Enhancing ICD-Code-Based Case Definition for Heart Failure Using Electronic Medical Record Data.

Area of Science:

• Medical Informatics
• Clinical Epidemiology
• Health Services Research

Background:

• Accurate heart failure (HF) patient identification is crucial for surveillance and outcomes research.
• Existing International Classification of Diseases (ICD) code algorithms have low sensitivity for HF detection in administrative data.
• Electronic medical records (EMRs) offer detailed clinical information to improve HF identification algorithms.

Purpose of the Study:

• To enhance the accuracy of HF case identification by improving the ICD algorithm.
• To incorporate comprehensive EMR data into HF definition algorithms.
• To compare the performance of various algorithms against a gold standard medical chart review.

Main Methods:

• Utilized a population-based cohort of 2106 inpatients in Calgary, Alberta, Canada.
• Compared the standard ICD algorithm with EMR-based algorithms: keyword search, machine learning-based HF concept (HFC), and structured data algorithms.
• Evaluated algorithm performance using medical chart review as the gold standard, assessing sensitivity and positive predictive value (PPV).

Main Results:

• The ICD algorithm showed high PPV (92.4%) but low sensitivity (57.4%).
• The HFC algorithm achieved 80.0% sensitivity and 88.9% PPV, outperforming the ICD and keyword algorithms.
• Combining HFC and ICD algorithms yielded 83.3% sensitivity and 83.3% PPV, demonstrating improved case capture.

Conclusions:

• Natural language processing and machine learning on inpatient EMR discharge summaries significantly improve HF case identification over traditional ICD algorithms.
• The HF concept (HFC) algorithm provides a straightforward and effective method for identifying HF cases.
• Enhanced algorithms utilizing EMR data are vital for accurate HF surveillance and outcomes studies.