Machine Learning Operations in Health Care: A Scoping Review

Area of Science:

• Health Informatics
• Machine Learning
• Clinical Operations

Background:

• Machine learning (ML) tools are increasingly used in healthcare.
• The operational processes supporting ML deployment, known as machine learning operations (MLOps), are still developing.

Purpose of the Study:

• To synthesize existing literature on healthcare MLOps.
• To identify knowledge gaps and provide insights for clinical practice adoption.

Main Methods:

• A comprehensive scoping review was conducted.
• Searches were performed across MEDLINE, PubMed, Google Scholar, Embase, and Scopus.
• 148 relevant articles were included after a two-phase screening process.

Main Results:

• Seven key topic areas in healthcare MLOps were identified: model monitoring, automated retraining, ethics/equity/bias, clinical workflow integration, infrastructure/HR/tech stack, regulatory, and financial considerations.
• The review outlines best practices and highlights knowledge gaps within these areas.
• Strengths and weaknesses of the current literature were assessed.

Conclusions:

• This review offers a valuable overview for healthcare ML practitioners and consumers.
• It identifies critical areas for developing and implementing MLOps in clinical settings.
• Further research is needed to address identified knowledge gaps for successful MLOps adoption.