Transdisciplinary team develops framework to ensure ethical, transparent AI use in critical care environments
Delirium is a common but often undiagnosed condition in intensive care unit (ICU) patients. To improve early detection, a transdisciplinary team has developed an artificial intelligence (AI) tool that analyzes facial expressions and movement data to identify behaviors linked to delirium severity. This technology aims to provide clinicians with passive, real-time feedback to support timely recognition and better-informed decisions.
The research team was led by Heidi Lindroth, PhD, R.N., a nurse scientist at Mayo Clinic and an affiliate scientist at Regenstrief Institute. Other members of the research team from Regenstrief included Andrew Gonzalez, M.D., J.D., MPH, and Malaz Boustani, M.D., MPH. Both Drs. Gonzalez and Boustani are also on faculty with the Indiana University School of Medicine.
Alongside the tool’s development, the team created an adaptive ethics roadmap to guide the responsible design and implementation of AI in critical care environments. Informed by Agile Science, it was developed with input from ICU survivors, caregivers, clinicians, ethicists and community representatives. The framework emphasizes transparency, accountability and patient-centeredness throughout the AI lifecycle.
“As we integrate AI into patient care, we must ensure it is not only technically sound but also ethically grounded,” said Dr. Gonzalez. “This roadmap offers a living, patient-informed guide to designing tools that are transparent and responsive to the people they serve.”
“AI in healthcare is evolving rapidly, but ethical considerations often lag behind,” said Dr. Lindroth. “Our goal is to involve patients, clinicians and communities from the start to build trust and ensure the technology reflects their needs and values.”
Insights gained from the design and testing of the AI delirium detection and severity measurement tool directly informed the development of the adaptive ethics framework. While created with ICU care in mind, the roadmap is flexible and applicable across diverse healthcare settings, providing practical guidance for ethical AI integration from concept to implementation.
Key components of the adaptive ethics framework include:
- Continuous stakeholder engagement throughout the AI lifecycle, involving patients, caregivers, clinicians and ethicists.
- Early identification of potential ethical risks during the design phase.
- Emphasis on transparency regarding AI functionality and data use.
- Iterative evaluation and adaptation as AI tools progress toward implementation.
- Accountability measures, including clear communication channels for feedback.
- Mechanisms to monitor and address unintended consequences.
- Flexibility to customize the framework for different clinical settings and evolving technologies.
- Maintaining a patient-centered approach as a core principle.
Next steps for the adaptive ethics framework include refining and validating it through real-world use across diverse healthcare settings. The team will collaborate with clinical partners to test the roadmap with new AI tools beyond ICU delirium detection, ensuring its ongoing relevance. Continued engagement with patients, caregivers, clinicians and ethicists will help adapt the framework to evolving technologies and ethical challenges. The team also plans to create educational resources and toolkits to support effective application, fostering transparency, access to care and accountability throughout the AI lifecycle.
“Applying an Agile Science Roadmap to Integrate and Evaluate Ethical Frameworks Throughout the Lifecycle and Use of Artificial Intelligence Tools in the Intensive Care Unit,” is published in Critical Care Nursing Clinics of North America.
Authors and affiliations as listed in the publication
Heidi Lindroth1, Juhi Sahajwani2, Mark Hudson3, Laura Heier4, Andrew A Gonzalez5, Anirban Bhattacharyya2, Zhi Zheng6, Malaz Boustani7, Vitaly Herasevich8, Michelle McGowan9, Barbara Barry10
1Division of Nursing Research, Department of Nursing, Mayo Clinic, Rochester, MN, USA; Center for Aging Research, Regenstrief Institute, School of Medicine, Indiana University, Indianapolis, IN, USA; Center for Health Innovation and Implementation Science, School of Medicine, Indiana University, Indianapolis, IN, USA. Electronic address: Lindroth.heidi@mayo.edu.
2Department of Critical Care Medicine, Mayo Clinic, Jacksonville, FL, USA.
3School of Psychology and Counselling, The Open University, Milton Keynes, UK.
4Department of Graduate Nursing, Viterbo University, La Crosse, WI, USA; Department of Neurosurgery, Mayo Clinic, Rochester, MN, USA.
5Center for Health Services Research, Regenstrief Institute, School of Medicine, Indiana University, Indianapolis, IN, USA; Surgical Outcomes and Quality Improvement Center, Department of Surgery, School of Medicine, Indiana University, Indianapolis, IN, USA.
6Department of Electrical Engineering, University of Notre Dame, Notre Dame, IN, USA.
7Center for Aging Research, Regenstrief Institute, School of Medicine, Indiana University, Indianapolis, IN, USA; Center for Health Innovation and Implementation Science, School of Medicine, Indiana University, Indianapolis, IN, USA.
8Department of Anesthesiology and Perioperative Medicine, Mayo Clinic, Rochester, MN, USA.
9Biomedical Ethics, Departments of Quantitative Health Sciences and Artificial Intelligence and Informatics, Biomedical Ethics Research Program, Mayo Clinic, Rochester, MN, USA.
10Kern Center Mayo Clinic, Rochester, MN, USA.
Andrew Gonzalez, M.D., J.D., MPH
In addition to his role as a research scientist and associate director for data science with the William M. Tierney Center for Health Services Research at Regenstrief Institute, Andrew Gonzalez, M.D., J.D., MPH, is a practicing vascular surgeon and an assistant professor of surgery at the Indiana University School of Medicine. Dr. Gonzalez is also a faculty affiliate with the Regenstrief Center for Healthcare Engineering at Purdue University.
Heidi Lindroth, Ph.D., R.N. FAAN
In addition to Dr. Heidi Lindroth’s role as a nurse scientist, she is an Associate Professor in the Mayo Clinic College of Medicine and Science, is a practicing Agile Scientist, and is pursuing a postdoctoral masters in Artificial Intelligence in Healthcare (NIAK23AG076662-04). Her vision is a world without delirium. She has expertise in neuroscience, advanced statistics, and implementation science.
Partners
The Latest
