Identifying rare diseases using electronic medical records: the example of allergic bronchopulmonary aspergillosis

The purpose of the study is to evaluate whether primary care electronic medical records (EMRs) from patients with severe asthma can be used to identify allergic bronchopulmonary aspergillosis (ABPA) cases.

This cross-sectional feasibility study was conducted in adults with active and severe asthma registered with the Clinical Practice Research Datalink. A set of keywords flagged terms potentially indicative of ABPA in free-text comments of patients' EMRs to produce a grid on the basis of keywords' hit or miss. The grid was examined for occurrence and concurrence of keywords to discern patterns of concurrence potentially indicative of an underlying diagnosis of ABPA.

The analyses included 3 653 169 free-text items from 21 054 patients. In total, 52 patients (0.25%) had at least one mention of 'ABPA' in their medical record; 67% of these patients also had a mention of 'aspergillus/aspergillosis', 54% of 'bronchiectasis', 42% of 'itraconazole' and 62% of 'IgE'. The term 'aspergillus/aspergillosis' occurred with a proportion of 1.84% (N = 387); 9% of these patients also had a mention of 'ABPA', and the remaining 91% were potential additional cases of ABPA. From the observed concurrence of keywords, we were able to devise a potential algorithm to identify cases with varying degrees of specificity.

This study suggests that analysis of free text within asthmatic patients' EMRs may be used to identify potential cases of ABPA. This could be an efficient approach to identify rare conditions and to quantify their potential burden. © 2017 The Authors. Pharmacoepidemiology & Drug Safety Published by John Wiley & Sons Ltd.