The Evidence-First Adjudication Lifecycle
Every draft claim artifact is constructed through five strictly bounded stages, ensuring that probabilistic models never dictate final billing truth.
Domain, Specialty & Payer Scoping
Prior to linguistic analysis, the system retrieves relevant clinical coding criteria, specialty guidelines, and applicable payer policies. This establishes a bounded context for evidence extraction rather than querying an unconstrained model.
- Retrieval of applicable CMS NCCI edit tables for the provider specialty.
- Mapping of relevant Medicare Local Coverage Determinations (LCDs).
- Loading Medically Unlikely Edit (MUE) practitioner caps for candidate code families.
Clinical Fact Parsing & Assertion Verification
The AI model is strictly confined to extracting verifiable clinical facts from unstructured documentation (SOAP notes, H&P charts, operative reports). The model does not propose billing codes at this stage; it extracts structured observations.
- Assertion Status: Distinguishes confirmed conditions from historical, family, or ruled-out diagnoses.
- Anatomical Specificity: Extracts laterality (right, left, bilateral) and specific anatomical structures.
- Encounter Parameters: Captures face-to-face time, medical decision-making (MDM) complexity markers, and procedure details.
Code Candidate Proposal via Authoritative Sets
Extracted clinical facts are mapped to candidate standard codes (CPT, HCPCS, ICD-10-CM) using verified nomenclature. Every candidate code must point directly to documented supporting evidence in the clinical text.
- Zero hallucinated codes: candidates must exist in authoritative code sets.
- Diagnosis pointers: primary and secondary ICD-10-CM codes mapped to procedural indications.
- Uncertainty surfacing: ambiguous documentation triggers structured clarification requests.
NCCI, MUE & Coverage Rule Execution
All candidate codes enter a deterministic evaluation engine written in compiled code with zero LLM involvement. Rules are executed against official CMS NCCI edits, MUE tables, and coverage criteria.
- NCCI PTP Conflict Resolution: Identifies Column 1 / Column 2 edits and evaluates correct coding modifier indicators (CCMI).
- Modifier Adjudication: Evaluates factual justification for Modifier 25 (separate E/M), Modifier 59 / X{EPSU}, and Modifier 50.
- MUE Cap Enforcement: Restricts billed units to published practitioner and outpatient hospital limits.
CMS-1500 & ANSI EDI 837 Formatting for Evaluation
Once adjudicated, the platform outputs standardized claim artifacts formatted for testing, workflow integration, and pre-clearinghouse review.
- Draft CMS-1500 PDF with box-by-box clinical field mapping.
- Draft ANSI ASC X12 EDI 837P transaction files structured per standard 5010 loops.
- Detailed validation summary citing each applied rule and evidence link.
Source Scope & Legal Disclaimer
Source Scope: The deterministic adjudication engine references published CMS National Correct Coding Initiative (NCCI) Policy Manuals, CMS Medically Unlikely Edit (MUE) datasets, Medicare National and Local Coverage Determinations (NCD/LCD), and Medicare Claims Processing Manual (CMS Pub. 100-04).
Limitations: HealthyClaim is a research platform in public preview. Artifacts generated by the platform are drafts intended for evaluation, simulation, and workflow testing. The platform does not provide legal, medical, or payer-specific billing advice. Final coding and billing determinations remain the legal responsibility of certified healthcare providers and medical coders.