Using Artificial Intelligence With Natural Language Processing to Combine Electronic Health Record's Structured and...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Public Health Surveillance

Background:

• Nonvalvular atrial fibrillation (NVAF) affects millions, increasing stroke risk but often remains undiagnosed and undertreated.
• Current diagnostic and treatment guidelines for NVAF are not fully implemented, leading to preventable adverse events.

Purpose of the Study:

• To evaluate if combining semisupervised natural language processing (NLP) with structured electronic health record (EHR) data enhances the discovery and treatment of NVAF.
• To assess the potential of this integrated approach to prevent strokes, reduce mortality, and lower healthcare expenditures.

Main Methods:

• Utilized NLP to index clinical notes from 96,681 participants in the University of Buffalo faculty practice EHR.
• Compared the identification of NVAF and related risk scores (CHA2DS2-VASc, HAS-BLED) using structured data alone versus a combination of structured and unstructured (free-text) EHR data.
• Analyzed large-scale databases (Optum, Truven) with over 63 million participants to determine NVAF prevalence, treatment rates, and outcomes in untreated populations.

Main Results:

• The structured-plus-unstructured method identified 3,976,056 additional true NVAF cases (P<.001), significantly improving sensitivity for NVAF detection.
• This integrated approach demonstrated a 32.1% improvement in identifying risk factors (CHA2DS2-VASc, HAS-BLED scores) compared to structured data alone (P=.002 and P<.001).
• Extrapolated to the US, this method could prevent an estimated 176,537 strokes, save 10,575 lives, and save over $13.5 billion annually.

Conclusions:

• AI-informed bio-surveillance, integrating NLP with structured EHR data, substantially improves data completeness for NVAF identification.
• This advanced method offers a powerful strategy to prevent thousands of strokes, save lives, and reduce healthcare costs.
• The applicability of this approach extends to numerous other disorders with significant public health implications.