This paper is concerned with the use of a combined judgmental/statistical approach to research data audits, with a focus on illustrating the use of a stratified sampling technique. In the first part of the paper, we propose a two‐stage approach to screening research data reports for further evaluation. The two stages are (1) the assessment (evaluation) of the organization's (and research study team's) internal control structure and (2) the testing of the values (data points) in the research data report tables and appendices.
In the second part of the paper we provide a general description of the stratified sampling method to test values in research data report tables and appendices. The major decisions to be made when using the stratified sampling method are described in this section. These decisions include defining what will constitute a discrepancy, what the tolerable discrepancy rate should be, what reliability level is desired, and identifying strata and their relative audit intensities. This part of the paper concludes with a discussion of how to evaluate the results of the1 stratified sampling method.
The third part of the paper presents an illustration of the use of this audit approach on a toxicology study report. We conclude that this approach could provide quality assurance auditors with an efficient audit while providing reasonable assurance regarding audit results. This approach allows the auditor to make a statistical inference about the results obtained and aids the auditor in conducting a more structured analysis of the data.
Notes
We would like to thank Dr. Steven Brecher for his invaluable help in conducting the audit of the toxicology study which is described in this paper.