Blogs

A program integrity team finds a real fraud signal. The pressure to act is legitimate: public dollars are at risk, beneficiaries may be harmed, and bad actors can move quickly. But a broad control can also touch legitimate providers, especially in markets where there are not many alternatives.

Share: Fraud controls can protect public dollars and still create access pressure. The tradeoff depends on review capacity, payment scope, provider supply, and beneficiary complexity. #Causality #Healthcare #FraudWasteAbuse

The Problem Is Not Always One Thing

The mistake is treating fraud control as a one-direction story. Stronger controls can improve program integrity, but access disruption depends on review capacity, payment-suspension scope, provider supply, existing capacity, and beneficiary complexity. A moratorium or payment hold may be the right tool, while still creating operational side effects that need to be managed.

Why The Obvious Explanation Can Mislead

The obvious explanation is that more control means better outcomes. It may mean better program integrity, but it does not automatically mean lower access risk. If targeted review capacity is low, broad payment holds and enrollment restrictions can create a bottleneck for legitimate providers too.

The Question For This Example

This example asks how access-disruption risk changes after observing a high-fraud-control scenario, and what changes if review capacity and payment-suspension scope are set more narrowly.

The target outcome is AccessDisruption, which means access disruption in this teaching model.

A Small Model Of The Situation

The model is a synthetic tradeoff graph. It does not judge a specific enforcement action, estimate fraud, decide provider sanctions, or model legal due process. It only shows why access risk is not the same thing as fraud signal intensity.

The graph has 17 nodes:

• FraudSignalIntensity: Fraud signal intensity (low / high)
• BillingAnomalyRate: Billing anomaly rate (low / high)
• OwnershipChangeRisk: Ownership change risk (low / high)
• GeographicProviderSupply: Geographic provider supply (adequate / low)
• ExistingProviderCapacity: Existing provider capacity (adequate / low)
• BeneficiaryComplexity: Beneficiary complexity (low / high)
• ScreeningStrictness: Screening strictness (targeted / high)
• MoratoriumActive: Moratorium active (no / yes)
• TargetedReviewCapacity: Targeted review capacity (adequate / low)
• InvestigationBacklog: Investigation backlog (low / high)
• PaymentSuspensionScope: Payment suspension scope (narrow / broad)
• LegitimateProviderBurden: Legitimate provider burden (low / high)
• NewProviderEntryBlocked: New provider entry blocked (no / yes)
• ProviderExitRisk: Provider exit risk (low / high)
• CareContinuityRisk: Care continuity risk (low / high)
• ProgramIntegrityRecovery: Program integrity recovery (strong / low)
• AccessDisruption: Access disruption (no / yes)

What Is Sourced And What Is Synthetic

CMS fraud and moratoria materials support including program integrity signals, enrollment controls, and moratoria. HHS OIG materials support treating fraud, waste, and abuse as a formal oversight domain. The access-disruption probabilities are synthetic.

Moratoria and enrollment controls

CMS materials describe temporary moratoria and enrollment restrictions as tools to prevent fraud, waste, and abuse.

Program integrity signals

CMS fraud materials support including billing anomalies, investigations, and analytics as program-integrity mechanisms.

Oversight boundary

HHS OIG material supports fraud prevention as a public oversight concern, not a private operational preference.

The Query The Graph Cannot Show By Itself

The visual graph is the best way to inspect the structure. The code-backed query is there for a narrower reason: we want to separate what is learned by observing a messy case from what changes if one condition is set differently.

Using the Darkstar Python SDK, a subscribed SDK user can load the same BBN JSON, ask for the target probability after evidence is entered, then run the action query by setting a node before querying again. The download includes this as illustrative SDK code, while the standalone query_example.py reproduces the article numbers without shipping the SDK.

import json

from darkstar import DiscreteModel

TARGET = "AccessDisruption"
OBSERVED_EVIDENCE = {
    "FraudSignalIntensity": "high",
    "MoratoriumActive": "yes",
    "TargetedReviewCapacity": "low",
    "GeographicProviderSupply": "low"
}
INTERVENTION_CONTEXT = {
    "FraudSignalIntensity": "high",
    "MoratoriumActive": "yes",
    "GeographicProviderSupply": "low"
}
INTERVENTION = {
    "TargetedReviewCapacity": "adequate",
    "PaymentSuspensionScope": "narrow"
}


def probability_after_observing(model, evidence):
    query = {"nodes": [TARGET], "evidence": evidence}
    result = json.loads(model.pquery(json.dumps(query)))
    return result[TARGET]["yes"]


def probability_after_setting(model, setting, evidence):
    changed_model = model.intervene(json.dumps(setting))
    return probability_after_observing(changed_model, evidence)


with DiscreteModel(model_json) as model:
    observed = probability_after_observing(model, OBSERVED_EVIDENCE)
    acted = probability_after_setting(model, INTERVENTION, INTERVENTION_CONTEXT)

That is the article’s main modeling distinction. Observing a pattern is not the same as setting one part of the system differently.

The standalone query code gives this synthetic result:

Baseline access disruption: 19.0%
Observed scenario (moratorium pressure with low review capacity): 24.0%
Action scenario (adequate review capacity and narrow payment holds): 21.0%
Same context with the action left unfavorable: 26.1%
Synthetic reduction versus the observed scenario: 3.0 percentage points

What We Learn By Observing The Situation

The observed scenario sets:

• Fraud signal intensity = high
• Moratorium active = yes
• Targeted review capacity = low
• Geographic provider supply = low

The observed scenario sets high fraud signal, an active moratorium, low targeted review capacity, and low geographic provider supply. In the synthetic model, that combination raises access-disruption risk because legitimate capacity can become scarce while investigations are still unresolved.

What Changes If We Set One Condition Differently

The action scenario sets:

• Targeted review capacity = adequate
• Payment suspension scope = narrow

The action query does not turn off fraud controls. It sets targeted review capacity to adequate and narrows payment-suspension scope. That asks whether a more focused operational response can keep integrity pressure while reducing access-side effects.

What This Example Cannot Prove

This example cannot evaluate a real moratorium, prove overreach, or recommend enforcement policy. It is designed to keep two truths visible at the same time: fraud controls can be necessary, and their operational design can affect access.

Try The Model

Open the companion Darkstar model: Open Health Care Fraud Control Tradeoff BBN

Set PaymentSuspensionScope to broad while keeping review capacity adequate. Then set review capacity low with a narrow suspension. The comparison shows why scope and capacity should be considered together.

Download the teaching package: Health Care Fraud Control Tradeoff BBN ZIP

The public package contains only synthetic teaching assets. It does not contain claims data, provider identifiers, investigations, enforcement criteria, SDK packages, or internal API details.

Sources

1. CMS: Crushing Fraud, Waste, & Abuse
2. CMS Announces Nationwide Hospice and Home Health Agency Enrollment Moratoria
3. CMS Provider Enrollment Moratoria
4. HHS OIG Fraud