Tools › Finance and Business › Auditing › Audit Sampling › Example

Overpayment audit of 100 claims in three strata: projected error and its 90 % lower limit

ToolAudit Sampling
SourceOriginal illustrative case (seeded synthetic data). Method: U.S. Department of Health and Human Services, Office of Inspector General, Office of Audit Services (2010). RAT-STATS Companion Manual, Rev. 5/2010, Stratified Variable Appraisal, pp. 3-15 and 3-16.
LevelProfessional

Problem

A payer's 3,480 paid claims are split into three strata by amount (2,400, 900 and 180 claims; book totals 239,860.44, 436,711.08 and 442,202.54). Random samples of 40, 40 and 20 claims are audited and 24 of the 100 are found in error. Project the overpayment (book minus audited) to the population with the difference estimator and give the two-sided 90 % confidence limits.

Inputs to Enter

  • Task: project a sample (classical variables); confidence level 90 %; estimator difference
  • Population summary: stratum 1: 2,400 claims, 239,860.44; stratum 2: 900 claims, 436,711.08; stratum 3: 180 claims, 442,202.54
  • Sample: 100 rows of id, book amount, audited amount, stratum (the tool's opening data)
Open the Audit Sampling tool (loaded with this example)

Answers

  • Stratum estimates N · mean error: 2,400 · 22.24 = 53,379.60; 900 · 51.41 = 46,272.83; 180 · 578.73 = 104,171.49; point estimate 203,823.92
  • Stratum variances N² · (1 − n/N) · s² / n with s = 43.75, 135.49 and 1,330.50; standard error of the total 56,350.84
  • Two-sided 90 % limits with Z = 1.644854: 111,135.03 to 296,512.80 (precision ± 45.47 %); the lower limit is 55 % of the estimate
  • Two-sided 80 % lower limit 131,607.41, which is also the lower limit of a one-sided 90 % interval
  • On the same sample the ratio estimator gives 181,130.53 (standard error 54,914.96) and mean per unit 63,655.58 (standard error 77,257.51)

Notes

The tool's default inputs are this case; press Submit, or open /analysis-tools/business/audit/audit-sampling/?try=stratified-overpayment. Every number is synthetic. The tool computes; whether a sample design is valid for a given audit is the auditor's judgment.

More Examples for This Tool