OIG: UnitedHealthcare of Wisconsin (H5253) risk scores drove $46.9M in estimated unsupported diagnosis overpayments

On September 10, 2026, HHS Office of Inspector General issued report A-07-24-01214, a Medicare Advantage compliance audit of specific diagnosis codes that UnitedHealthcare of Wisconsin, Inc. (contract H5253) submitted to CMS. The report was posted September 15, 2026. In the report, OIG shortens the audited entity to "United."

OIG found that most of the selected high-risk diagnosis codes in its sample did not comply with Federal requirements. For 183 of 250 sampled enrollee-years, medical records did not support the diagnosis codes (or United could not locate the records) and resulted in $722,280 in sample overpayments. Sixty-seven enrollee-years validated. On that basis, OIG estimated that United received at least $46.9 million ($46,913,121) in overpayments for payment years 2020 and 2021 (service years 2019 and 2020).

Read that dollar figure carefully. OIG estimates overpayments at the lower limit of a two-sided 90-percent confidence interval. The report states the exact estimate as at least $46,913,121 ($46,668,024 for the statistically sampled enrollee-years plus $245,097 for the nonstatistically sampled mis-keyed enrollee-years). This is a performance-audit projection from a sample, not a court judgment, not a CMS RADV final determination, and not a fraud allegation. Footnote 29 in the report says OIG audit recommendations do not represent final determinations; CMS action officials decide whether an overpayment exists and how to recoup.

This audit is the companion series report to the live HumanaChoice H5216 high-risk diagnosis audit (A-05-24-00010), issued and posted in the same window. Same OIG high-risk-group family, different contract, different sample, different estimate. Do not collapse the two into one story.

What OIG sampled

OIG identified 28,410 unique enrollee-years tied to 11 high-risk diagnosis groups and limited review to the payment portions associated with those codes ($64,458,294). It selected 250 enrollee-years for audit:

• a stratified random sample of 200 (out of 28,279) enrollee-years for the first 10 high-risk groups (20 per group)
• a nonstatistical sample of 50 (out of 131) enrollee-years for the potentially mis-keyed diagnosis-code group

An independent medical review contractor validated HCCs against the medical records United provided. When an HCC was not validated, OIG counted the payment effect in the overpayment calculation (with limited offsets when a less severe related HCC was supported). United could not locate medical records for 3 of the 250 sampled enrollee-years.

For context, OIG states that under contract H5253 United provided coverage to 676,374 enrollees and that CMS paid United approximately $16.6 billion for payment years 2020 and 2021. Those coverage figures are background; the overpayment estimate applies only to the high-risk diagnosis portions OIG audited.

The 11 high-risk groups: sample results

OIG's Table 4 and the findings sections break out unsupported enrollee-years and sample overpayments by group. Desk meaning below is what the pattern usually looks like on a coding queue, framed the way OIG defined each high-risk group.

OIG's methodology example of a potential mis-key (payment years 2020 and 2021 under the CMS-HCC models then in force): ICD-10 E43 (protein-calorie malnutrition, which in that audited period mapped to the protein-calorie malnutrition HCC) could be mis-keyed as I43 (which in that audited period mapped to the congestive heart failure HCC). In OIG's example that is a first-character substitution (E to I), one form of the report's broader screen for codes that may have been mis-keyed because numbers were transposed or other data-entry errors occurred. Those E43/I43 HCC mappings are OIG's PY2020-2021 / audited-model-year example only; they are not stated here as current V28 payable mappings.

The coding pattern OIG keeps finding

Across stroke, AMI, embolism, the cancer groups, sepsis, and pressure ulcer, the failure modes repeat:

1. Acute or active disease coded when history-of or sequelae may be what the record supports. Acute stroke and AMI findings repeatedly note past medical history of the event without justification for an acute code at the physician service. Breast and colon cancer were unsupported for all 20 sampled enrollee-years in each group, nearly always because history-of was what the record supported.

2. Encounter pattern that does not match how the condition is typically treated. Embolism: Embolism HCC on only one DOS and no anticoagulant dispensed; OIG says history-of embolism typically should have been used. Active cancers and severe pressure ulcers: a single DOS without matching treatment evidence. Sepsis: a single physician/outpatient claim without corresponding inpatient sepsis.

3. Potentially mis-keyed unrelated conditions. One stray code that maps to a different HCC, sitting next to repeated codes for another condition, is enough to land in the mis-keyed stratum. The mis-keyed group produced the largest sample-dollar hit in this audit ($245,097 across 22 of 50 sampled enrollee-years).

That is the same family of documentation risk the desk already covers for acute stroke in the live OIG acute stroke overpayments article: acute event codes need acute-event documentation, and history-of codes do not carry the same HCC weight.

What OIG recommended, and what United said

OIG recommends that UnitedHealthcare of Wisconsin, Inc.:

1. refund to the Federal Government the $46,913,121 of estimated overpayments;

2. determine, for the 81 enrollee-years OIG did not review in the potentially mis-keyed group, whether medical records support the unrelated diagnosis, and refund any resulting overpayments;

3. identify similar noncompliance for these high-risk diagnoses after the audit period and refund resulting overpayments; and

4. continue examining compliance procedures and strengthen controls so high-risk diagnosis codes submitted for risk adjustment meet Federal requirements.

United disagreed with some of OIG's findings and requested that OIG withdraw all of the recommendations. Among other points in the report's response appendix, United contested findings for a subset of sampled enrollee-years, challenged aspects of OIG's statistical methodology (including use of the 90-percent confidence interval), and objected to the refund and follow-up review recommendations. OIG maintained its findings and recommendations in the final report (with adjustments noted where United's additional information changed individual sample items).

Again for the desk: an OIG recommendation to refund an estimated amount is not the same thing as CMS collecting that amount. CMS decides next steps under its policies, and MA organizations retain RADV appeal rights if a disallowance is taken (42 CFR § 422.311, as cited in the report).

What this means for your risk-adjustment queue

If you work MA encounter or chart-review queues, treat this report as a priority list for documentation QA, not as a fraud story.

• Acute stroke: confirm a same-year inpatient or outpatient hospital acute-stroke diagnosis when the code is acute on physician claims (1-5 DOS). AMI / sepsis: confirm the corresponding inpatient record OIG required for those strata. Physician-only acute coding without the matching hospital corroboration is how several of these strata were built.
• Active cancer HCCs on a single DOS: ask whether the record supports active disease versus history-of. History-of cancer typically does not map to the active cancer HCC. In this sample, breast and colon cancer were unsupported for every sampled enrollee-year.
• Embolism: when an Embolism HCC appears on only one DOS and no anticoagulant was dispensed, ask whether history-of embolism is what the record supports (OIG's inclusion screen for this stratum; OIG says history-of embolism typically should have been used).
• Severe pressure ulcer on one DOS: look for evidence of ongoing evaluation or treatment consistent with the severity implied by the HCC.
• Mis-keys: watch for a single unrelated HCC-driving code beside repeated codes for another condition. OIG identified scenarios where diagnosis codes could have been mis-keyed because numbers were transposed or other data-entry errors occurred (its E43-to-I43 methodology example is a first-character substitution from the audited model years, under OIG's transposed-numbers-or-other-data-entry-errors screen).
• RADV context: this OIG contract-level audit is separate from CMS's broader RADV schedule, but the documentation standard is the same. See the live RADV quarterly audits article for the plan-side audit calendar.

For documentation review, the MEAT Criteria Guide covers what counts as evidence a condition was monitored, evaluated, assessed, or treated at the DOS. The Evidence Checker helps pressure-test whether a record actually supports the HCC before submission. Use the ICD-10 Encoder when you need to confirm billable status and related history-of alternatives without inventing specificity the chart does not support.

This page stays on OIG report A-07-24-01214 as written. It does not assert that UnitedHealthcare of Wisconsin committed fraud, does not treat the $46.9 million figure as collected dollars, and does not substitute for CMS or plan-specific RADV guidance.