Last Thursday (13 June 2024) UK's Medicines and Healthcare Products Regulatory Agency (MHRA) released updated guidance on "Machine Learning Medical Devices: Transparency Principles" that outlines guidelines for communicating clear and relevant information about machine learning-enabled medical devices. This guidance aims to ensure that ML medical devices are developed, deployed, and monitored transparently, fostering trust and safety for patients and healthcare providers.
Machine learning (ML) medical devices refer to medical devices that use algorithms capable of learning from data to improve performance over time without being explicitly programmed.
In 2021, the FDA, Health Canada, and the UK's MHRA collaboratively established 10 guiding principles for Good Machine Learning Practice (GMLP) to ensure the development of safe, effective, and high-quality AI/ML technologies. Building on these principles, the agencies have introduced additional guiding principles specifically for transparency in machine learning-enabled medical devices (MLMDs).
“Transparency” describes the degree to which appropriate information about an MLMD (including its intended use, development, performance, and, when available, logic) is communicated to relevant audiences.
“Logic” refers to information about how an output or result was reached or the basis for a decision or action.
The degree to which this logic can be explained in a way a person can understand is known as “explainability”.
Logic and explainability are aspects of transparency.
The guiding principles for transparency of MLMDs consider the following:
Who: Relevant Audiences
Transparency should address health care professionals, patients, caregivers, support staff, administrators, payors, and governing bodies.
Why: Motivation
Transparency is crucial for patient-centered care, safety, effectiveness, and informed decision-making. It helps detect errors, evaluate risks and benefits, and maintain device safety.
What: Relevant Information
Information should include the device's medical purpose, intended users, environments, target populations, inputs and outputs, performance, benefits and risks, and lifecycle management activities. It should also cover the logic of the model, biases, confidence intervals, and data gaps.
Where: Placement of Information
Information should be accessible through the user interface, optimized for responsiveness, and personalized. Modalities can include audio, video, on-screen text, alerts, diagrams, and document libraries.
When: Timing of Communication
Communication should consider the total product lifecycle, providing detailed information during acquisition, implementation, and use. Timely notifications about updates or new information are essential.
How: Methods to Support Transparency
Applying human-centered design principles ensures the appropriate level of detail is provided for the intended audience, and information is arranged to support decision-making. This involves using plain language or technical language as needed.
These guiding principles aim to foster the development of transparent MLMDs, ensuring safety, effectiveness, and trust in the technology. Continued engagement and feedback from stakeholders will support the evolution of these practices.
For a detailed understanding of the transparency principles and their implementation, refer to the full guidance document available on the UK government website: Machine Learning Medical Devices: Transparency Principles.
In addition, check the recent guidance "Software and artificial intelligence (AI) as a medical device" that recently updated the guidance to include reference to guiding principles on transparency for machine learning-enabled medical devices.
Comments