Optimization Of Manual Review Of Adverse Event Data With SAS

Data quality plays an important role in the outcome of the clinical study. Edit checks and manual reviews are the most common methods used for clinical data management (CDM) to ensure data quality. However, edit checks in clinical data management system have limitations, and manual review process can be time consuming, labor intensive and subject to human errors. This paper is to mainly discuss using SAS tool to assist manual data review as part of data cleaning process focusing on Adverse Event data. The purpose is to demonstrate the improved efficiency and data quality by using SAS tool to identify data discrepancies therefore to optimize data review process, improve work efficiency and ensure the quality of clinical safety data.ObjectiveA method was introduced in this article for optimizing manual review of adverse event (AE) data, the effect of which on clinical data management workload, efficiency and accuracy were evaluated. A new manual review process was established to improve efficiency and quality, manage clinical safety data more effectively and quickly, and ensure data quality.MethodsA CDASH standard variable database was established. Traditional data listing was automatically manipulated by SAS. Matched records were filtered in advance, while the remaining records were produced in a more concise and convenient SAS output listing. The optimization method was deployed in eight clinical trials, the number of AE to be checked was compared, and the effect on workload was studied. The changes of manual review time and accuracy were investigated by a cross-over study.ResultsManual review process of AE data was optimized by SAS. This optimization method can identify and screen matched records quickly and exactly, and produce matching reports. Meanwhile, the unmatched reports were labeled with different colors to facilitate data managers to distinguish data from different modules. This optimization method reduces the number of AE to be reviewed manually and workload, significantly improves efficiency (P<0.05) without changing the accuracy (P>0.05). ConclusionThis method as an auxiliary tool to traditional manual review has favorable optimization effects, and can reduce human dependence. It could ensure data quality, and possess high value of practical application.