AI Medical AuditorClaims audit & documentation review

Claims Audit Dashboard

The engine checks whether the documented diagnosis and treatment support the billed ICD-10 codes and procedures, and flags items for auditor review.

1. Provide claims data

Download claims template (.xlsx)
⇧
Drop the claims file here or browse
Excel (.xlsx) or CSV with columns: Claim ID, Patient ID, Patient Age, Patient Gender, Admission Date, Discharge Date, Clinical Notes, Billed ICD-10 Codes (semicolon-separated), Billed Procedures, Billed Amount (AED)
📄
Optional: clinical notes (.docx / .txt)
Linked to a claim by file name (e.g. CLM-2026-013.docx) or a "Claim ID: …" line in the document. Text is appended to that claim's notes.
Connect ERP… Sample dataset: 25 claims for evaluating the checks.

How it works

1
Extract
A clinical NER model (d4data/biomedical-ner-all) pulls diagnoses, symptoms, procedures and medications out of the free-text notes.
2
Match
Sentence embeddings (all-MiniLM-L6-v2) compare the documentation with the official description of every billed ICD-10 code and procedure.
3
Check
Rule-based checks add code validity, upcoding, duplicates, demographic consistency and billed-amount outliers.
4
Review
Every result is presented as flagged for auditor review with the exact note text or data that triggered it.