Background: Administrative data is often used to identify patients with chronic obstructive pulmonary disease (COPD), yet the validity of this approach is unclear. We sought to develop a predictive model utilizing administrative data to accurately identify patients with COPD.
Methods: Sequential logistic regression models were constructed using 9573 patients with postbronchodilator spirometry at two Veterans Affairs medical centers (2003-2007). COPD was defined as: 1) FEV1/FVC <0.70, and 2) FEV1/FVC < lower limits of normal. Model inputs included age, outpatient or inpatient COPD-related ICD-9 codes, and the number of metered does inhalers (MDI) prescribed over the one year prior to and one year post spirometry. Model performance was assessed using standard criteria.
Results: 4564 of 9573 patients (47.7%) had an FEV1/FVC < 0.70. The presence of ≥1 outpatient COPD visit had a sensitivity of 76% and specificity of 67%; the AUC was 0.75 (95% CI 0.74-0.76). Adding the use of albuterol MDI increased the AUC of this model to 0.76 (95% CI 0.75-0.77) while the addition of ipratropium bromide MDI increased the AUC to 0.77 (95% CI 0.76-0.78). The best performing model included: ≥6 albuterol MDI, ≥3 ipratropium MDI, ≥1 outpatient ICD-9 code, ≥1 inpatient ICD-9 code, and age, achieving an AUC of 0.79 (95% CI 0.78-0.80).
Conclusion: Commonly used definitions of COPD in observational studies misclassify the majority of patients as having COPD. Using multiple diagnostic codes in combination with pharmacy data improves the ability to accurately identify patients with COPD.