κ=0.97: A Practical Framework Any Hospital Can Implement for Research-Grade Data Quality

Background: Medical device registries in transitional healthcare systems face substantial challenges achieving reliable data extraction due to documentation fragmentation and absence of integrated electronic health records. Conventional single-source or dual-source extraction methods demonstrate moderate inter-rater reliability (κ=0.50–0.70), limiting research validity. We hypothesised that systematic multi-source data triangulation could achieve near-perfect reliability comparable to advanced registry systems whilst providing a replicable framework for resource-limited settings. Methods: We conducted a retrospective methodological validation study of 176 rotational atherectomy procedures performed between January 2020 and December 2023 at Hospital Regional General Ignacio Zaragoza, Mexico. Three independent, blinded investigators extracted device-specific data from thirteen documentary sources organised into six validation domains. Inter-rater reliability was assessed using Cohen's κwith 95% confidence intervals. External validation was performed by an investigator from a geographically separate institution. Results: The thirteen-source framework achieved overall inter-rater reliability of κ=0.97 (95% CI: 0.94–0.99), with 96.6% concordance across three extractors. External validation demonstrated κ=0.94, confirming reproducibility without subspecialty expertise. Complete device identification was achieved in 100% of procedures. Comparative bootstrap analysis revealed 49% improvement over single-source extraction (κ=0.65, p<0.001) and 24% improvement over dual-source methods (κ=0.78, p<0.001). Conclusions: Systematic multi-source data triangulation enables transitional healthcare systems to achieve research- grade inter-rater reliability exceeding advanced registry benchmarks. Documentation multiplicity, when leveraged through structured protocols, transforms from methodological limitation to asset.