Skip to main content

How ANICA Works

From clinical document to audit-ready codes in 5 stages — powered by Jivica's multi-agent AI engine

How ANICA Works — 5-stage pipeline: Intake, Analysis, Coding, QA, Audit. Seconds to complete.
Stage 1

Upload

Clinical documents ingested — PDF, text, scanned images, or HL7

Stage 2

Extract

Parses clinical documents to identify and extract individual progress notes, sections, and provider attestation blocks.

Stage 3

Code

Analyzes clinical text and generates ICD-10-CM code suggestions with confidence scores and evidence citations.

Stage 4

Map & Score

Maps validated ICD-10 codes to HCC categories using V24 and V28 models. Applies hierarchy rules and calculates RAF score impact.

Stage 5

Validate

Final validation checking MEAT criteria, guideline compliance, and RADV readiness. Challenges false positives and recovers missed diagnoses.

Stage Details

Stage 1: Administrative Extraction

~2 seconds

Agent: Document Intake

Clinical documents ingested — PDF, text, scanned images, or HL7

Patient ID (redacted for PHI)Provider name and credentialsDate of serviceDocument typeEncounter type

Stage 2: Progress Notes Extraction

~5-10 seconds

Agent: Clinical Analysis

Parses clinical documents to identify and extract individual progress notes, sections, and provider attestation blocks.

Individual progress notesSection boundariesProvider attestation blocksSignature detection

Stage 3: Diagnosis Coding

~10-20 seconds

Agent: Code Assignment

Analyzes clinical text and generates ICD-10-CM code suggestions with confidence scores and evidence citations.

Candidate ICD-10 codesConfidence scoresSupporting evidence citationsSpecificity recommendations

Stage 4: HCC Mapping

~3-5 seconds

Agent: HCC Mapping

Maps validated ICD-10 codes to HCC categories using V24 and V28 models. Applies hierarchy rules and calculates RAF score impact.

HCC category assignmentsV24 and V28 mappingsHierarchy adjustmentsRAF score contribution

Stage 5: QA Review

~5-10 seconds

Agent: Quality Assurance

Final validation checking MEAT criteria, guideline compliance, and RADV readiness. Challenges false positives and recovers missed diagnoses.

MEAT compliance statusGuideline check resultsRADV risk scoreFinal recommendationsComplete audit trail

Powered by Neuro-Symbolic AI

Every stage of the pipeline combines advanced AI for clinical understanding with a deterministic rule engine that enforces CMS coding rules — zero hallucinations, guaranteed compliance.

Explore the architecture