Enhancing the value to users of machine learning-based clinical decision support tools: A framework for iterative,...

Area of Science:

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

Background:

• Healthcare organizations are increasingly adopting machine learning (ML)-based clinical decision support (CDS) tools.
• A lack of clear implementation guidance hinders the effective use of these tools by practitioners.
• Effective integration is crucial for ML-CDS tools to assist end-users in their daily work.

Purpose of the Study:

• To identify strategies for health care organizations to facilitate collaborative development of ML-based CDS tools.
• To enhance the value of ML-CDS tools for real-world healthcare delivery.
• To provide a framework for successful ML-CDS tool implementation.

Main Methods:

• Qualitative study employing 37 interviews within a large health system.
• The health system had developed and implemented two operational ML-based CDS tools.
• Thematic analysis was used to develop an explanatory framework and recommendations.

Main Results:

• ML-based CDS tool development and implementation occur in four iterative phases: coidentification, coengagement, coapplication, and corefinement.
• Each phase involves a collaborative, back-and-forth process between developers and users.
• This iterative process transforms both user workflows and the ML-CDS technology itself.

Conclusions:

• Organizations embracing iterative collaboration in ML-CDS development and implementation are more likely to succeed.
• A traditional technology innovation process is less effective for ML-CDS tools.
• Anticipating and planning for iterative collaboration is key to deploying effective ML-CDS tools.