Statistical Audits

Statistical audits of claims and invoices are performed when the number of such is large and the dollar amounts of each are relatively small. A random sample of claims is drawn, a thorough audit is performed on the sample and the results are extrapolated to the population of claims or invoices. Standards for conducting such an audit have been developed by statisticians and accountants and are described in several textbooks and journal articles.
(linked files are in pdf format requiring Adobe Acrobat plug-in or reader)

Mt. Carmel Medical Center v Kansas SRS

• Spectrum critiqued an audit conducted by Kansas on a random sample of Medicaid claims submitted by the Center. Among the problems we found were 1) using an attribute sample design instead of a variables estimation sampling design, 2) failing to compute confidence limits so that the estimated number of invalid invoices could be tested for statistical significance, 3) using too small of a sample size to achieve the required degree of precision, 4) failing to test for and failing to correct for a skewed distribution of invoices, 5) failing to check the sample of invoices for both over and under charges.

Oak Hill Hospital v Kansas SRS

• Spectrum found similar problems with an audit that had been conducted for claims submitted by Oak Hill Hospital.