Objective: To i) quantify the agreement between comorbidities documented within medical records and an orthopaedic trauma dataset; and ii) compare agreement between these sources before and after the introduction of new comorbidity coding rules in Australian hospitals. Study design and setting: A random sample of adult (≥ 16 years) orthopaedic trauma patients (n = 400) were extracted from the Victorian Orthopaedic Trauma Outcomes Registry (VOTOR). Diagnoses of obesity, arthritis, diabetes and cardiac conditions documented within patients’ medical records were compared to ICD-10-AM comorbidity codes (provided by hospitals) for the same admission. Agreement was calculated (Cohen's kappa) before and after the introduction of new coding rules. Results: All comorbidities had the same or higher prevalence in medical record data compared to coded data. Kappa values ranged from <0.001 (poor agreement) for coronary artery disease to 0.94 (excellent agreement) for type 2 diabetes. There was improvement in agreement between sources for most conditions following the introduction of new coding rules. Conclusion: There has been improvement in the coding of certain comorbidities since the introduction of new coding rules, suggesting that, since 2015, administrative data has improved capacity to capture patients’ comorbidity profiles. Consideration must be taken when using the ICD-10-AM data due to its limitations.