Health & Justice: To Err, and to Use A.I., is Human
Artificial Intelligence and Medical Malpractice Litigation
By K. Zoe Jessie, JD, with Andrew D. DeSimone, JD
Did you know the FDA has already approved more than 800 Artificial Intelligence enabled medical products and devices in the United States? With emerging AI technology come emerging laws, regulations, and guidelines on how to use it safely and ethically.
Since those are still in the development stage, both the legal and medical communities are navigating what impact AI will have on future litigation. Specifically, how will the new standard of care change? And when AI inevitably errs, who will be liable? To begin to answer these questions, we must look at how AI is currently used, how it is currently regulated, and how it could apply to medical negligence litigation.
Current Use of AI in Medicine
Despite what movies may portray, AI systems do not “think” the way humans do. Instead, AI uses vast amounts of data to solve problems based on algorithms, patterns, and connections.
The two prominent uses of AI in the healthcare field are machine learning (ML) and natural language processing (NLP). ML is a predictive system that learns from data, then makes decisions or predictions based on that data. In the medical field, machine learning is commonly used in early disease detection. NLP converts human language into a format a computer can use to perform tasks, analyses, and basic commands. In healthcare, NLP models range from simple dictation of operative notes, to transcribing patient information and adding data to EMR and EHRs across platforms.
The type of AI yet to be FDA-approved is Generative AI (think ChatGPT). Generative AI is a form of ML technology that creates novel content from the vast amounts of data it reviews. This can include creating new text, images, videos, and audio. In medicine, Generative AI is mostly being used to take chemical or biological data to create new molecular and protein structures with desired characteristics for use in drug development. It is also being tested for creating medical documentation, and patients are using it to answer medical questions regarding symptoms and possible diagnoses.
Medical Regulatory and Statutory Landscape
The medical and legal communities are slowly, but surely, tackling the use of AI in healthcare. Two prominent federal bills have been introduced in Congress: The Better Mental Health Care for Americans Act (S293) and The Health Technology Act of 2023 (H.R.206).
- S293 would require Medicare Advantage organizations to conduct studies on the use of AI in “nonquantitative treatment limitations” for mental health and substance use disorder benefits. Under S293, if AI is used in the denial of benefits, the organization would also be required to provide information on those denials.
- R.206 would allow AI or ML technologies to be considered a “prescribing practitioner” for medications and drugs. H.R.206 would allow state law to authorize AI to prescribe a specific drug that is approved, cleared, and authorized under federal law governing medical products.
Though few laws on healthcare AI have been implemented, federal agencies have filled the gap by publishing guidelines and proposed regulations. For example, in April 2019, the Food and Drug Administration (FDA) published a “Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device,” also known as “Clinical Decision Support” Model, to discuss potential approaches to reviewing AI/ML technology in devices and device modifications. In 2018, the American Medical Association (AMA) posted policy recommendations for “augmented intelligence” (rather than “artificial intelligence”) to reflect that the physician, and not the AI, is the “primary” or the one in control.
To use AI or not to use it? AI in medical liability litigation
Though the executive and legislative branches are progressing, the judicial branch still lacks clear direction in addressing AI. The federal courts as well as Kentucky’s lack both statutory and case law on the use of AI in healthcare.
A few Kentucky courts have discussed the ethics of AI by analyzing it under Kentucky Rules of Evidence. Kentucky does prohibit the use of AI technology in certain telehealth settings. Under 907 KAR 3:170, each asynchronous telehealth service (a store and forward telehealth service that is electronically mediated) “shall not be solely the result of reviewing an artificial intelligence messaging generated interaction with a recipient.”
In the context of medical standard of care, Kentucky law is limited. Currently, a plaintiff cannot succeed on a medical malpractice claim in Kentucky unless they prove a breach of duty of care and a deviation from the standard of care. This requires expert testimony on the level of care a reasonably competent medical professional would provide in same or similar circumstances. Thus, if AI/ML medical devices and products become ubiquitous, will the standard of care for a practicing physician change?
Most litigation in other jurisdictions addresses the question of “who is liable”” rather than “did the physician deviate?” For example, the Washington Supreme Court in Taylor v. Intuitive Surgical, Inc., 187 Wash. 2d 743 (2017) held that under state product liability law, the manufacturer of a robotic prostatectomy machine could not escape liability from failing to warn the hospital of the robot’s risks.
To think about the possible outcomes of this ruling, consider that over 70% of AI/ML devices authorized by the FDA are in radiology and imaging. In this context, medical negligence may come from a failure in the AI’s ability to recognize liver cancer on an MRI. But is the failure (and therefore the liability) in the programming of the AI, the corporate supervision of the AI programming, the actions of physician in relying on the AI, or the algorithm itself? These questions must be considered as litigation moves forward.
Comparing the Federal Health Technology Act of 2023 to potential state law issues might aid the legal and medical community in predicting the source of future litigation. The Act would allow AI to be a “prescribing practitioner of medicine” if state law allows.
So, if the Kentucky legislature or the Kentucky Board of Medical Licensure implemented laws or regulations stating AI that diagnoses, predicts, and treats illness or disease constitutes the “practice of medicine,” physicians may be required to be “licensed” in the use of AI. The best medical practice could then become using AI to review medical imaging, to give second opinions for diagnosis, or to interpret personal patient data.
If AI becomes ubiquitous, the reliance or failure to rely on AI medical devices could make its way into medical expert testimony. For example, if the best medical practice at the time is the use of AI to review an MRI, and a radiologist refuses to use it, is he deviating from the standard of care? In addition, what if said radiologist misses liver cancer on radiology imaging that AI would have spotted?
These questions and others are, unfortunately, far from being answered. However, AI-based medical negligence and liability litigation is inevitable because humans—and artificial intelligence—will err. For now, it is essential for practitioners to consider the liability and practical implications of this technology and to ensure if they do choose to explore using AI in practice, they ensure appropriate safeguards to protect patients.
Zoe Jessie is a healthcare and civil litigation attorney with Sturgill, Turner, Barker & Moloney, PLLC. She can be reached at zjessie@sturgillturner.com or 859.255.8581. Member Attorney Andrew DeSimone chairs Sturgill Turner’s Healthcare Law and Medical Negligence Defense practice group. He can be reached at adesimone@sturgillturner.com or 859.255.8581. This article is intended as a summary of state and/or federal law and does not constitute legal advice.
