pipeline for clinical data, which is a tool-level development.

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

LLM pipeline struggles with FHIR R4 validation, highlighting data integrity issues

COVERAGE [1]

dev.to — LLM tag TIER_1 English(EN) · Jesus Pantoja · 2026-08-03 22:33

The server rejected the norepinephrine — and that was the best thing that happened

I'm a critical care physician in Colombia. For months I've been building OMAXI, an LLM pipeline that reads Spanish clinical documents — nursing records, resident notes, lab reports — and turns them into structured data plus a note a physician can sign. …

COVERAGE [1]

The server rejected the norepinephrine — and that was the best thing that happened

RELATED TOPICS