Dive Brief:
- The Food and Drug Administration and its international partners have created guiding principles for transparency for machine learning-enabled medical devices, the agency said Thursday.
- Working with its counterparts in Canada and the U.K., the FDA has developed the principles to ensure information that could affect risks and patient outcomes is communicated.
- The agencies explained what information device developers should communicate and how they should get it to users to provide the industry with best transparency practices.
Dive Insight:
The FDA, Health Canada and the U.K.’s Medicines and Healthcare products Regulatory Agency jointly identified 10 guiding principles for good machine learning practice in 2021. The principles touched on the need for transparency by recommending users are given clear, essential information, but lacked a detailed look at how to address that requirement.
Now, the three agencies have published a second document focused on transparency, which they define as the degree to which appropriate information about a machine learning-enabled medical device is clearly communicated to relevant audiences.
Algorithms can provide an output, such as a diagnostic recommendation, without explaining how they reached their conclusion. A device has “explainability” if it can describe the logic behind its output.
It is valuable to provide users with the logic of the algorithm “when this information is available and easily understood,” the FDA said. The information can equip users to “critically assess the device and its output when patient care decisions are being made.”
Effective transparency entails considering the information users need and the best way to communicate it, according to the document. The FDA and its collaborators want device developers to consider the who, why, what, where, when and how of transparency.
“What” covers the type of information that companies should share. Best practices include providing a clear and accurate description of a product and how it fits into the healthcare workflow.
Other sections of the document cover where, when and how to provide information. Device information can be accessed through the user interface, the agencies said, and it is “a good practice is to optimize use of the software user interface so that the information it conveys is responsive to the user.” Another proposal is to provide “timely notifications” when a device is updated or new information is discovered.
