Explainable Peer-Comparison Screening for Medicare Payment Integrity: Methods and Validation Against the HHS-OIG Exclusion List

Improper payments in Medicare and Medicaid are concentrated in categories that come to light only when a reviewer requests and examines the records behind a claim: in fiscal year 2025, missing or insufficient documentation accounted for 65% of Medicare fee-for-service improper payments, and incorrect coding for 11.1%. Screening methods that direct reviewers to the right claims therefore carry real value, and federal program-integrity guidance asks that each selection carry a supporting rationale. This paper presents transparent screening methods that attach a plain-language reason, with its figures, to every flag. Three claim-level rules flag possible duplicate payments, impossible service dates, and payments far above comparable claims, the last using a robust z-score built on the median and the median absolute deviation; a provider-level rule applies the same statistic to standardized Medicare payment per beneficiary. In a controlled test on 90,705 CMS synthetic Medicare claims, the rules recovered 200 of 200 planted duplicates, 195 of 200 planted inflated payments, and 200 of 200 planted date errors, against 0.22% to 0.33% expected from random selection. On CMS’s 2023 Medicare Physician and Other Practitioners data (9,660,647 provider-service records; 1,175,281 providers; $93.7 billion in payments), flagged providers were 2.73 times as likely as providers overall to appear on the HHS-OIG List of Excluded Individuals/Entities (73 against 26.7 expected; one-sided binomial p = 1.2 × 10−13), and 71 of the 73 were excluded after the data year. At an equal review budget, the method identified more excluded providers than a classic mean-and-standard-deviation outlier score (63) or a largest-billers ranking (46), and it is the only one of the three that states a reason for each selection. Code, settings, and full run records are public.