External validation of the PUMA COPD diagnostic questionnaire in a general practice sample and the PLATINO study population.

Background: A seven-item prescreening questionnaire (gender, age, pack-years smoking, dyspnea, sputum, cough, previous spirometry data) was developed for COPD detection in the primary care setting (PUMA Study) of four Latin America countries.

Objectives: To validate the PUMA prescreening COPD questionnaire externally in two different populations (primary care and general).

Methods: The PUMA prescreening COPD questionnaire score was applied to subjects from the Hospital Maciel, Montevideo (primary care), case-finding program and the PLATINO population (general) using PUMA study inclusion criteria. Post-bronchodilator FEV/FVC <0.70 and lower limit of normal (LLN) criteria were used to define COPD. Area under the received operator curve (ROC), sensitivity, specificity, predictive positive and negative values (PNV), number needed to treat (NNT), and best cut-points of the score were calculated.

Results: 974 individuals from Hospital Maciel and 2512 from the PLATINO population were eligible, using post-bronchodilator FEV/FVC <0.70, 45.1% and 18.7% had COPD, respectively, and using LLN 38.4% and 15.4% had COPD, respectively. From Hospital Maciel (post-bronchodilator FEV/FVC <0.70), the best cut-point of ≥6 had moderate discriminatory power (ROC 0.70), sensitivity 69.9%, specificity 62.1%, PNV 70.9%, and NNT of 3. The discriminatory power was 0.73 (ROC) in the PLATINO population with three potential cut-points (Youden's index): ≥3 (sensitivity 85.4%, specificity 46.9%), ≥4 (sensitivity 66.7%, specificity 66.5%), and ≥5 (sensitivity 51.5%, specificity 81.6%); the PNV at each cut-point was 93.3%, 89.9%, and 88.0%, respectively. The NNT was 5 for scores ≥3 and ≥4, and 4 for ≥5. The mean accuracy using the LLN for Hospital Maciel and PLATINO was 0.67 and 0.70, respectively.

Conclusion: External validation of the PUMA prescreening questionnaire in two Latin American populations (general and primary care) suggests moderate accuracy, similar to the original study in which the questionnaire was developed.