Development of a chronic obstructive pulmonary disease severity classification system using a Japanese health insurance claims database

Abstract Background: Healthcare services provided to patients should vary depending on disease severity. However, disease severity bias, a type of selection bias, is a commonly encountered problem in administrative database studies. Herein, we selected chronic obstructive pulmonary disease (COPD), which commonly affects elderly Japanese citizens, for the development and validation of a severity classification system based on a health insurance claims database. Methods: Patients who received COPD-related diagnostic codes in 2011 were selected from a commercially based health insurance claims database. COPD patients were randomly divided into two groups to develop and validate severity scores. A principal component analysis was used to estimate factor loadings used to weight calculations of COPD severity scores. Score validity was evaluated using a linear trend test to predict COPD treatment costs and acute exacerbation events. Results: Using records from 880 patients, ten variables were created: acute exacerbation events, emphysema diagnoses, laboratory test and oxygen therapy procedures, prescribed anticholinergic, inhaled corticosteroid (ICS), short acting beta-agonist, and long acting bronchodilator (LABA) agents, asthma diagnosis and patient birth years. Factor loadings from LABA and ICS prescriptions had the strongest impacts on estimated severity scores (0.50 and 0.49, respectively). Among 300 validation group patients, scores were found to associate with increasing trends of median costs and exacerbation risks (p for trend < 0.05). Conclusions: Estimated severity scores would help to predict COPD-related medical costs and exacerbation events. For further clinical implementation, this classification system should be re-evaluated using clinical lung functions information indicative COPD severity and treatment choices.