OBJECTIVES: The validation of diagnoses is particularly important in health care database studies, where the use of codes to identify outcomes of interest can lead to potential misclassification. The aim of this study was to validate records of treated COPD (Chronic Obstructive Pulmonary Disease) registered in the SIDIAP primary care Database.

METHODS: A random sample of 1,000 COPD patients identified using pre-specified ICD-10 codes registered between birthdate to 31^st December 2016 and treated with LAMA/LABA either in fixed or free combination in SIDIAP (www.sidiap.org) was selected. Patients were considered to have COPD if it was confirmed by recorded spirometry (post-bronchodilator FEV1/FVC < 0.70). If no spirometry data was available, specialist confirmation in linked hospital data or code/s for clinical symptoms (dyspnea, chronic cough or sputum production) and 2+ prescriptions of bronchodilator/s within one year was required. Finally, where none of the above were present, free text was used: a GP reviewed manually the GP records within 3 months before/after COPD diagnosis.

RESULTS: Out of the 1,000 patients with COPD disease codes and treated with LAMA/LABA, 29 had no free text available and were therefore excluded, leaving 971 patients for validation. A total of 709 (73.0%) had COPD confirmed by spirometry and an additional 104 (10.7%) by the combination of symptoms and bronchodilator/s use. A total of 158 patients had their GP record reviewed manually, of which 68 were confirmed to have COPD. The overall PPV for COPD diagnosis was 90%.

CONCLUSIONS: