Rationale

Chronic obstructive pulmonary disease (COPD) is associated with an increasing mortality worldwide. COPD is classified according to the airflow limitation and the symptoms/risk of exacerbation. In a recent review, a new etiological classification for COPD was proposed including 5 subtypes: genetics, early-life events, infection, smoking and environmental exposure.

Aim

To validate the newly proposed types of COPD in a Swiss population of patients with COPD.

Methods

Electronic reports were retrospectively extracted from 211 consecutive patients with COPD who underwent cardio-pulmonary exercise testing at our institution between Jan. 2019 and Jul. 2022. Text information included in clinical reports was matched with the text description of the new COPD types using text mining. Term usage frequency was analyzed using correspondence analysis and each patient was classified to a COPD subtype.

Results

Fig. 1 depicts the correspondence analysis of the 5 COPD subtypes. Individual clinical reports are shown as supplementary entries. Our COPD cohort was mostly classified as Infection and Smoking types (38% and 37% respectively). Early life event types represented 19% of the population, whereas environmental and genetic types corresponded to 4% and 2%, respectively. There were important overlaps among COPD types.

Conclusions

Using AI-based text mining of clinical reports, we were able to identify newly proposed COPD types in a cohort of patients with COPD.