MHRA issues new guidance on transparency for machine learning medical devices

The MHRA has issued new guidance relating to transparency principles for machine learning medical devices (MLMDs), to help ensure that information is clearly communicated to relevant audiences.

The guidance sets out the guiding principles of transparency for who, why, what, where, when, and how. The ‘who’ explores relevant audiences, looking at who will be using the device, and who will be receiving healthcare with the device, as well as additional parties who may be involved in decision-making on a device’s use.

The ‘why’ looks at the motivations for MLMD use, with information needing to be communicated around identifying and evaluating the device’s risks and benefits, to promote informed decision-making, to help in the detection and investigation of errors, and to promote health equity through the identification of potential bias.

Under the header of ‘what’, the guidance outlines the kind of information to be shared around MLMDs, including a “clear and accurate description of the device” which covers information about its medical purpose and function, the conditions it is designed to address, and its intended users or user environments and target populations. Information on product development and risk management across the device’s lifecycle is also recommended.

On the placement of information, or ‘why’, the guidance recommends optimising use of the software user interface which covers elements of the device the user interacts with, so that the information conveyed “is responsive to the user” and their needs.

The timing of this communication is covered under ‘when’, with recommendations including a consideration of information needs across every stage of the total product lifecycle, including when decisions are to be made about acquiring a device, and on how to use it. According to the guidance, it may also be useful to provide ‘timely notifications’ about updates to the device, or new information that is discovered, as well as targeted information such as on-screen instructions or warnings upon specific triggers.

Finally, the guidance looks at ‘how’, recommending the application of human-centred design principles, that involves relevant parties throughout the design and development process, providing “the appropriate level of detail for the intended audience”, and choosing between plain language and technical language dependent upon that audience.

To read the guidance in full, please click here.

Last month, the MHRA announced the launch of its regulatory sandbox for AI as a Medical Device (AIaMD), AI Airlock, a pilot project which forms a “key part” of the agency’s recently published strategic approach to AI.

For a recent interview, we sat down with Ben Farrar, business development manager for the healthcare sector for UK-based company Traka, to discuss operational efficiencies for device management in healthcare.