Healthcare claims fraud

Pay for care. Not for phantom visits.

Fraudulent providers bill for visits that never happened, quote prior authorizations issued for other patients and bill the same service twice. Omnareth checks every claim against the member's coverage and history before it is paid, and recognizes the provider when it bills across many members.

12 reference rules8/8 scenario decisions verified75 ms median decision, lab test

Healthcare claims decisions · reference scenariohealthcare-claims/demo.sh
#EventDecision
1 Genuine MRI with its prior authorization ALLOW
2 A clinic bills an office visit that never happened DENY SERVICE_NOT_VERIFIED
3 The clinic bills an MRI on another member's prior authorization
▲ Repeat offender case; clinic watched
DENY PRIOR_AUTH_INVALID · PRIOR_AUTH_REUSE
4 The clinic bills a therapy session already paid
▲ One provider across three members: case opened
DENY DUPLICATE_BILLING · WATCHLIST_MATCH · REPEAT_OFFENDER · PROVIDER_HOPPING
5 A genuine visit billed above what the diagnosis supports FLAG UPCODING_SUSPECTED
6 A procedure the plan does not cover DENY SERVICE_NOT_COVERED
– Analyst confirms the provider case
▲ Provider blocked everywhere
analyst · state
7 A claims processor resubmits the clinic's claim, visit marked verified DENY WATCHLIST_BLOCK · REPEAT_OFFENDER · PROVIDER_HOPPING
8 Genuine office visit ALLOW
8/8 decisions as expected · every record verified on chain

How it fits

A check before money moves

Healthcare fraud hides in volume: each claim looks routine, and the pattern only appears across members. Your system sends each request to Omnareth and acts only on what it allows.

The record

The member

Eligibility, the services the plan covers, the prior authorizations issued and the services already paid.

The events

Provider claims

A claim must be for an eligible member, a covered service, a verified encounter and an authorization issued for this member, once.

The memory

One provider, many members

Omnareth holds the billing provider responsible, never the members billed for, and recognizes it across members.

Reference rules

12 rules, ready to adapt

Authorization checks fail closed, and every rule names who it holds responsible when it fires.

ThreatWhat Omnareth checksRule codeDecision
Possible UpcodingFlag claims whose diagnosis does not support the level of service billed.UPCODING_SUSPECTEDFLAG
Duplicate BillingA service by a provider on a date is paid once.DUPLICATE_BILLINGDENY
Service Not VerifiedA billed service must match a verified encounter (eligibility check, check-in or visit record).SERVICE_NOT_VERIFIEDDENY
Member Not EligibleClaims are paid only for members with active coverage.MEMBER_NOT_ELIGIBLEDENY
Prior Authorization InvalidA prior authorization quoted on a claim must have been issued for this member.PRIOR_AUTH_INVALIDDENY
Excluded ProviderProviders on a government exclusion list may not be paid.PROVIDER_EXCLUDEDDENY
Service Not CoveredThe procedure must be covered by the member's plan.SERVICE_NOT_COVEREDDENY
Provider Across MembersFlag a provider or referrer implicated in denied claims for two or more other members in 24 hours.PROVIDER_HOPPINGFLAG
Prior Authorization ReuseA prior authorization belongs to one member; seen on another, it is likely copied.PRIOR_AUTH_REUSEFLAG
Repeat OffenderFlag events involving an identity implicated in two or more denied events in the last 30 days.REPEAT_OFFENDERFLAG
Blocked IdentityIdentities blocked on the fraud watchlist may not take part in any member event.WATCHLIST_BLOCKDENY
Watched IdentityIdentities under watch may proceed only with a flag for review.WATCHLIST_MATCHFLAG

The insider case

Marked verified. Still not paid.

After the analyst confirms the case, a claims processor resubmits the clinic's claim with the visit marked verified. Every member check passes. Omnareth refuses it on the watchlist, and the audit record shows the block it acted on.

Fair

Victims are not suspects

Only the identities a rule implicates are held responsible. Genuine customers named in a fraud attempt are never counted against.

Controls

Policy as rules

Your checks run on every request, with the reason and a remediation step recorded.

Evidence

For auditors and regulators

Every refused request is sealed with what was known at the time, and verifiable later.

Your process

Adapt the model

The Member model, its rules and its fraud identifiers are plain JSON you can change.

Pilot program

Put healthcare claims fraud prevention to work in 90 days

One lane, facility or process, in shadow mode, with results measured against criteria we agree with you. Qualified companies and organizations can apply.