Identifying QT prolongation from ECG impressions using natural language processing and negation detection

Area of Science:

• Cardiology
• Medical Informatics
• Natural Language Processing

Background:

• Electrocardiogram (ECG) interpretations are vital for clinical decisions and research.
• QT prolongation is a significant risk factor for sudden cardiac death.
• Automated corrected QT (QTc) calculations by ECG machines may have limitations.

Purpose of the Study:

• To evaluate a natural language processing (NLP) system for identifying QT prolongation in ECG impressions.
• To compare NLP-identified QT prolongation with automated QTc calculations.
• To assess the utility of NLP for clinical decision support and research.

Main Methods:

• Integrated a negation tagging algorithm into the KnowledgeMap concept identifier (KMCI).
• Applied KMCI to 44,080 ECG impressions to identify Unified Medical Language System concepts.
• Compared QT prolongation instances identified by KMCI with calculated QTc values.

Main Results:

• The negation detection algorithm achieved high recall (0.973) and precision (0.982).
• A concept query for QT prolongation matched 2,364 ECGs with 1.00 precision.
• Automated QTc cutoffs showed low positive predictive values (6-21%), with 96% of discrepancies explained by miscalculations.

Conclusions:

• NLP systems can effectively identify QT prolongation from ECG impressions.
• This approach enhances decision support and clinical research capabilities.
• NLP offers a more reliable method for detecting cardiac conditions compared to automated QTc calculations.