Accuracy of administrative coding in identifying hip and knee primary replacements and revisions

Rationale, aims and objectives: Hospital discharge notes can be used to identify rates of revisions in hip and knee arthroplasty surgeries if such administrative codes are accurate. In order to trust the data taken from the hospital discharge abstracts it is important to assess their reliability. The purpose of this study is to evaluate the accuracy of the administrative coding used in measuring revision rates for total hip and knee arthroplasty.

Methods: Validation coding was determined by two orthopaedic surgery residents who reviewed the operative, radiological and discharge summaries in order to identify the revision rates for total hip and knee arthroplasty. A random sample of 637 patients from two tertiary care hospitals was studied. These patients had total hip or knee arthroplasty between 1996 and 2006. All of these patients had an International Classification of Disease (ICD)-9CM or ICD-10CM code indicating what procedure they had done. The validation reviewers were blinded to the administrative codes used. The sensitivity, specificity and positive and negative predictive values of the administrative codes for revision rates were measured.

Results: Based on 1201 procedures performed on 637 patients, when comparing validation review versus hospital administrative chart coding for primary and revision surgeries of total hip and knee arthroplasty, the following data were obtained: for total hip arthroplasty sensitivity is 99%, specificity is 91%, positive predictive value is 91% and negative predictive value is 99%; for total knee arthroplasty sensitivity is 89%, specificity is 98%, positive predictive value is 97% and negative predictive value is 93%. The accuracy of ICD-9CM and ICD-10CM were 96% and 95%, respectively.

Conclusion: This study demonstrates that ICD-9CM and ICD-10CM codes can be used accurately when analysing hip and knee arthroplasty. This study was conducted in a large tertiary academic centre where a significant number of records analysts are employed; therefore, there should be little inter-hospital error. These results should help researchers understand the potential accuracy of classification for these procedures as part of an audit or quality assurance project.