Clinical plausibility before payment, not after.
CMS estimates over $100B in improper Medicare and Medicaid payments annually. Most fraud detection is reactive — post-payment recovery after the money is gone. Skippy Integrity screens claims against clinical evidence before adjudication: drug-indication matching, dose validation, and diversion pattern detection — with PAY / FLAG / HOLD / REFER_MFCU recommendations and an immutable audit trail per claim.
No clinical evidence supports oxycodone for allergic rhinitis (J30.9)
What a response looks like
Claim plausibility scoring, Holy Trinity diversion detection, and batch pre-payment screening — three capabilities, one API.
No clinical evidence supports oxycodone for allergic rhinitis (J30.9)
Opioids are not indicated for allergic conditions per FDA label or major guidelines
Simulated output representative of real API responses. Clinical indication matching, dose validation, and diversion pattern detection. Immutable SHA256 audit trail per claim scored.
Seven capabilities. One API.
Claim plausibility scoring
Screen any drug claim for clinical plausibility before payment. Clinical indication matching returns PLAUSIBLE / IMPLAUSIBLE / UNCERTAIN with a fraud risk score (0–10) and recommendation: PAY, FLAG, HOLD_FOR_REVIEW, or REFER_MFCU. Every decision generates an immutable audit record.
Regimen fraud screening
Screen a patient's full medication regimen for multi-drug diversion patterns. Detects Holy Trinity (opioid + benzodiazepine + carisoprodol), opioid-benzo combinations, and high-dose single-opioid patterns — with per-pattern risk score increments and an overall risk classification.
Batch claims screening
Submit up to 10,000 claims in a single request for concurrent pre-payment screening. Returns per-claim verdicts alongside summary counts (approved / flagged / held / referred) and total processing time. Designed for MMIS claim file integration.
Dose plausibility validation
Verify submitted quantity and days-supply against clinical maximum daily dose references for 14 controlled substances: 8 opioids (oxycodone 120 mg/day, hydrocodone 90 mg/day, fentanyl 200 mcg/hr, morphine 200 mg/day, methadone 120 mg/day, buprenorphine 24 mg/day, codeine 360 mg/day, tramadol 400 mg/day), 4 benzodiazepines (alprazolam 4 mg/day, diazepam 40 mg/day, lorazepam 10 mg/day, clonazepam 20 mg/day), and carisoprodol 1400 mg/day. Returns daily dose calculated, reference maximum, and DOSE_IMPLAUSIBLE flag when exceeded.
Diversion patterns catalog
Retrieve the full library of diversion pattern definitions — drug class combinations, risk score increments, severity, and clinical rationale. Used by the regimen screen and can be queried directly for integration with existing FWA platforms.
Audit explanation generation
Generate plain-language explanations of fraud screening results scoped to three audiences: auditor (OIG federal auditor — precise, flag-citing, cites specific clinical evidence), investigator (state Medicaid investigator — actionable findings with referral package), or reviewer (compliance reviewer — summary with recommended next steps). Structured fallback if LLM unavailable — explanation never blocks the audit record.
Prescriber pattern analysis
Analyze a prescriber's controlled substance prescribing pattern for potential fraud indicators. Accepts prescriber NPI and state. Returns risk_flags, risk_score, and diagnostic detail. Requires prescriber database integration for full statistical outlier analysis — conservative fail-open behavior by default: returns no flags rather than generating false positives that could block legitimate payments.
Four verdicts. Risk-scored. Auditable.
Every claim receives a structured recommendation based on its fraud risk score (0–10). Thresholds are configurable per program; the fail-open design ensures legitimate claims are never blocked by infrastructure issues.
Drug is clinically supported for the stated diagnosis, dose is within range, and no diversion patterns detected. Risk score < 6.0.
Clinical plausibility concern identified — indication mismatch, dose question, or pattern proximity. Risk score 6.0–6.9. Claim may proceed but should be reviewed.
Significant clinical implausibility or flagged diversion pattern. Risk score 7.0–7.9. Payment held pending manual clinical review before adjudication.
Fraud indicators meet the threshold for referral to the Medicaid Fraud Control Unit. Risk score ≥ 8.0. Structured referral package generated with evidence citations and audit trail.
If clinical evidence services are unavailable, the system returns PAY with UNCERTAIN plausibility rather than blocking legitimate claims. Designed to never hold up valid payments due to infrastructure issues.
Every claim scored produces an audit record with a unique audit_id linked into an append-only SHA256 hash chain. Records are tamper-evident and retained for 7 years — OIG and FDA 21 CFR Part 11 compatible.
Known diversion signatures.
Pattern detection runs against a patient's full medication regimen during a time window — not individual claims in isolation. Each pattern match increments the risk score independently, and patterns compound.
The highest-known diversion risk combination — opioid sedation amplified by benzo and muscle relaxant. Strongly associated with diversion and overdose mortality.
Combined CNS depression is associated with respiratory depression risk. Concomitant prescribing requires documented clinical justification.
Daily dose calculated from quantity / days_supply exceeds evidence-based maximum for the specific opioid. Hardcoded clinical reference limits for 8 opioid drugs.
Seven structured flag types.
Every adverse finding is classified into one of seven typed flags — each with severity, confidence, and clinical evidence citation. Flags are the building blocks of the audit record and the basis for any denial rationale cited in OIG submissions.
Skippy Ground finds no clinical evidence supporting the submitted drug for the stated ICD-10 diagnosis. Fired when ground verdict is NOT_COVERED.
Clinical evidence explicitly contradicts prescribing this drug for the stated diagnosis — a positive contraindication, not merely absence of evidence. Fired when ground verdict is CONTRADICTED.
Daily dose derived from quantity ÷ days-supply exceeds the clinical reference maximum for the drug. Fired by the dose-plausibility engine against hardcoded controlled substance limits.
Patient regimen matches a named diversion signature — Holy Trinity (+4.0), Opioid-Benzo (+2.5), or High-Dose Opioid (+1.5). Patterns compound additively against the base risk score.
Drug class is clinically used, but not for the specific indication on the claim. Distinguishes general implausibility from a specific mismatch with documented alternatives.
Controlled substance prescribed by multiple prescribers within the screening window — a known diversion indicator for doctor-shopping. Requires prescriber NPI field in claim input.
Prescriber's controlled substance volume or distribution pattern is statistically anomalous relative to specialty peers. Powered by prescriber risk module (currently configurable stub awaiting database integration).
CMS recovers approximately 4% of identified improper payments. Once a fraudulent claim is paid, recovery is expensive, slow, and often incomplete. The only economically rational intervention is pre-payment screening.
Most FWA systems are statistical — they flag outliers compared to peer behavior. Skippy Integrity is clinical: does this drug have clinical evidence for this diagnosis? Is this dose consistent with the indication? Does this regimen match any known diversion pattern? Statistical and clinical detection catch different fraud — they should both run.
“Oxycodone prescribed for allergic rhinitis looks statistically normal — opioids are common, J30.9 is common — but it's clinically implausible. Statistical outlier detection won't catch it. Clinical indication matching will.”
Skippy Integrity adds the clinical plausibility layer that statistical platforms don't have — grounded in drug-indication evidence from FDA labels, clinical guidelines, and the Skippy evidence base. Denial rationale is evidence-cited, not algorithmic, which satisfies CMS-0057-F documentation requirements for AI-driven prior authorization decisions.
Program integrity teams and payers
OIG-compatible from first call
Every screening decision produces an OIG-compatible audit record: the specific clinical evidence evaluated, confidence score, sources consulted, and the full rationale chain. Records are cryptographically hash-chained in an append-only SQLite audit log and retained for 7 years — meeting FDA 21 CFR Part 11 and CMS audit documentation requirements.
Denial rationale is grounded in specific clinical evidence citations, not statistical flags. This satisfies CMS-0057-F requirements for AI-driven prior authorization decisions — every FLAG or HOLD is traceable to versioned clinical criteria a provider can review and appeal.
HIPAA-ready with BAA available. RxNorm and ICD-10 normalized inputs integrate directly with existing claims processing infrastructure and MMIS workflows.
Often deployed together
ICD-10 and CPT verification feeds upstream into fraud screening — accurate diagnosis codes are the input the plausibility engine evaluates. Coding errors and fraudulent coding both surface differently.
Clinical indication evidence is shared infrastructure — the same drug-diagnosis evidence base that powers CDS recommendations is what Integrity uses to evaluate claim plausibility.
Diversion pattern detection covers known dangerous combinations. DDI covers the broader interaction landscape — together they flag both the clinically dangerous and the clinically implausible.
Skippy Integrity is a decision-support tool for qualified program integrity, compliance, and claims review professionals. Screening verdicts are recommendations — all FLAG, HOLD_FOR_REVIEW, and REFER_MFCU decisions should be reviewed by qualified personnel before adverse action. Clinical plausibility screening does not substitute for clinical review, legal counsel, or investigative judgment in MFCU referral decisions. HIPAA-ready; BAA available.
See Integrity in your program integrity workflow
We work with MACs, Medicaid PIUs, PBMs, and health plan program integrity teams. Let's talk about your pre-payment screening problem.