denial-shield

name

denial-shield

description

Adversarial Revenue Intelligence Engine for US healthcare. Treats hospital-payer relationship as an adversarial game — profiles payer behavioral 'genomes', predicts denial attack vectors, detects underpayments, optimizes DRG, auto-generates appeals with legal citations, and recommends optimal submission timing. Use whenever the user mentions: denial management, revenue cycle, claim denial, prior authorization, CPT/ICD coding, payer behavior, appeal letter, underpayment, DRG optimization, CDI clinical documentation, medical necessity, CARC/RARC codes, timely filing, modifier 25, CMS LCD NCD, Medicare Advantage denial, revenue leakage, charge capture, or any US healthcare billing/reimbursement topic. Also use for competitive analysis of US healthtech RCM companies.

Denial Shield — Adversarial Revenue Intelligence Engine

Paradigm Shift

Traditional denial management is reactive: claim denied → investigate → appeal.

Denial Shield is adversarial: profile the opponent → predict their moves → prevent denials before submission → detect underpayments → weaponize appeals.

Think of it like chess against the payer. The Payer Genome tells you their opening book.

8 Modules

#	Module	What It Does	Category
1	RULES	CPT/ICD-10 compatibility from external YAML	Denial Prevention
2	GENOME	Payer behavioral profiling & denial prediction	Adversarial Intel
3	CDI	Clinical documentation integrity analysis	Denial Prevention
4	MISS	Missing charges + underpayment detection	Revenue Recovery
5	DRG	DRG optimization through accurate coding	Revenue Recovery
6	RISK	Predictive denial scoring (ML-ready features)	Prioritization
7	APPEAL	Auto-generate appeals with legal citations	Denial Recovery
8	TEMPO	Optimal submission timing strategy	Denial Prevention

Quick Start

cp -r /mnt/skills/user/denial-shield/scripts/* /home/claude/
cp -r /mnt/skills/user/denial-shield/rules/* /home/claude/

# Install PyYAML for external rules
pip install pyyaml --break-system-packages

# Demo with realistic US hospital claim
python /home/claude/denial_shield.py demo

# Payer genome analysis
python /home/claude/denial_shield.py genome --payer united_healthcare

# Clinical documentation analysis
python /home/claude/denial_shield.py cdi --text "55yo male with morbid obesity..."

# Appeal strategy for CO-50 denial
python /home/claude/denial_shield.py appeal --denial-code CO-50 --input claim.json

# Submission timing
python /home/claude/denial_shield.py timing

Key Innovation: Payer Genome

Each payer has a behavioral "genome" — a fingerprint of:

• Baseline denial rate
• Top denial codes and frequencies
• High-risk CPT families
• Prior auth strictness score
• Timely filing deadline
• Temporal patterns (quarter-end spikes, Monday surges)
• Contract tactics (silent amendments, bundling without notice)
• Effective counter-strategies (peer-to-peer success rates, regulatory leverage)

The genome turns denial management from guessing to game theory.

Key Innovation: Underpayment Detection

Most hospitals only track denials (paid $0). But payers also underpay — paying 8-15% below contracted rates on individual line items. This silent revenue loss often exceeds denial losses.

The MISS module compares allowed_amount vs paid_amount per line to detect contract breaches.

Key Innovation: Appeal Weaponization

Appeals succeed based on specificity. The APPEAL module generates targeted strategies citing:

• CMS NCDs/LCDs
• OIG reports on payer behavior (OIG-22-06-11)
• CMS Interoperability Rule (CMS-0057-F)
• No Surprises Act provisions
• ERISA full-and-fair-review requirements
• State prompt payment laws

Plus payer-specific counter-strategies from the genome.

Key Innovation: Temporal Strategy

Denial rates follow temporal patterns:

• Monday surge (+30% prior auth denials)
• Quarter-end spike (+15-25% denial rate)
• Q4 audit season (RAC targeting)
• Policy lag (30-60 day propagation delay)

TEMPO module recommends when to submit, when to hold, and when to escalate.

External Rules (YAML)

All rules live in rules/denial_rules.yaml for hot-reload:

• CPT/ICD-10 incompatibility rules
• Payer genomes
• DRG optimization triggers
• Documentation requirements
• CARC/RARC denial code intelligence
• Temporal patterns

Hospital billing teams can update rules without touching code.

Payers Profiled

6 payer genomes with behavioral fingerprints:

• United Healthcare, Anthem BCBS, Aetna, Cigna
• Medicare Traditional, Medicare Advantage

Output Format

{
  "risk_score": { "risk_score": 0.68, "risk_level": "HIGH" },
  "payer_genome": { "attack_vectors": [...], "behavioral_warnings": [...] },
  "financial_summary": {
    "total_charges": 85000,
    "value_at_risk": 57800,
    "missed_revenue": 1250,
    "underpayment_detected": 3600,
    "total_financial_opportunity": 62650
  },
  "appeal_readiness": { "regulatory_citations": [...], "escalation_path": [...] },
  "timing_strategy": { "recommendation": "REVIEW TIMING" }
}

References

• references/us_revenue_cycle.md — US healthcare billing primer, regulatory landscape
• rules/denial_rules.yaml — All external rules (hot-reloadable)

v3.0 Modules: Contract Compiler + Claim Digital Twin

Contract Compiler (scripts/contract_compiler.py)

The missing foundation — converts payer contracts into executable pricing logic.

# Price a CPT code against compiled contract
python /home/claude/contract_compiler.py price --cpt 43775 --charge 45000

# Price inpatient DRG
python /home/claude/contract_compiler.py drg --drg 619 --charge 58750 --los 7

# Full variance analysis (expected vs paid)
python /home/claude/contract_compiler.py demo

Components:

• PricingEngine — calculates expected reimbursement per contract rules
• VarianceEngine — classifies every dollar of difference by root cause (15 variance types)
• RecoveryPrioritizer — ranks recovery opportunities by net expected value minus cost to pursue

Supports: fee schedule, % of Medicare, DRG-based, per diem (tiered), case rate, cost-plus, carve-outs, implant markup with caps, modifier adjustments, stop-loss, outlier, lesser-of, annual escalators.

Claim Digital Twin (scripts/claim_twin.py)

Each claim has a "twin" with 4 layers:

1. Clinical — documentation quality, DRG opportunities, medical necessity score
2. Regulatory — auth status, CCI compliance, CMS coverage, guideline defensibility
3. Economic — expected vs optimized reimbursement, carve-outs, stop-loss
4. Behavioral — payer denial probability, attack vectors, appeal success rates

# Full twin analysis with strategy recommendation
python /home/claude/claim_twin.py demo

# Monte Carlo simulation across all strategies
python /home/claude/claim_twin.py simulate

The twin generates 3-5 submission strategies and recommends the optimal play:

• Submit as-is (baseline)
• Strengthen documentation first
• Optimize coding (DRG shift via CDI query)
• Pre-emptive peer-to-peer
• Full optimization package

Monte Carlo simulation (1000+ runs) produces probability distributions of payment outcomes for each strategy.

Roadmap

• v3.1: 835/837 EDI parser for automated claim/remittance ingestion
• v3.2: FHIR R4 integration for EHR clinical data
• v3.3: Network Genome — cross-institutional payer behavior intelligence
• v4.0: ML model trained on historical denial data (XGBoost/LightGBM)
• v4.1: Automated appeal letter generation with local LLM
• v5.0: Multi-hospital Revenue Intelligence Platform (SaaS)