Enhanced Detection of Heart Valve Disease Using Integrated Artificial Intelligence at Scale.

Area of Science:

• Cardiology
• Health Informatics
• Medical Software Development

Background:

• Undertreatment of heart valve disease poses significant patient risks.
• Fragmented healthcare data systems hinder effective management of heart valve disease.
• A novel software solution is proposed to address these challenges by leveraging electronic health records.

Purpose of the Study:

• To develop and evaluate a rules-based software algorithm for detecting heart valve disease.
• To assess the software's ability to identify moderate or severe heart valve disease.
• To measure disease progression and compliance with care pathways.

Main Methods:

• A software platform integrated electronic health records and other data sources.
• A rules engine was programmatically applied to search structured and unstructured data.
• Guideline-reported values were used for identifying moderate or severe heart valve disease.

Main Results:

• The software identified moderate or severe aortic stenosis in 9.1% of 60,145 patients across two US institutions.
• The algorithm demonstrated high precision (92.9%) and accuracy (98.6%) in detecting moderate or severe aortic stenosis.
• Thirty-five percent of patients with moderate stenosis progressed to severe stenosis, and 70.3% lacked timely follow-up.

Conclusions:

• A rules-based software program significantly improves the detection of heart valve disease.
• The software facilitates the measurement of disease progression and adherence to care pathways.
• This technology offers a scalable solution for managing heart valve disease within healthcare systems.