Original Research
Saturday, November 11, 2023
11:30 AM–01:00 PM
Abstract
Introduction: Artificial intelligence (AI) can be used to improve critically-ill patient outcomes, and the development of a common data model (CDM) is required for seamless data exchange among various AI applications. Creation of this CDM aims to standardize drug and clinical features for longitudinal use in clinical decision support systems (CDSS) and pharmacotherapy optimization in the intensive care unit (ICU).
Research Question or Hypothesis: Is creating a standardized CDM of ICU medications feasible for supporting clinical decision-making software?
Study Design: Retrospective cohort study
Methods: An expert panel comprised of nine critical care pharmacists participated in a five-round modified Delphi process to compile medication and clinical features for a CDM involving ICU medications. A list of drug formulations based on the medication regimen complexity-intensive care unit (MRC-ICU) scoring tool was derived from the electronic health record of adult patients admitted to an institution’s ICU from 2015 to 2020. The primary outcome was to develop features for inclusion in the CDM. The secondary outcome was to utilize drug information resources to input data in all features of the CDM for each drug formulation.
Results: The panel finalized a list of 1,463 drug formulations, including 332 unique drugs, with 889 formulations pertaining to the MRC-ICU score and 574 formulations related to common ICU medications. Eighty-seven drug formulation features were defined by the expert panel, which were then split into subcategories: medication clinical features (n=73) and drug product features (n=14). The finalization of CDM features satisfied the primary outcome, and coding of the 1,463 drug formulations fulfilled the secondary outcome.
Conclusion: A CDM that analyzes MRC-ICU related drug formulations was developed and has the potential to enhance an AI program’s feedback and prediction of various events in the ICU. Without an established CDM, use of AI in CDSS is limited. Future research should validate this CDM and gauge its impact on AI applications.
Presenting Author
Kara E. Phillips PharmD CandidateThe University of Georgia College of Pharmacy
Authors
Merrie Barnett-Brock PharmD Candidate
The University of Georgia College of Pharmacy
Amber D. Fraley PharmD Candidate
The University of Georgia College of Pharmacy
Liana Ha BSCh, PharmD Candidate
The University of Georgia College of Pharmacy
Alexander Durant BS, PharmD Candidate
The University of Georgia College of Pharmacy
Ciana Wallace Pharm.D. Candidate
University of Georgia
Andrea Sikora Newsome PharmD, MSCR, BCCCP, FCCM
Augusta University Medical Center/UGA College of Pharmacy
Susan E. Smith PharmD, BCPS, BCCCP, FCCM
University of Georgia College of Pharmacy