Authors' Reply: Critique of ICD Code Accuracy in Identifying Cardiovascular Events in CKD and Kidney Failure Populations

We appreciate the thoughtful Letter to the Editor from Wang et al.1 and thank them for their generous remarks regarding the value of our JASN study.2 We also acknowledge and agree with their observations concerning the study's limitations. First, it is true that International Classification of Diseases (ICD) codes versus physician adjudication for cardiovascular events had low sensitivity overall. As we acknowledged in the discussion, this may have been because of the limitations of the data collection. Unfortunately, we were unable to determine whether the low sensitivity was due to missed diagnoses by ICD codes, underreporting diagnoses, or incorrect ICD code usage by the treating patient clinician. For that reason, we advise that use of ICD codes versus physician adjudication may not be appropriate for all studies. Second, given that ICD-10 code updates were meant to more accurately capture diagnoses than ICD-9 codes, we agree that it is unexpected that we observed lower positive predictive values for all four outcomes when using ICD-10 codes. Prior studies in other populations have also shown variable results when comparing ICD-9 versus ICD-10 codes.3–5 This certainly warrants further investigation. Third, the authors raised concerns about using primary ICD codes in our analyses. Our decision to use primary codes came out of practicality of ICD code collection changes in our data from the Chronic Renal Insufficiency Cohort study. This approach allowed us to capture the greatest number of events during study follow-up. However, we also performed a secondary analysis in which we reported results when using both primary and secondary codes (Figures 1 and 2). In conclusion, we are grateful for the thoughtful engagement with our work. While our study represents an initial step toward understanding the accuracy of ICD codes compared with physician adjudication for identifying cardiovascular events in individuals with CKD and kidney failure, it also highlights important areas for future research. ICD codes may not be ideal for outcome identification in all research settings. A clear understanding of their strengths and limitations of this approach can help investigators choose the most appropriate methods based on their specific research question.