A common data model for the standardization of intensive care unit medication features

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Data Standardization

Background:

• Common data models (CDMs) are crucial for artificial intelligence (AI) applications but have not been established for intensive care unit (ICU) medications.
• Standardizing ICU medication data is essential for advancing AI-driven clinical decision support and improving patient outcomes.

Purpose of the Study:

• To develop a comprehensive common data model (CDM) for intensive care unit (ICU) medications.
• To standardize medication features essential for supporting future AI initiatives in the ICU.

Main Methods:

• A 9-member multidisciplinary team of ICU clinicians and AI experts utilized a 5-round modified Delphi process.
• Methods included conference calls, web-based communication, and electronic surveys to identify and score medication features.
• Features were evaluated for clinical relevance, feasibility of collection, and adherence to Findable, Accessible, Interoperable, and Reusable (FAIR) principles.

Main Results:

• An initial list of 889 ICU medications yielded 106 potential medication features.
• 71 features were identified as core components for the CDM, categorized into drug product-related (43 features) and clinical practice-related (63 features).
• Standardized definitions and suggested response values were established for each feature, with the CDM made available online.

Conclusions:

• The developed CDM for ICU medications is a foundational step for AI research in critical care.
• This initiative aims to enhance the use of AI for improving patient outcomes in the ICU.
• Ongoing engagement and refinement of the CDM are necessary for its continued success and evolution.