Data Quality Assessment

DUCkS analyses will compare healthcare systems data with trial-collected data for mortality, treatment, second cancer incidence, and major adverse events. Trial-level analyses will compare hazard ratios derived from trial-based data and healthcare systems data using absolute and relative differences with confidence intervals. When events appear in both sources, their dates will be compared using median and interquartile differences with graphical displays. Completeness, sensitivity, specificity, and Cohen's kappa will be used as patient-level data quality indicators. The DUCkS analyses will classify each outcome as recorded in both sources, only in trial data, only in healthcare systems data, or absent from both datasets.