EHR-Oriented Knowledge Graph System: Toward Efficient Utilization of Non-Used Information Buried in Routine Clinical...

Area of Science:

• Health Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support Systems

Background:

• Clinicians often overlook clinical information outside their specialty, leading to diagnostic delays and suboptimal patient management.
• Unused clinical data within electronic health records (EHRs) negatively impacts healthcare quality and patient outcomes.
• Effective utilization of comprehensive EHR data is crucial for improving diagnostic accuracy and treatment efficacy.

Purpose of the Study:

• To develop and evaluate an EHR-oriented knowledge graph system for leveraging neglected clinical information.
• To enhance clinical decision-making by identifying important findings within EHR data that may be overlooked by non-specialist clinicians.
• To assess the system's effectiveness in detecting underdiagnosed conditions, such as chronic kidney disease (CKD), in non-nephrology departments.

Main Methods:

• Transformed EHR data into a semantic, patient-centralized information model using a knowledge graph ontology.
• Developed a reasoning engine with semantic rules to identify critical clinical findings and create EHR data trajectories.
• Utilized a graphical reasoning pathway to visualize and explain the clinical significance of neglected information to clinicians.

Main Results:

• The system identified 2,774 patients meeting CKD diagnostic criteria and 10,377 requiring high attention among 71,679 patients in non-nephrology departments.
• A follow-up study confirmed CKD in 82.1% of diagnosed patients and 61.4% of high-attention patients.
• The knowledge graph approach proved feasible and effective for clinical information utilization and provided interpretable recommendations.

Conclusions:

• The EHR-oriented knowledge graph system effectively utilizes neglected clinical information, improving healthcare quality.
• The system acts as an explainable artificial intelligence tool, providing interpretable recommendations for comprehensive clinical decision-making.
• This approach enhances the identification and management of conditions like CKD among non-specialist physicians, demonstrating significant clinical value.