Transparency in healthcare AI: Testing EU regulatory provisions against users' transparency needs

Area of Science:

• Medical Informatics
• Health Law and Policy
• Human-Computer Interaction

Background:

• Algorithmic systems are integral to modern healthcare, embedded in medical devices and software.
• European Union regulations (AI Act) mandate transparency information sharing via Instructions for Use (IFU) documents.

Purpose of the Study:

• To assess the alignment between end-user transparency needs and the structure of IFU documents.
• To identify potential gaps in information accessibility within current IFU formats.

Main Methods:

• An online survey was conducted with diverse deployer groups: managers, healthcare professionals, patients, and IT workers.
• Participants evaluated the importance of specific transparency needs and indicated the most relevant IFU section for each.

Main Results:

• Significant differences in transparency priorities were observed across different user categories.
• The study identified specific transparency needs that may be difficult to locate within existing IFU structures.
• User feedback indicated potential weaknesses in the current IFU document design for AI transparency.

Conclusions:

• Recommendations are proposed to enhance IFU documents, ensuring they are more locally meaningful and effectively meet diverse user transparency requirements.
• Improving IFU structure can enhance compliance with the EU AI Act and foster trust in healthcare AI.