Background and purpose — Studies describing time-related change in reasons for knee replacement revision have been limited to single regions or institutions, commonly analyze only 1st revisions, and may not reflect true caseloads or findings from other areas. We used revision procedure data from 3 arthroplasty registries to determine trends and differences in knee replacement revision diagnoses. Patients and methods — We obtained aggregated data for 78,151 revision knee replacement procedures recorded by the Swedish Knee Arthroplasty Register (SKAR), the Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR), and the Kaiser Permanente Joint Replacement Registry (KPJRR) for the period 2003–2017. Equivalent diagnosis groups were created. We calculated the annual proportions of the most common reasons for revision. Results — Infection, loosening, and instability were among the 5 most common reasons for revision but magnitude and ranking varied between registries. Over time there were increases in proportions of revisions for infection and decreases in revisions for wear. There were inconsistent proportions and trends for the other reasons for revision. The incidence of revision for infection showed a uniform increase. Interpretation — Despite some differences in terminology, comparison of registry-recorded revision diagnoses is possible, but defining a single reason for revision is not always clear-cut. There were common increases in revision for infection and decreases in revision for wear, but variable changes in other categories. This may reflect regional practice differences and therefore generalizability of studies regarding reasons for revision is unwise.