Overpayment models for medical audits: multiple scenarios

Abstract

Comprehensive auditing in Medicare programs is infeasible due to the large number of claims, therefore, the use of statistical sampling and estimation methods is crucial. We introduce super-population models to understand the overpayment phenomena within the claims population. The zero- and one-inflated mixture-based models can capture various overpayment patterns including the fully legitimate or fraudulent cases. We compare them with the existing models for symmetric and mixed payment populations that have different overpayment patterns. The distributional fit between the actual and estimated overpayments is assessed. We also provide comparisons of models with respect to their conformance with Centers for Medicare and Medicaid Services (CMS) guidelines. In addition to estimating the dollar amount of recovery, the proposed models can help the investigators to detect overpayment patterns.