Background The Clinical Practice Research Datalink (CPRD) is a UK database of primary care health records covering 11 million residents of England and Wales, including at least 200,000 COPD patients. We have recently validated both the recording of COPD and algorithms to identify acute exacerbations of COPD (AECOPD) treated in primary care. It is unclear if primary care records alone can be used to identify hospitalisations for AECOPD. We aimed to validate strategies for identifying hospitalisations for acute exacerbations of COPD (AECOPD) using CPRD.

Methods We identified 22,599 patients with a validated diagnosis1 of COPD who had HES data linked to CPRD. We assessed the positive predictive value (PPV) and sensitivity of four strategies to identify hospitalisations for AECOPD using CPRD: 1) AECOPD hospitalisation code; 2) AECOPD identified using our validated algorithm; 3) generic hospitalisation code; or 4) AECOPD identified using our validated algorithm and generic hospitalisation code on the same day. We identified hospitalisations for AECOPD in HES using ICD codes, and used HES identified AECOPD hospitalisation as the reference standard. We used ICD-10 codes J44.0 and J44.1 in any position and J44.9 in first position to identify hospitalisations for AECOPD in HES. We searched primary care records over a 30 day window after a record for hospitalisation for AECOPD for recording consistent with AECOPD hospitalisation defined by the four strategies. Patients were followed up between January 2004 and July 2013. As a sensitivity analysis, we repeated the analysis using a more specific definition of hospitalisation for AECOPD (J44.0 or J44.1 in first position only).

Results 19,507 hospitalisations for AECOPD were identified based on HES during the study period. The PPV and sensitivity of the various strategies to identify hospitalisations for AECOPD from CPRD alone are presented in Table 1. Sensitivity analysis did not significantly change the results.