Case Study
Clinical Document Intelligence
The Situation
A regional health plan needed to identify patients matching complex risk criteria — across thousands of clinical documents including PDFs, scanned records, and handwritten clinical notes. The volume was too large for manual review. Traditional keyword search broke down against clinical abbreviations, inconsistent provider terminology, and unstructured data that standard analytics workflows couldn’t read.
The downstream impact was real: HCC gaps were being missed, RAF scores weren’t reflecting true member acuity, and clinical teams had no scalable way to surface the evidence they needed.
Combined OCR, clinical NLP, vector search, and structured filters into a single end-to-end system — capable of reading, understanding, and searching clinical documents the way a trained coder would, but at machine speed.
Applied domain-specific embeddings to normalize clinical entities across conditions, medications, and procedures — so the system understood that “DM2,” “Type 2 diabetes,” and “insulin-dependent diabetes” all mean the same thing.
Built a production-grade ingestion, indexing, and retry engine with full audit trail — exposing results through both a REST API and a web-based analytics dashboard accessible to clinical and coding teams.
Talk to an Onyx expert about applying clinical document intelligence to your risk adjustment or care gap program.
