(Image Credit: AdobeStock/Lee)
As the integration of AI in health care continues to evolve, it is vital to see how the technology is being used for diagnostics, patient management, and the streamlining of administrative tasks in ophthalmology practices.
With a number of possible benefits, AI technology does have the potential to change the way practices will work in the coming years, but it also demonstrates that it needs additional advancements in terms of relieving physician burnout and driving organization.
Travis Redd, MD, MPH, assistant professor of ophthalmology at Casey Eye Institute Oregon Health and Science University, explained in a conversation with Ophthalmology Times that AI has mostly been implemented for autonomous screening and diabetic retinopathy in practices. As of now, Redd said there also are only 3 FDA-authorized, AI-enabled software being used as medical devices for automated diabetic retinopathy screening in the US.
With only 3 AI models available for practices, there are some significant barriers for ophthalmologists and optometrists.
“The main barriers, the biggest ones, are that the data sets that we have to train and evaluate AI models are relatively small compared to most deep learning models,” Redd said. “It’s much more difficult to share and aggregate data in health care than it is in non-healthcare settings because of the need to protect patient privacy and administrative barriers between institutions and sharing data.”
Redd noted there’s also challenges with logistical issues, as a lot of the medical imaging used in ophthalmology is not in a standardized format. Right now, practices use the DICOM standard for formatting medical images, but a lot of the equipment used for medical imaging in practice is not DICOM compliant, making it extremely difficult to share images obtained by different devices or institutions.
However, AI software in eye care is looking to have a positive impact on the future of increasing accessibility to care for patients.
“One of the bigger potential benefits in the future is just increasing accessibility to care,” Redd said. “A lot of people have difficulty getting in to get a thorough eye exam by an expert, so if we have technology that can perform automated screening and help identify patients who actually really do need to come in and get more advanced care, versus people who can continue to go about their lives, that drastically decreases the number of patients who actually need to be seen in a optometry or ophthalmology office.”
Looking ahead, Redd pointed out it would be helpful to have large language models implemented into electronic health records since a lot of physicians’ tasks in clinical documentation could be easily abstracted. He also says this implementation could decrease the administrative burden on clinical care if used in practices’ administrative, documentation, and scheduling software.
After going through these benefits and consequences in his own experience, Redd has many suggestions for how to smoothly and efficiently integrate AI into practices.
“All of those models really require much larger training and evaluation data sets that we currently have, so finding better ways to share data across institutions, make data from different sources more easily compatible, and making sure that all medical imaging data is DICOM compliant in eye care, and then also coming up with the reimbursement models,” Redd said. “That’s another big barrier to actually implementing this technology is we don’t have a great framework for how people get paid if they use it.”
Redd also advises that companies creating AI software and equipment for practices should collaborate with eye care professionals, allowing them to be involved in the beginning of development.
“Most people with machine learning expertise don’t have clinical expertise, so I think we really need to ask eye care professionals, we need to be involved from the beginning of the model development process to make sure that the models that are developed makes sense from a clinical perspective, provides an added value, and answer clinically meaningful questions,” Redd concluded.
The results of a recent survey, conducted by the American Medical Association and released earlier this year, indicated that physicians are intrigued by the potential of AI to enhance diagnostic accuracy, personalize treatments and reduce administrative burdens. However, they also expressed worry about the potential to introduce bias, put privacy at risk and create new concerns for liability.1
Nearly two-thirds of the 1081 physicians responding to a the survey said they can see advantages to using AI, though just 38% indicated they were using it when the survey was completed in 2023.
Moreover, 56% said they see opportunity in addressing administrative burdens through automation, and 41% said they were both equally excited and concerned about potential uses of AI in healthcare.
“Whatever the future of health care looks like, patients need to know there is a human being on the other end helping guide their course of care,” AMA President Jesse M. Ehrenfeld, MD, MPH, said. “That’s essential.”
Reference:
-
November 2023 AMA Augmented Intelligence Research Physician Sentiments around the Use of AI in Heath Care: Motivations, Opportunities, Risks, and Use Cases 1 Background and Objectives. https://www.ama-assn.org/system/files/physician-ai-sentiment-report.pdf
