75 Internal Validity for Large Collaborative Clinical Datasets: Assessment of the CONGRESS Database

Abstract Aim Multi-centre clinical research collaboratives generate large, generalisable datasets, but concerns remain regarding data quality due to variability in validation practices and the involvement of trainees with limited clinical or academic experience. This study aims to evaluate the feasibility, methodology, and effectiveness of internal data validation within the CONGRESS database, to inform best practices for ensuring data integrity in collaborative research settings. Method The multicentre CONGRESS dataset of early oesophago-gastric cancer was assessed. A random 20% sample of patients was selected to meet a >15% target validation size. Patient, disease and outcome data were re-abstracted from medical records and entered into a validation dataset which was compared to the original database. Cohen’s kappa coefficient (κ) and Pearsons corelation (r) were calculated to express the strength of agreement between categorical and continuous variables, respectively. Results In total, 302 patients (18.1%) from the original CONGRESS database were included in the validation dataset and 3320 data points were compared between datasets (6640 total). The percentage of exact agreement for variables ranged from 82.5-98.7% (median 92.3%, IQR 86.3-95.7%). 9 variables (1645/2946, 55.8% data points) showed “almost-perfect” agreement (κ or r >0.8), 5 (1301/2946, 44.2%) showed substantial agreement (κ > 0.6). None showed weak or poor agreement. Conclusions This study provides strong evidence of internal validity for the CONGRESS collaborative clinical database. It presents key learning points and a methodology for data validation, which can be applied to other large collaborative databases. This approach aims to enhance confidence in the quality and reliability of research conducted through these platforms.