Machine Learning-Enabled Clinical Information Systems Using Fast Healthcare Interoperability Resources Data...

Area of Science:

• Health Informatics
• Machine Learning in Healthcare
• Clinical Information Systems

Background:

• Machine learning-enabled clinical information systems (ML-CISs) are crucial for advancing healthcare delivery and research.
• The Fast Healthcare Interoperability Resources (FHIR) standard is increasingly adopted in ML-CIS development.
• Current methods for applying FHIR to ML-CISs exhibit variability.

Purpose of the Study:

• To evaluate and compare existing FHIR-based ML-CIS functionalities, strengths, and weaknesses.
• To propose guidelines for optimizing the development of future ML-CISs.
• To identify best practices for integrating machine learning with FHIR standards in clinical settings.

Main Methods:

• A systematic literature search was conducted across Embase, PubMed, and Web of Science.
• Articles describing FHIR-compliant machine learning systems for clinical data analytics or decision support were analyzed.
• Systems were compared based on functionality, data sources, formats, security, performance, resource needs, and scalability.

Main Results:

• 39 articles on FHIR-based ML-CISs were categorized into decision support (n=18), data management (n=10), and auxiliary modules (n=11).
• Strengths included novel cloud system applications, Bayesian networks, visualization, and unstructured data translation to FHIR.
• Identified weaknesses included lack of electronic health record interoperability and external validation of clinical efficacy.

Conclusions:

• Current ML-CIS shortcomings can be mitigated through modular, interoperable data management and analytics platforms.
• Secure interinstitutional data exchange and scalable APIs are essential for real-time and prospective clinical applications.
• Optimizing ML-CISs requires addressing diverse electronic health record platform implementations and ensuring robust interoperability.