As artificial intelligence’s role in healthcare rapidly expands, a comprehensive new report co-authored by UCLA Health states that the same technology that can help doctors detect strokes or seizures could also worsen health disparities unless proper safeguards are in place.
The report1, published in the journal Neurology, the medical journal of the American Academy of Neurology, examined AI’s growing role in neurological care. While the technology has already shown benefits such as allowing doctors to make faster decisions in classifying brain tumors or analyzing stroke imaging, researchers say AI’s reliance on large datasets poses a risk for patients in vulnerable populations who are already underrepresented in research and underdiagnosed.
At the same time, AI presents potential to allow for healthcare providers in resource limited settings to recognize early signs of neurological diseases based on clinical notes, for clinics to improve enrollment of underrepresented groups in research studies, or for health systems to ensure all patient groups are receiving high quality care and improved health outcomes.
“That means that AI could help doctors in areas with a shortage of neurologists to recognize neurological diseases months earlier, ensure medications match what patients can afford, automatically write medication instructions in the patient’s primary language and flag when certain populations are being systematically excluded from clinical trials.”Dr. Adys Mendizabal, a neurologist and health services investigator at UCLA Health.
“The technology exists. We just need to build it with equity as the foundation,” said Dr. Adys.
Consulting with experts in healthcare, AI experts, Food and Drug Administration officials and one AI company, Mendizabal and researchers from nine other universities identified both the benefits and pitfalls of AI implementation in neurological care and created three guiding principles for future implementation:
Diverse perspectives must shape AI development: healthcare institutions must involve community advisory boards reflecting the demographics of populations they serve to ensure AI tools are culturally sensitive and linguistically appropriate.
AI education for neurologists: researchers must understand AI is not an infallible source of information and should be trained to recognize potential biases in algorithmic outputs.
Strong governance: Independent oversight with clear accountability must be established to monitor AI performance, investigate failures and give patients the ability to report concerns or delete their health care data.
Investigators said the governance of AI must evolve continuously alongside the technology itself, requiring constant collaboration between government regulators, healthcare institutions, AI developers and patients.
“We are at a critical moment,” Mendizabal said. “The decisions we make now on how to develop and deploy AI in healthcare will determine whether this technology becomes a force for equity or another barrier to care.”
Reference:
1) https://www.neurology.org/doi/10.1212/WNL.0000000000214356
(Newswise/HG)