(167) A retrospective look at the efficiency of 100% source data verification and the possibilities presented by risk-based quality management in increasing monitoring efficiency while maintaining good data quality.

Background: Risk-based quality management (RBQM) is a process in clinical projects to detect data quality and identify potential risk centrally using various data sources. The main goals of RBQM are to improve data quality, patient safety, and efficiency. Data quality is currently monitored using 100% source data verification (SDV). SDV is a method used in clinical trials to ensure the data entered in a Case Report Form (CRF) is accurate, by comparing the CRF with source data. This process, typically done manually, costs a lot of resources and might not be the most efficient way of achieving clean data.

Objectives: To identify the impact of 100% SDV on all data in a Phase 2 multi-center, randomized, double-blind, placebo-controlled, parallel group study, compared to reducing SDV / Queries to a few targeted unique forms.

Methods: CRFs contain fields that must be entered by site staff. This can, for example, be the subject's demographics, a questionnaire, or laboratory test results. Each field can then be queried by an authorized Clinical Research Assistant (CRA) if, after digital review, the data seems out of order. SDV is performed at study end and/or during periodic monitoring visits. For the purposes of this analysis, we collected weekly exports of the CRF and compared each new export with the one from the preceding week. The changes in data were then linked to whether a CRA performed a query and/or SDV took place.

Results: In the given study, we identified 53 unique forms, containing a total of 29’027 fields. On these fields, a total of 1’687 queries were placed, of which about a quarter (27%, n=448) resulted in a data change. Of these 448 data changes, almost all changes (96%) were made in the same 55% of unique forms (e.g., demographics, medical history). When we look at changes made after SDV, we noticed that 5’497 fields (19%) were changed. However, 98% of required changes occurred in a similar subset of unique forms (46%; e.g., demographics, medical history).

Conclusions: Our retrospective analysis allowed us to pinpoint the unique CRFs costing the most investigative resources. Furthermore, our findings suggest, despite intensive resource allocation, SDVs had a relatively low impact on data quality on more than half of the unique forms. Ongoing work will build off these findings, evaluating data from two additional randomized placebo-controlled trials to determine whether we can effectively improve the data quality of the study in real time by targeting only the high-impact forms prior to SDV.