Clinically meaningful categorisation of ICD-10-AM (Australian modification)

Background: Current methods of categorising the International Statistical Classification of Diseases and Related Health Problems (ICD) have limitations when deciphering administrative data and monitoring healthcare outcomes. These include many-to-one relationships, non-linear sequencing, collinearity, and ambiguous miscellaneous (residual) codes. Objective: Describe novel methodology for clinically meaningful categorisation of 12th Edition of ICD Version 10 Australian modification (ICD-10-AM). Setting: State of Victoria (Australia), population of 6.6 million with over 3 million separations per annum. Method: Diagnosis codes from ICD-10-AM were aggregated into Clinical Diagnosis Group (CDG) sets according to clinical features and associated risk of in-hospital death and complications. Residual codes were excluded. Administrative data from July 2020 to June 2023 were interrogated to ascertain frequency of diagnoses captured by CDG sets. Results: 12,716 (87.9%) of 14,470 total ICD-10-AM codes were aggregated into 406 CDG sets; mean 32 (range 1-288) codes per set. One thousand seven hundred fifty-three (12.1%) were excluded (not allocated): 775 (5.4%) residual codes; 702 (4.9%) indicating reason for healthcare encounter; and 276 (1.9%) ill-defined clinical symptom codes. Over 36-months, 11.8 million separations were coded with 11,898 (82.2%) unique ICD-10-AM diagnoses, including 10,721 (90.1%) present in a CDG set. Of the 8571 (59.2%) codes associated with death or complications, 7813 (91.2%) were present in a CDG set. Conclusion: The CDG list provides a clinically meaningful method of categorisation and interrogating datasets based on ICD-10-AM and complements existing methods.