Implementing an Artificial Intelligence Decision Support System in Radiology: Prospective Qualitative Evaluation...

Area of Science:

• Radiology
• Medical Imaging
• Health Informatics

Background:

• Digital health technologies, including AI clinical decision support systems, are increasingly integrated into medical imaging practices.
• While AI shows clinical validity, its long-term implementation and impact on radiologists' workflows are not well understood.

Purpose of the Study:

• To qualitatively evaluate the real-world implementation of an AI decision support tool in radiology across different phases.
• To identify factors influencing AI adoption and sustainability using the NASSS framework.
• To inform strategies for effective and safe AI integration in public hospitals.

Main Methods:

• A prospective qualitative study involving 43 semi-structured interviews with radiologists and radiographers.
• Interviews were conducted across pre-, peri-, and postimplementation phases at a tertiary hospital.
• The Nonadoption, Abandonment, Scale-up, Spread, and Sustainability (NASSS) framework guided the identification of implementation factors.

Main Results:

• Organizational barriers were prominent early on, shifting to technological issues like accuracy and information overload post-implementation.
• Enablers increased over time, with some clinicians using AI as a safety check.
• Trust and adoption were hindered by inconsistent performance, poor communication, and medicolegal uncertainties.

Conclusions:

• AI implementation in radiology is a complex sociotechnical process involving organizational and cultural elements.
• Sustainable AI adoption requires addressing user experiences, enhancing communication and training, and fostering feedback loops.
• The NASSS framework provides insights into dynamic interactions for effective AI integration.