VARDR Program Review
Missouri

Synthetic data. Practice names are invented. Several practices carry planted billing patterns, and several benign look-alikes are planted to show what the analysis declines to flag. Cases, people and outcomes are invented too.

Program integrity · Medicaid · Missouri

Recipient integrity

Enrollment records checked against outside sources and against each other. An indicator is a question to verify with the person or the source. No adverse action may rest on an unverified match, and most of what looks like a pattern here is how people actually live.

Cases screened
92
in the synthetic enrollment file
Indicators to verify
6
from outside matches and identity checks
Set aside by rule
4
shared addresses and representatives that are not links
Family pairs removed
1
related cases, filtered before any link is drawn

Verification queue

One row per indicator. The reference is a pseudonym that is stable within this deployment and meaningless outside it.

ReferenceIndicatorPriorityStandingObserved
5a78f02f8c4cf2b4Death record matchIdentityVITAL-001Priority reviewIndependently verified.Matching agreement CMA-2026-012026-08-20
f9c3ad48814050eaIncarceration record matchIdentityINCAR-001Scheduled reviewUnverified. Confirm with the person or the source before anything follows from it.Matching agreement CMA-2026-022026-09-02
094dc5977004e59aAdditional identity documents requestedIdentityIDV-002RoutineUnverified. Confirm with the person or the source before anything follows from it.2026-07-06
6bb23f2bfe7567d2Record suggests this person may have been impersonatedIdentityIDV-003RoutineRoute to help, not review: the record suggests this person may have been impersonated.2026-07-04
67f544e792996af2Identity could not be confirmed at applicationIdentityIDV-001RoutineUnverified. Confirm with the person or the source before anything follows from it.2026-07-01

Re-identification happens in the State’s own case system, under its own permission, with the access logged; it is not offered here. Your role does not include that permission.

Linked cases

Groups of cases joined by a high-confidence link, such as one representative filing from one device for cases at different addresses. A shared address, phone or email alone never forms a group.

  • e5cfedb6956099fb
    <11 cases (small count, not shown exactly)
    For verification
    • Same authorized representative across many households · high confidence
    • Applications submitted from the same device · high confidence

    Addresses: 4 single-family homes.

Seen and set aside

Shared addresses and representatives the rules decline to treat as links, with the reason. A nursing facility, an apartment building and an enrollment assister are how the program reaches people.

  • Same resolved residential address36 cases

    Address class CONGREGATE_EXCLUDED is not edge-eligible (nursing-facility). No edges drawn between its 36 households.

  • Same resolved residential address12 cases

    Address class MULTI_UNIT_UNRESOLVED is not edge-eligible. No edges drawn between its 12 households.

  • Same resolved residential address8 cases

    Address class MAIL_ONLY is not edge-eligible. No edges drawn between its 8 households.

  • Same authorized representative across many households30 cases

    30 households share this same authorized representative across many households, above the cap of 12. Authorized representatives are lawful and often institutional — a community agency, a shelter caseworker, a care home operator. Volume alone is expected for those; the cap exists so an agency representative does not generate a cluster. Treated as infrastructure; no edges drawn.

Coverage lost and regained

Aggregate only. There is no list of people behind these numbers, by construction.

Why coverage endedEndedBack within 6 monthsShare who returned
Procedural closure (missed paperwork or deadline)643758%
Over income or resource limit22418%
Voluntary withdrawal or moved away1218%
Deceased600%

Returning after a procedural closure is ordinary churn: mostly people who were eligible throughout and lost coverage to paperwork. A high rate is a signal about the agency’s closures, which cost staff time to reopen, and never an indicator about a person. None of these reasons can produce a per-case signal.

Does the rule fall evenly?

How often the shared-representative rule raises a case, by rurality. A stratum raised at more than 1.25× the lowest rate is a prompt to examine the rule before relying on it.

StratumCasesRaisedRateAgainst the lowest rateReading
Rural counties1,900170.89%1.19×Within the threshold
Micropolitan counties3,100260.84%1.12×Within the threshold
Metropolitan counties5,200390.75%1.00×Within the threshold
Frontier countiesToo few cases to measure. Reported, not dropped.

Flag rates by stratum, with the impact ratio measured against the lowest non-zero rate. A breach is a prompt to examine the rule, not a determination that it is unlawful — whether a disparity is justified is a legal and policy judgement for the agency.