A management tool for controlling the rate of non-acceptable inpatient hospital claims

This paper demonstrates a tool for substantially improved monitoring of the validity of health insurance claims. Using a Bayesian regression model, we predict the probability of a non-acceptable claim (NAC) for each claim record and the expected number of NACs for any set of claims. When applied to a large set of hospital discharge claims, the tool shows a substantial improvement in the ability to estimate the actual number of NACs in the set. The tool permits ongoing monitoring of claims, more precise control, and a substantial reduction in audit cost in claims administration. It is conceptually applicable to other ongoing quality control systems, where inexpensively obtained information can be used to predict events that are costly to measure directly.