Internal validation protocol for large collaborative clinical data sets: assessment of the CONGRESS database.

Area of Science:

• Clinical research methodology
• Data quality assurance
• Oesophagogastric cancer research

Background:

• Multicentre collaboratives generate large datasets but face challenges with data quality due to trainee involvement.
• Validation practices in such studies are inconsistent, potentially leading to bias.
• This study addresses these concerns by evaluating internal data validation within the CONGRESS database.

Purpose of the Study:

• To outline the methods, feasibility, and outcomes of internal data validation for a multicentre clinical database.
• To establish a reproducible framework and benchmarks for validating large collaborative clinical datasets.
• To ensure reliable, high-quality research outcomes from multicentre databases.

Main Methods:

• A random 20% sample of patients from the multicentre CONGRESS dataset of early oesophagogastric cancer was selected for validation.
• Patient, disease, and outcome data were re-abstracted from medical records and compared with the original database.
• Cohen's kappa coefficient (κ) and Pearsons correlation (r) were used to assess agreement for categorical and continuous variables.

Main Results:

• 302 patients (18.1%) were included in the validation set, with 3,320 data points compared.
• The percentage of exact agreement for variables ranged from 82.5% to 98.7% (median 92.3%).
• Nine variables showed 'almost perfect' agreement (κ or r > 0.8) and five showed substantial agreement (κ > 0.6), with no weak or poor agreement observed.

Conclusions:

• Internal data validation using the CONGRESS database proved feasible and effective.
• A reproducible framework and benchmarks for validating large collaborative clinical datasets have been proposed.
• This approach provides a standard for ensuring high-quality research outcomes across multicentre databases.