Clinical Intelligence Agents for ACA Coding
ACA ChartCopilot™ deploys AI pipelines powered by NLP and fine-tuned LLMs trained on CMS coding guidelines, HHS-HCC model logic, and EDGE server supplemental diagnosis rules. Intelligent agents scan structured and unstructured clinical data, identify conditions that map to HHS-HCC categories, validate supporting evidence against CMS documentation standards, and generate compliant supplemental diagnosis records — with full audit trails linking every code to its source documentation.
Why Supplemental Diagnoses Matter in ACA Risk Adjustment
Under the ACA risk adjustment program, issuers can submit supplemental diagnosis codes derived from medical records and EDI sources through the EDGE server. These supplemental diagnoses capture conditions that may not appear on adjudicated claims but are clinically documented and supported by evidence in the patient's chart.
The impact is significant: supplemental diagnoses can reveal HHS-HCCs that drive higher individual risk scores and improve plan-level actuarial risk — directly affecting whether a plan receives or pays transfer funds. But CMS requires that every supplemental diagnosis be derived from an acceptable source, properly coded, and traceable to supporting documentation.
Most issuers leave substantial risk score value on the table because chart review for ACA populations is resource-intensive, coding teams are focused on MA programs, and the supplemental diagnosis file format has its own validation rules that differ from encounter submissions. ACA ChartCopilot™ solves all three problems.
Who This Is For
ACA ChartCopilot™ is designed for:
- ✓ QHP issuers seeking to maximize HHS-HCC capture from clinical documentation
- ✓ Coding and clinical documentation improvement (CDI) teams managing ACA chart reviews
- ✓ Risk adjustment teams responsible for supplemental diagnosis file submissions to the EDGE server
- ✓ TPAs and delegated coding partners supporting multiple ACA issuers
- ✓ Provider groups participating in value-based arrangements tied to ACA risk scores
- ✓ Compliance teams ensuring chart-derived codes meet CMS acceptable source requirements
The Challenges We Solve
Undocumented HHS-HCCs
Claims data alone misses conditions documented in progress notes, discharge summaries, and problem lists. Without chart review, these HCCs never reach the EDGE server and never impact risk scores.
ACA-Specific Coding Models
The HHS-HCC model differs from CMS-HCC (used in MA). Different condition categories, different hierarchies, different interaction terms. Coding teams trained on MA models miss ACA-specific opportunities.
Supplemental Dx File Compliance
CMS mandates specific acceptable sources for supplemental diagnoses: medical records and EDI. Health assessments have their own guidance. File format, void/add sequencing, and duplicate rules add operational complexity.
Resource Constraints
ACA populations are large and diverse. Manual chart review at scale requires significant coder capacity. Most organizations prioritize MA reviews and underinvest in ACA chart coding.
Evidence Traceability
Every supplemental diagnosis must be defensible in an HHS-RADV audit. Without systematic evidence linking, issuers risk having codes overturned during audit review.
Chronic Condition Recapture
Chronic conditions must be documented annually to maintain their HCC contribution. Without systematic recapture programs, prior-year HCCs drop off and risk scores decline.
How ACA ChartCopilot™ Works
Target High-Impact Members
ACA RiskLens™ identifies members with suspected undocumented HHS-HCCs based on claims patterns, prescription history, and prior-year diagnoses. ACA ChartCopilot™ receives a prioritized worklist ranked by revenue impact, enabling your coding team to focus where it matters most.
Ingest Clinical Documentation
Pull structured and unstructured clinical data from EHRs, HIEs, provider portals, and document management systems. ACA ChartCopilot™ processes progress notes, discharge summaries, operative reports, lab results, problem lists, and specialist consultations.
AI-Powered Chart Analysis
NLP and fine-tuned LLMs scan clinical text to identify conditions that map to HHS-HCC categories. The AI highlights relevant clinical passages, extracts ICD-10-CM codes with maximum specificity, and evaluates whether documentation meets CMS evidence standards for supplemental diagnosis submission.
Human-in-the-Loop Review
Certified coders review AI suggestions in a side-by-side viewer: source documentation on one side, proposed codes with evidence highlights on the other. Coders accept, modify, or reject each suggestion. Every decision is logged with rationale for audit defense.
Generate Supplemental Diagnosis Records
Approved codes are packaged into EDGE-compliant supplemental diagnosis file records with proper source type identification (medical record or EDI), member linkage, and void/add sequencing. Records feed directly into EDGESync™ for validation and submission.
Track Impact & Close the Loop
Monitor which supplemental diagnoses were accepted by the EDGE server, which mapped to HHS-HCCs, and the resulting risk score impact. Feed outcomes back into targeting algorithms to improve future chart review prioritization.
Key Capabilities
HHS-HCC Code Mapping
ICD-10-CM to HHS-HCC mapping engine covering Adult, Child, and Infant models. Applies hierarchical exclusion rules, disease interactions, and age/sex applicability — specific to the ACA risk adjustment model, not Medicare Advantage.
Clinical NLP Engine
Extracts conditions from unstructured clinical text with context-aware understanding. Distinguishes active diagnoses from historical mentions, ruled-out conditions, and family history. Handles negation, uncertainty, and temporal references.
Evidence Linker
Every proposed code is linked to specific passages in the source documentation. Evidence packages include page numbers, section headers, relevant clinical phrases, and documentation type classification — ready for audit defense.
Coder Workbench
Side-by-side review interface with source document viewer, AI-highlighted evidence, proposed ICD-10 codes, HHS-HCC mapping, and revenue impact estimates. Supports bulk review, quality scoring, and inter-rater reliability tracking.
Supplemental Dx File Builder
Generates EDGE-compliant supplemental diagnosis records with proper source type codes, member identifiers, and processing metadata. Handles add/void sequencing and duplicate detection per CMS business rules.
Prospective Suspecting
Identify chronic conditions due for annual recapture. Generate provider-facing alerts and care gap notifications to support prospective documentation improvement at the point of care.
Provider Feedback Reports
Automated reports showing providers their documentation completeness rates, missed coding opportunities, and improvement trends. Supports targeted provider education campaigns and documentation quality initiatives.
Quality & Compliance Monitoring
Track coding accuracy rates, inter-rater agreement, over-coding risk flags, and CMS-guideline adherence. Dual-control approval workflows ensure no code reaches submission without proper review.
CMS-Compliant Source Validation
CMS specifies acceptable sources for supplemental diagnoses submitted through the EDGE server. ACA ChartCopilot™ validates every record against these requirements:
| Acceptable Source | Description | ChartCopilot™ Handling |
|---|---|---|
| Medical Records | Progress notes, discharge summaries, operative reports, specialist consultations, problem lists, and other provider-authored clinical documentation | NLP extraction with evidence linking, MEAT validation, and source document classification |
| Electronic Data Interchange (EDI) | Diagnosis codes received through standard EDI transactions from providers or trading partners | Cross-reference with claims data, validate against enrollment records, and reconcile with existing EDGE submissions |
| Health Assessments | Diagnosis codes derived from health risk assessments (with CMS-specific guidance) | Flagged for additional compliance review per CMS guidance on health assessment-derived diagnoses |