Anticipatory moral distress in machine learning-based clinical decision support tool development: A qualitative...

Area of Science:

• Health Informatics
• Artificial Intelligence in Medicine
• Clinical Decision Support Systems

Background:

• Machine learning (ML) systems are increasingly integrated with electronic health records (EHRs) to create clinical decision support (CDS) tools.
• These tools aim to enhance medical care, diagnostics, and therapy by analyzing diverse patient data and identifying risk factors.
• End-user adoption and integration into clinical practice remain significant challenges for CDS tools.

Purpose of the Study:

• To investigate the ethical issues encountered by clinicians during the co-design of an ML-based CDS tool.
• To understand clinicians' anticipated symbolic relationships with CDS tools and potential sources of moral distress.
• To explore the utility of participatory, experience-based co-design in developing ethically sound CDS tools.

Main Methods:

• Applied a symbolic interaction framework and an interpretive descriptive approach.
• Conducted three participatory, experience-based co-design focus groups with clinicians.
• Analyzed qualitative data from clinicians involved in developing a CDS tool for adverse outcome risk detection in outpatient obstetrics.

Main Results:

• Clinicians' anticipated relationships with the ML-based CDS tool were characterized as either promising or morally distressing.
• Anticipatory moral distress was categorized into three sub-types: clinical conflict, partial conflict, and resource conflict.
• Findings highlight ethical considerations related to clinical judgment, comprehensive care, and structural barriers.

Conclusions:

• Participatory co-design is essential for identifying and addressing end-user concerns early in CDS tool development.
• Addressing clinicians' ethical concerns, particularly anticipatory moral distress, is vital for successful CDS tool implementation.
• Continued co-design throughout the development process can foster better integration and acceptance of ML-based CDS tools in clinical practice.