An integration engineering framework for machine learning in healthcare.

Area of Science:

• Healthcare Systems Engineering
• Clinical Machine Learning Integration
• Artificial Intelligence in Medicine

Background:

• Machine Learning (ML) in healthcare offers significant potential for improving patient outcomes, team performance, and cost reduction.
• However, a low rate of successful clinical integration of ML models leads to inefficiencies, increased costs, and potential patient harm due to a lack of standardized guidelines.
• Systems engineering principles, widely used in industry for complex system integration, provide a robust model for addressing these challenges in healthcare.

Purpose of the Study:

• To propose a novel framework for the development and integration of Machine Learning models in healthcare, grounded in systems engineering principles.
• To address the limitations and challenges identified in applied systems engineering, software engineering, and healthcare ML development.
• To create a harmonized approach that integrates ML software development with systems engineering practices for the healthcare domain.

Main Methods:

• A critical appraisal of applied systems engineering, software engineering, and healthcare ML software development practices was conducted.
• Principles of systems engineering were adapted and applied to develop solutions for identified integration problems.
• The developed framework was aligned with the ML software development lifecycle.

Main Results:

• An integration framework for healthcare Artificial Intelligence (AI) was developed, considering the entire 'system of systems'.
• The framework employs a combined software and integration engineering approach, structured into four phases: Inception, Preparation, Development, and Integration.
• Each phase addresses critical elements within the domains of The Human, The Technical System, and The Environment, including their interactions.

Conclusions:

• Clinical ML models are technical systems requiring careful integration into the broader healthcare 'system of systems'.
• A systems engineering approach ensures that essential elements are considered throughout the model design and integration process.
• The proposed framework serves as a guide for ML model development, enhancing the probability of successful clinical translation and integration.