[Optimal search conditions for identification of patients with chronic myeloid leukemia from Japanese National Health Insurance data].

This study investigated which conditions could be used to identify patients with chronic myeloid leukemia (CML) from a National Health Insurance claims dataset. During April 2012 and September 2018, 1,789,462 employees were enrolled in the dataset for Shizuoka Prefecture residents. The number of patients with the ICD-10 code for CML was 761. Among them, 246 who had been prescribed a tyrosine kinase inhibitor were considered as having true CML. The positive predictive value was calculated as 32.3% when CML was identified by ICD-10 code alone. Combination of ICD-10 code with prescribed drugs was required to accurately identify patients with CML from the insurance database.