Asthma in electronic health records: validity and phenotyping

Abstract

This PhD thesis explores the validation of asthma in electronic health records (EHR) and the characteristics of asthma phenotypes in the UK using CPRD GOLD, HES and ONS data. The absence of a universal case definition, the overlap with other diseases and the incomplete recording of diagnostic markers makes the identification of asthma patients in EHR challenging. Furthermore, asthma phenotypes have previously been established based on cluster analysis in small populations, but their prevalence, treatment and outcomes in the general population have not been investigated. Firstly, I conducted a systematic review to understand how past epidemiological studies have validated asthma recording in EHR, including a critical appraisal and list of test measure values for the selected studies. Secondly, I validated algorithms to reliably ascertain the asthma status of patients in CPRD GOLD. This validation study identified multiple algorithms with PPV greater than 80%. The most practical algorithm (presence of a specific asthma diagnostic code) had a PPV of 86.4 (95% CI:77.4-95.4). Thirdly, I quantified the concomitant occurrence of asthma in COPD patients and vice versa in CPRD GOLD. After detailed case review, concomitant asthma and COPD was concluded in 14.8% of validated asthma patients and in 14.5% of validated COPD patients. However, asthma diagnoses may be unreliable in COPD patients, as over 50% of COPD patients had received an asthma code. Finally, I examined the prevalence, treatment, outcomes and characteristics of established asthma phenotypes in CPRD GOLD. Only a minority of patients (37.3%) were classified into these phenotypes using stringent inclusion criteria. Exacerbation rates/1000PY were lowest for benign asthma (106.8 [95% CI:101.2-112.3]), and highest for obese non-eosinophilic asthma (469.0 [95% CI:451.7-486.2]). In conclusion, this thesis provides information on the validation of asthma diagnoses in EHR and the prevalence, treatment, outcomes of predefined asthma phenotypes in the UK primary care population.