Implementation framework for AI deployment at scale in healthcare systems.

Area of Science:

• Digital Health
• Artificial Intelligence in Medicine
• Machine Learning Applications

Background:

• Artificial intelligence (AI) and digital health technologies are increasingly integrated into medicine.
• Significant barriers hinder the large-scale deployment of AI in healthcare systems.
• Existing approaches often fall short in addressing the full lifecycle of AI model implementation.

Purpose of the Study:

• To introduce a novel implementation framework for co-creating and deploying AI and machine learning (ML) models in healthcare.
• To guide diverse stakeholders through the entire AI model lifecycle, from design to maintenance.
• To promote human-centered AI design for enhanced clinical utility and personalized medicine.

Main Methods:

• A design thinking approach is employed to facilitate collaboration among designers, developers, patients, and policymakers.
• The framework integrates privacy preservation with clinical parameters to create a reward function for reinforcement learning.
• Explainable AI (xAI) methods are utilized for clinical interpretability, supported by governance and orchestration platforms.

Main Results:

• The framework enables the co-creation and problem-solving throughout the AI model lifecycle.
• It promotes clinical utility beyond mere prediction by ranking competing models based on a defined reward function.
• Integration of governance and orchestration platforms allows for effective model monitoring and management.

Conclusions:

• The proposed framework facilitates the systematic implementation of AI and ML in complex health systems.
• It supports the development of AI-enhanced health solutions with a focus on clinical utility and interpretability.
• This human-centered approach is crucial for overcoming barriers and realizing the potential of AI in healthcare.