Is the American Joint Replacement Registry Able to Correctly Classify Revision Total Knee Arthroplasty Procedural Diagnoses?

Background: The American Joint Replacement Registry (AJRR) is a powerful tool for the study of revision total knee arthroplasty (rTKA). The AJRR uses International Classification of Diseases-10 (ICD-10-CM) codes for recording surgical diagnoses. However, the validity of this methodology is unknown. The purpose of this study was to determine the accuracy of ICD-10-CM codes, as used by AJRR, in classifying rTKA diagnoses. Methods: There were 988 rTKAs performed from 2015 to 2021 identified in our institutional total joint registry (TJR). Revision diagnoses were obtained from TJR, in which trained abstractors prospectively record diagnoses independent of ICD-10-CM data. The ICD-10-CM diagnosis codes submitted to AJRR were retrieved for the same procedures. The accuracy of ICD-10-CM codes for classifying rTKA diagnoses as septic versus aseptic, aseptic loosening, instability, and periprosthetic fracture was assessed using Cohen's Kappa statistics, sensitivities, and specificities. Results: Concordance between AJRR-submitted codes and TJR was excellent (97.3%, k = 0.9) for identifying septic versus aseptic revisions. Agreement for aseptic diagnoses varied from very good for loosening (k = 0.65) and instability (k = 0.64) to fair for periprosthetic fracture (k = 0.36). Specificity was high (> 94%) for all three diagnoses, but sensitivity was lower at 71%, 63%, and 28% for loosening, instability, and periprosthetic fracture, respectively. Conclusion: The AJRR submitted ICD-10-CM diagnosis codes correctly classified rTKA cases as septic or aseptic with remarkable accuracy, but accuracy for more granular diagnoses varied. These data demonstrate the potential for diagnosis-specific limitations when using administrative claims data for registry reporting and have important implications for researchers using ICD-10-CM data.