Artificial intelligence (AI), specifically machine learning (ML), has revolutionized the medical practice and has become a vital part of the healthcare industry. However, while these technologies have tremendous potential to improve healthcare, there are still regulatory concerns surrounding their use. In an article published in the HealthLawProf Blog, Sara Gerke, a professor at Pennsylvania State University, emphasizes the urgent need for labeling standards for AI/ML-based medical devices.
Gerke notes that the U.S. Food and Drug Administration (FDA) has already cleared or approved over 520 AI/ML-based medical devices, with many more in development. These devices are not just used by healthcare providers but are also increasingly offered directly to consumers through applications and wearables. Although these devices are promising, there are currently no standards for labeling them, which is a significant problem given that some of these devices are opaque and prone to biases.
Gerke argues that labeling is critical to ensuring that patients and consumers can use these devices safely and effectively and understand their potential risks and limitations. To this end, she proposes a model of labeling akin to the nutrition facts labels found on food products. This would include 11 key types of information, such as indications for use, data sets, model limitations, warnings, and precautions related to privacy and security. Gerke also suggests that the labeling should be dynamic, which would allow adaptive algorithms to continuously learn and evolve.
Gerke concludes by emphasizing the need for labeling standards for AI/ML-based medical devices to protect patients and consumers, promote transparency, and help shape the broader AI/ML-based products that are not subject to FDA regulation. She suggests that implementing such labeling standards will require coordination between regulatory agencies, technology developers, and healthcare providers to ensure the effective and appropriate use of these powerful technologies.
