Accuracy of Electronic Health Record-Based Definitions for Patients with Heart Failure

Area of Science:

• Cardiology
• Health Informatics
• Epidemiology

Background:

• Electronic health records (EHRs) are widely used but lack standardized methods for identifying heart failure (HF) patients.
• Accurate identification of HF patients in EHR data is crucial for epidemiological studies and clinical care.

Purpose of the Study:

• To develop and validate multiple definitions for identifying patients with heart failure using EHR data.
• To assess the performance of these definitions in capturing true HF cases.

Main Methods:

• Developed eight distinct HF definitions combining ICD-10 codes and ejection fraction (EF) ≤ 40%.
• Validated definitions using stratified sampling and physician chart review against the Universal Definition of Heart Failure.
• Evaluated performance using sensitivity and positive predictive value.

Main Results:

• Identified over 41,000 patients meeting at least one EHR-based HF definition.
• Physician chart review of 528 sampled patients showed high positive predictive values (80.7%–98.6%) for the definitions.
• Sensitivities ranged from 10.3% to 42.0%, indicating underestimation of HF prevalence.

Conclusions:

• EHR-based definitions can identify true heart failure cases but capture less than half of all HF patients.
• Current definitions severely underestimate HF prevalence, necessitating more comprehensive methods for epidemiological burden assessment.
• Improved EHR data utilization is needed to accurately understand the full scope of heart failure.