Validation of diagnosis codes for pleural diseases and procedure codes for relevant respiratory procedures in a healthcare database in Hong Kong: a single tertiary centre study

INTRODUCTION: There are insufficient population-based epidemiological data on various pleural diseases in Hong Kong. We aimed to validate ICD-9-CM (International Classification of Diseases, Ninth Revision, Clinical Modification) codes for pleural diseases and relevant procedures prior to conducting epidemiological analyses using local electronic health records. METHODS: Hospitalisation episodes coded as 'pneumothorax', 'pleural effusion', and trauma-related pleural events, as well as procedures beginning with ICD-9-CM codes 33 and 34 between 2013 and 2022, were retrieved from the Hospital Authority. Paediatric patients and uninterrupted hospitalisation episodes were excluded. The cohort was filtered to include those hospitalised at Prince of Wales Hospital (PWH). Up to 50 hospitalisation episodes were randomly selected for manual validation. Positive predictive values (PPVs) with 95% confidence intervals of individual codes were calculated; successful validation was defined as a PPV ≥0.700. The primary endpoint was the PPV of individual diagnosis and procedure codes. RESULTS: A total of 26 757, 218 018, 1269, 185 154, and 106 450 hospitalisation episodes with non-traumatic pneumothorax, non-traumatic pleural effusion, trauma-related pleural events, procedures with code 33, and procedures with code 34, respectively, were retrieved. Within the PWH cohort, PPVs for these diagnosis and procedure codes were 0.853 (0.787-0.904), 0.928 (0.903-0.948), 0.957 (0.907-0.981), 0.932 (0.913-0.948), and 0.933 (0.916-0.948), respectively. Procedures involving indwelling pleural catheterisation and open drainage of the pleural cavity failed validation due to frequent miscoding. CONCLUSION: This is the first validation study of clinical codes for pleural diseases and related procedures in Hong Kong. All diagnosis codes and most procedure codes were successfully validated.