Predicting the costs of managing patients with chronic obstructive pulmonary disease

Abstract

The economic consequences of chronic obstructive pulmonary disease (COPD) are considerable, although the factors that best predict costs are largely unknown. This study used a population-based cohort to identify the clinical factors during an index year that were most predictive of increased direct medical costs in the subsequent year, and to develop a predictive model that described the cost variations in COPD. The medical records of 2116 patients enrolled in one regional health system who had COPD and health-care resource utilisation data for 1998 and 1999, were abstracted for information about symptoms, smoking history, chronic illnesses, and pulmonary function data. All inpatient, outpatient and pharmacy utilisation data for each subject for 1999 were extracted from the database. Total costs for each individual were transformed to a log scale. Potential causes of cost variability (predictor variables) were defined and classified into sets (or domains). Multiple linear regression models were fitted for each domain. The study demonstrated that severity of airflow obstruction, as assessed by FEV(1)% predicted, is a significant but weak predictor of future health-care resource utilisation-prior hospitalisation and home oxygen use, the presence of comorbid conditions and symptoms of dyspnoea are better predictors of costs. Those interested in the economic benefits of new COPD treatments and disease management programs need to carefully account for these factors.