OAC

OAC performs semantic natural language processing of electronic health record text to extract oral anticoagulant prescription patterns and compute CHA2DS2-VASc and HAS-BLED risk scores for atrial fibrillation patients.

Key Features:

• Natural Language Processing Pipeline: Employs an NLP pipeline to identify risk factors for atrial fibrillation patients by extracting relevant information from discharge summaries.
• Risk Scoring Calculation: Computes CHA2DS2-VASc and HAS-BLED scores from text-extracted risk factors.
• Validation Against Experts: Validated calculated scores against two independent clinical experts with average kappa 0.78 for CHA2DS2-VASc and 0.54 for HAS-BLED.
• Trend Analysis: Analyzes temporal trends demonstrating a significant increase in oral anticoagulant prescriptions among high-risk AF patients linked to guideline changes and the introduction of direct oral anticoagulants (DOACs).
• Factors Influencing OAC Use: Identifies factors independently associated with OAC use, including components of CHA2DS2-VASc and HAS-BLED, discharging specialty, and patient frailty, with cardiology departments showing the highest OAC rates.

Scientific Applications:

• Clinical Decision Support: Automates risk score calculations to support anticoagulation decision-making for atrial fibrillation patients.
• Epidemiological Research: Enables analysis of large EHR datasets to replicate registry findings and study trends in clinical practice.
• Healthcare Policy Analysis: Provides insights into prescription patterns and risk factor associations to inform guidelines and policy for AF management.

Methodology:

Processes discharge summaries from 1 January 2011 to 1 October 2017 for 10,030 atrial fibrillation patients using an NLP pipeline to extract clinical information, compute CHA2DS2-VASc and HAS-BLED scores, and validate them against two independent clinical expert assessments.