BACKGROUND: Arthroscopic knee surgery has been a topic of significant controversy in recent orthopaedic literature. Multiple studies have used administrative (Victorian Admitted Episodes Dataset and Centre for Health Record Linkage) data to identify trends in practice. This study explored the usage and reporting of arthroscopic knee surgery by conducting a detailed audit at a major Victorian public hospital. METHODS: A database of orthopaedic procedures at St Vincent's Hospital Melbourne was used to retrospectively identify cases of knee arthroscopy from 1 December 2011 to 1 April 2014. Procedures were categorized as diagnostic or interventional, and native and prosthetic joints were analysed separately. Procedure codes were reviewed by comparing a registrar, auditor and hospital coders for agreement. RESULTS: Of the 401 cases for analysis, 375 were conducted in native knees and 26 in prosthetic joints. Of native knees, 369 (98.4%) were considered interventional. The majority of these were conducted for meniscal pathology (n = 263, 70.1%), osteoarthritis (OA) (n = 25, 6.7%) and infection (n = 28, 7.6%). Comparison of codes assigned by different parties were found to be between 57% (k = 0.324) and 70% (k = 0.572) agreement, but not statistically significant. CONCLUSIONS: In this study, the most common indication for arthroscopy was meniscal pathology. Arthroscopy was rarely performed for OA in the absence of meniscal pathology. Diagnostic arthroscopy was rarely performed in the native knee, and fair to moderate agreement existed between parties in assigning Medicare Benefits Schedule procedure codes.