Introduction:

The use of early screening questionnaires for chronic obstructive pulmonary disease (COPD) in primary health care could improve underdiagnosis. Several instruments are currently available, but there is scant information on their diagnostic performance.

Objective:

To determine the validity of different questionnaires for COPD diagnosis.

Materials and methods:

A systematic review and a meta-analysis of diagnostic test accuracy were carried out. A search of the literature published between July 1, 1997, and June 30, 2019 was performed in PubMed, EMBASE, and LILACS databases using MeSH and DeCS terms and the PICO strategy. Based on the inclusion and exclusion criteria, two reviewers selected the articles for complete analysis. Article quality was assessed using the QUADAS instrument.

Results:

19 articles were included for analysis. Overall results were: sensitivity: 68.1% (95%CI: 66.7% -69.4%); specificity: 64.9% (95%CI: 64.3-65.5); positive likelihood ratio: 2.024 (95%CI: 1.7152.388); negative likelihood ratio: 0.407 (95%CI: 0.289-0.573); and receiver operating characteristic area under the curve (ROC AUC): 0.75. The COPD-PS questionnaire reported the highest performance with sensitivity of 0.673 (95%CI: 0.653-0.692), specificity of 0.663 (95%CI: 0.65.5- 0.651), and ROC AUC of 0.750. It was followed by LFQ with sensitivity of 0.840 (95%CI: 0.806-0.871), specificity of 0.312 (95%CI: 0.289-0.336), and ROC AUC of 0.730. Finally, CDQ had sensitivity of 0.798 (95%CI: 0.764-0.829), specificity of 0.517 (95%CI: 0.495-0.538), and ROC AUC of 0.727.

Conclusion:

Clinical prediction instruments for COPD diagnosis have an acceptable performance. The COPD-PS, LFQ and CDQ questionnaires show a similar performance.

        · abstract in Spanish     · text in English     · English ( pdf )