Priority: Ensure artificial intelligence (AI) regulation supports physician practices but does not replace physicians’ medical expertise.
Background: The success of AI in medicine depends on physician expertise at every step in the life cycle of an AI product. “This includes training of the algorithm, testing, roll-out, validation and surveillance,” says San Antonio radiologist Zeke Silva, MD. And the Texas Medical Association would like to keep it that way.
Dr. Silva uses the technology in his practice for several narrow applications, like detecting a pulmonary embolus on a chest scan. The AI-enabled app he uses cannot, however, consider the findings within the context of the patient’s broader clinical presentation, such as lab studies, symptoms, or prior history, he cautions. These clinical determinations require the expertise of physicians.
For that reason, the chair of TMA’s Council on Legislation stresses that any Texas legislation being drafted on the topic of AI must recognize that medicine has distinct risks and considerations which mandate physician involvement.
“TMA [has] a responsibility to review legislation [that] could affect, in my opinion, what is the most foundational relationship in the entire discussion: the physician and the patient,” Dr. Silva told Texas Medicine.
Dr. Silva captured that importance in testimony to the Texas Senate Committee on Business and Commerce during an interim hearing last summer. There he highlighted an important facet of TMA policy that draws an important distinction between the use of “augmented intelligence” – which is intended to co-exist with human decision-making in patient care – and “artificial intelligence” – which makes decisions autonomously, without human involvement.
Under the direct supervision of himself and others on his staff, Dr. Silva provides feedback on the output of AI-enabled applications to improve their function and output going forward. That training includes labeling medical images, for example, to teach AI how to differentiate between items, or using simulated scenarios to discover where improvements could be made.
AI shows promise in medicine for reducing administrative burdens for physicians and freeing time for patient care, which makes practices more viable and care more accessible. However, Dr. Silva stresses that AI gains – like helping doctors analyze data more efficiently – are only possible through a physician-led approach.
“Our accountability in the outcomes of our decisions and the decisions influenced by AI are greater than other industries, often times literally life or death, but also potentially affecting patient outcomes, patient well-being, and even public health,” Dr. Silva told the committee.
TMA is vigilant of potential adverse impacts of AI in the health care setting, including the potential of creating patient privacy risks.
Solutions: As Texas considers wholesale regulation of AI, TMA wants to ensure physicians play a role in educating lawmakers about its benefits and limitations in health care. The association will seek balanced regulation of AI use in medicine that includes physician oversight, transparency, accountability, and privacy protections, but does not go so far as to restrict practices’ ability to take advantage of its administrative benefits.
“AI has the potential for greatness, but TMA’s focus remains on the physician and the patient, and potential risks to that relationship,” Dr. Silva said.
Legislation adopted in 2023 – like House Bill 2060, which established the Texas Artificial Intelligence Advisory Council and charged it to develop guidelines on best practices – created a foundation for future AI legislation that will likely shape the 2025 session. Likewise, interim charges by the Texas Senate called for a “responsible regulatory framework for AI development” that includes input from the council to inform possible legislation.
TMA plans to examine legislation this session that could become a blueprint for how Texas regulates AI, including in health care. That could extend to efforts by the Texas Department of Insurance to oversee payers’ use of AI, for example in prior authorization, says Matt Dowling, TMA director of public affairs.
TMA will also continue to monitor the rollout of federal AI guidelines implemented last year that may play a role in how Texas develops its own AI framework in the future. Those guidelines established principles for the safe and secure use of AI via federal oversight across different sectors, including in health care.
Likewise, the association will ensure health information technology used by physicians and payers protects patient privacy.
TMA also will follow up on last session’s efforts – like Senate Bill 1467 – that sought to guard against disclosing sensitive test results to patients electronically until three days after the results are finalized.
Federal regulations require patient test results be available to patients electronically as soon as they are published. Without boundaries, however, physicians worry they risk sensitive information – like a cancer diagnosis – reaching patients without physicians first having a chance to review or explain it.
For instance, the presence of a “tumor or mass doesn’t always indicate malignancy or a poor prognosis,” Dr. Silva said.
Medicine argues the three-day window would allow physicians time to review results, schedule follow-up appointments if necessary, and better prepare for potentially difficult conversations with their patients.
“From the legislative perspective, we [want] a reasonable hold on that information [that is] long enough to allow the physician to process the information [and] to communicate it to the patient in a way that’s most beneficial,” Dr. Silva said. “We want to be really mindful that patients receive [sensitive] information … from a person they can trust, which is their physician,” Dr. Silva said.
TMA’s Top 10 Priorities for the 2025 Legislative Session
– Develop a noncompete agreement law that works for all physicians