Finding the Clinical Traces of Cognitive Impairment in Patients with Heart Failure: A Natural Language Processing Study of Clinical Letters from Routine Care

Cognitive impairment is common in patients with heart failure, but to which extent cognitive complaints are evaluated and listed in clinical practice is unknown. Therefore, this study aims to identify whether cognitive complaints are listed in clinical notes of patients with heart failure, consistent with listed complaints in clinical notes of patients attending memory clinics, by using natural language processing (NLP) techniques. Patients with heart failure and patients attending a memory outpatient clinic were identified by using echocardiography reports and presence of memory outpatient clinic codes stored in the Utrecht Individual-Oriented Database (UPOD) from 2011 to 2023. Named Entity Detection and Linking (NER+L) strategies MedCAT and MedCATTrainer were used to extract listed complaints in clinical notes by patient group, and it was assessed whether cognitive complaints were listed in clinical notes of patients with heart failure. Among 5803 patients with heart failure, dyspnea (57.1%), chest pain (48.4%), and oedema (43.6%) were the most listed complaints. In 967 patients attending memory clinics, memory problem (80.9%), getting lost (24.1%), and being morose (22.5%) were the most listed complaints. Notably, in patients with heart failure, the reporting of memory problems was low at 2.6%. This study shows a low reporting frequency of cognitive complaints in clinical notes of patients with heart failure, even though both conditions often co-occur according to cross-sectional studies. This points towards a potential underrecognition of cognitive complaints in patients with heart failure during clinical practice.