Use of electronic health record data mining for heart failure subtyping

Area of Science:

• Cardiology
• Medical Informatics
• Data Science

Background:

• Accurate heart failure (HF) subtyping is crucial for effective treatment and prognosis.
• Register-based HF subtyping traditionally relies on structured data, potentially missing nuanced information.
• Electronic health record (EHR) data offers a rich source of unstructured clinical information.

Purpose of the Study:

• To evaluate the efficacy of text mining of EHR data for improving register-based heart failure (HF) subtyping.
• To assess the accuracy of algorithm-derived ejection fraction (EF) in classifying HF subtypes.
• To compare survival analyses based on traditional HF diagnosis versus algorithm-based subtyping.

Main Methods:

• Text mining of unstructured EHR data from 43,405 individuals to identify ejection fraction (EF) mentions.
• Validation of algorithm-identified EF against clinical assessment in 200 randomly selected patients.
• Integration of structured laboratory data for HF subtyping into HF with mildly reduced EF (HFmrEF), HF with preserved EF (HFpEF), and HF with reduced EF (HFrEF).

Main Results:

• The algorithm correctly assigned EF to the appropriate HF subtype range in 86% of cases.
• High sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were observed for HFrEF and HFmrEF.
• Mortality risk increased progressively from HFmrEF to HFpEF to HFrEF, consistent with traditional diagnoses.

Conclusions:

• Quantitative ejection fraction (EF) data can be efficiently extracted from EHRs using text mining.
• EHR data combined with laboratory data enables reasonably accurate HF subtyping, particularly for HFrEF.
• The developed method shows potential for enhancing HF classification and epidemiological research.