Stephanie Helman, CCRN-K, CNS, MSN, RN
Predoctoral Research Fellow
University of Pittsburgh
Pittsburgh, Pennsylvania
Disclosure information not submitted.
Martha Ann Terry, PhD
Professor
University of Pittsburgh Graduate School of Public Health
Pittsburgh, Pennsylvania, United States
Disclosure information not submitted.
.jpg "Marilyn Hravnak, PhD, RN photo")
Marilyn Hravnak, PhD, RN
Professor Emeritus
University of Pittsburgh
Pittsburgh, Pennsylvania
Disclosure information not submitted.
Pellathy Tiffany, PhD, RN, ACNP
Postdoctoral Fellow
Veterans Administration Center for Health Equity Research and Promotion
Pittsburgh, Pennsylvania, United States
Disclosure information not submitted.
Betsy George, CCRN, MSN, PhD, RN
Programmatic Nurse Specialist
University of Pittsburgh Medical Center, United States
Disclosure information not submitted.
Michael Pinsky, MD, Dr hc,MCCM
Professor Critical Care Medicine, Bioengineering, Cardiovascular Disease and Anesthesiology
University of Pittsburgh School of Medicine, United States
Disclosure information not submitted.
Salah Al-Zaiti, PHD, RN, ANP-BC, FAHA
Associate Professor
University of Pittsburgh School of Nursing, United States
Disclosure information not submitted.
Gilles Clermont, MD, MS
Professor CCM, Mathematics, Clinical Translational Sci and IE
VA Pittsburgh Health System
Pittsburgh, Pennsylvania
Disclosure information not submitted.
Title: User-Engaged Design of a Graphical User Interface for Instability Decision Support in the ICU
INTRODUCTION/HYPOTHESIS: Critical instability forecast and treatment can be optimized by machine learning (ML)-facilitated clinical decision support (CDS). We previously used ML to develop an instability risk score from continuous ICU vital sign (VS) data, and identified physiologic explanations for the score. We are now developing a graphical user interface (GUI) to present instability forecast, explanations, and treatment recommendations for CDS at the bedside. We propose using focus groups to employ clinician driven iterative design of the GUI.
Methods: We recruited ICU clinicians to participate in focus groups regarding instability risk forecast, physiologic status, and action recommendations presented in a prototype bedside GUI. Six online focus group sessions were held in two rounds, each moderated by a focus group methodologist. Each round consisted of three groups (nurses, providers [APP and physicians], hybrid group). Iterative design changes were made, and the modified GUI design presented in the next round. Focus groups were recorded and transcribed, and the de-identified transcripts were independently coded by three researchers. After resolving coding discrepancies, codes were coalesced into emerging themes.
Results: 23 clinicians were recruited (11 RN, 2 NP, 1 PA, 9 MD). Six themes emerged from the focus groups: 1) analytics transparency, 2) graphical interpretability, 3) impact on practice, 4) value of trend synthesis of dynamic patient data, 5) decisional weight (weighing ML output during decision making), and 6) location (usability, concerns for patient/family GUI view). Nurses emphasized having GUI objective information to support communication with providers, and optimal GUI location. Providers emphasized need for interpretability of recommendations, and concern for impairing trainee critical thinking. Both groups valued synthesized views of VS, interventions, and risk trends, but were skeptical of placing decisional weight on ML output until proven trustworthy in testing and practice. Thematic saturation was achieved, and feedback informed three iterative GUI design versions, with substantive changes in each version.
Conclusions: User-engaged iterative design is useful in adjusting prototype GUI presentation, and may lead to enhanced clinical usability.
Funding: R01GM117622, F31NR019725