How Will AI Change Health Care?
Doctors say AI has the potential to reshape patient care for the better, but the technology needs to be implemented responsibly.
As the capabilities of artificial intelligence (AI) surge, the healthcare sector is rapidly finding new ways to use algorithms and machine learning for patient care. In 2022 alone, the Food and Drug Administration (FDA) approved over 90 different medical devices using AI, including algorithms that track atrial-fibrillation history or rapidly analyze X-rays to diagnose collapsed lungs. As these new technologies multiply, it’s becoming clear to doctors and patients alike that AI has the potential to reshape medicine.
“Radiology is probably one of the specialties most impacted by artificial intelligence,” says Nicholas Lim, MD, a radiology resident physician.
Of the over 500 submissions of AI medical devices authorized to date by the FDA, three-quarters are for radiologists. These tools improve image quality, detect difficult-to-see features and flag the most severe patients for immediate treatment. For example, in one study, Jefferson researchers identified 22 new AI-based brain imaging tools to speed stroke diagnosis and promote post-stroke recovery.
Radiologists are also experimenting with AI tools that help patients outside of the exam room. In a recent study, Ryan Lee, MD, chair of radiology at Jefferson Einstein Philadelphia Hospital, used an algorithm that analyzed physician reports to automatically recommend follow-up appointment dates. He found patients who received those automatic reminders were more likely to return for important follow-up visits than those who did not have these reminders.
Other medical fields are employing AI to improve patient care, too. Chen Wu, MD, a professor of neurological surgery and radiology, recently published a study that used machine learning to predict quality-of-life changes in patients with Parkinson’s disease to help guide treatment. Based on clinical symptoms alone, the algorithm flagged patients who were likely to experience a drop in quality of life in the upcoming year for further testing and medical attention.
Dr. Wu imagines AI can help patients with Parkinson’s even more in the future. One of his areas of expertise is deep brain stimulation (DBS), a method that helps restore function to damaged motor circuits in Parkinson’s patients’ brains. These devices often require months of delicate calibration, and success is highly dependent on the experience of the managing physician. Dr. Wu believes AI could be leveraged to rapidly calibrate DBS devices based on a patients’ individual neurological features.
“The uniform results you get from AI could help standardize outcomes and care,” says Dr. Wu. “That’s where the big benefit is.”
But while AI has the potential to improve patient care, there are risks if it isn’t used carefully. One of the biggest problems is ensuring the data an algorithm is trained on is representative of the patients it’s eventually used on. “If you train your model off of biased data, you’re going to generate biased results,” says Dr. Wu.
Dr. Lim says it’s also important to understand how the algorithm works. Many AI algorithms are “black boxes,” meaning programmers may not be sure how they generate their outputs. This is important as AI tools lacking in transparency could be more likely to demonstrate biases.
“It’s important to understand the steps and calculations that the algorithm is going through to produce its output as best we can. This allows us to find potential weaknesses where health disparities may be perpetuated,” says Dr. Lim.
But if these factors are addressed, Dr. Lim believes AI tools could be a pathway towards health equity: the attainment of good health for everyone regardless of gender, race or socioeconomic status. If algorithms are built in a way that accounts and corrects for current biases and inequities, they could reduce health disparities by improving diagnostic accuracy and overall patient care.
Patricia Henwood, MD, executive co-sponsor of Jefferson’s newly formed AI Center of Expertise and the executive vice president, chief clinical officer and James D. and Mary Jo Danella Chief Quality Officer - Jefferson Health, agrees. “There’s incredible promise and potential, and I think there’s much more to come.”
Ultimately, the doctors say that AI still won’t be replacing them any time soon — it’ll just be another way for them to serve patients.
“AI is not taking over what we do,” says Dr. Lee. “The take home message is that the physician in concert with these tools will ultimately result in better patient care.”
