Validation of a hospital-based joint registry: quantification of errors and maximizing utility

Data quality standards for multicenter arthroplasty registries have been previously established. However, no studies have assessed a hospital-based arthroplasty registry. A total of 6912 hip and knee procedures were queried in a hospital-based arthroplasty registry (HJR) and an administrative research database (Research Patient Data Registry). The HJR demonstrated a 93.8% capture rate. Four hundred seventy-seven cases contained data errors, with 65.7% occurring in the HJR. Most of these conflicts were attributable to random error. Systematic miscoding of procedures was observed in both databases and was most likely due to unclear definitions of procedural classifications. Our results confirm the validity of data in the HJR and suggest that using automated data capture systems and establishing a clear consensus on procedural terminology can improve the quality of data registries.