Anomaly Detection in Electronic Health Records Across Hospital Networks: Integrating Machine Learning With Graph...

Area of Science:

• Health Informatics
• Data Science
• Network Analysis

Background:

• Electronic Health Records (EHRs) are crucial for clinical decision-making in hospital systems.
• The reliability of Health Information Technology (HIT) systems is vital for patient safety.
• Existing anomaly detection methods for EHR data often focus on single hospitals, limiting system-wide insights.

Purpose of the Study:

• To develop a novel approach for detecting anomalies in EHR data across a network of hospitals.
• To improve the understanding of system-wide anomalous events and their impact on patient safety.
• To create a scalable tool for swift identification and response to deviations in EHR data.

Main Methods:

• Combined advanced machine learning models (five distinct models) with graph algorithms.
• Developed a tool to represent detected anomalies as graphs for pattern recognition across the hospital network.
• Utilized real-world data for extensive testing and validation.

Main Results:

• The proposed approach effectively identifies anomalies spanning multiple medical facilities, indicating potential system-level risks.
• Demonstrated actionable insights superior to existing methods.
• The system proved to be scalable for seamless integration into existing HIT infrastructures.

Conclusions:

• The novel approach enhances the detection of system-wide EHR anomalies, improving patient safety across hospital networks.
• This method offers a robust and scalable solution for HIT system reliability.
• The integration of machine learning and graph algorithms provides a powerful tool for proactive risk management in healthcare.