Medical researchers today are increasingly leveraging large language models (LLMs) and various AI technologies to streamline labor-intensive research processes. By utilizing these advanced tools to analyze extensive data sets, researchers can accelerate the journey from hypothesis development to practical applications that benefit patients.
However, this advancement raises important questions about the ethical use of AI in healthcare and its broader implications in the medical field.
Dr. Anai Kothari, an assistant professor at the Medical College of Wisconsin (MCW) specializing in surgical oncology, is exploring the role of AI as a research aid. He serves as the inaugural director of the Bud and Sue Selig Hub for Surgical Data Science, a program focused on addressing real-world challenges across Wisconsin with the help of data science and AI. In a discussion with Lake Effect’s Audrey Nowakowski, Dr. Kothari shared his insights on AI's influence on healthcare.
“It’s vital to have experts with medical knowledge collaborating with tech companies and research institutions to ensure we integrate the best iterations of artificial intelligence into clinical settings,” Kothari emphasizes.
He observes that the launch of ChatGPT in December 2022 significantly changed perceptions of artificial intelligence. "Before 2022, our understanding of AI was mostly limited to response-based models that required having data in advance to predict specific outcomes," he explains.
The traditional research process, which often involved strenuous human effort, was "labor-intensive and challenging, yet it was the best approach available." Kothari cites an instance where a colleague spent a year devising a model to estimate surgery durations.
"After twelve months of testing 40 various models and dedicating significant human input, the advent of generative AI allowed him to recreate the project in just one week," he notes. "Before generative AI, the process entailed extensive manual labor to ensure precise labeling, data acquisition, and model training; now we possess the capability to conduct similar research much more swiftly."
Despite the expedited data processing, Kothari underscores the necessity of medical expert oversight to ensure accuracy and to protect patient privacy. His responsibilities at MCW include establishing governance frameworks for AI utilization, ensuring that the same privacy safeguards governing electronic medical records also apply when employing AI with patient data.
"The message for patients is clear: many healthcare systems are committed to handling clinical data in the same trusted manner as traditional electronic systems," he clarifies.
Human supervision is also crucial for mitigating bias in AI outputs. “We must maintain human involvement to examine and scrutinize the outputs for bias or any misleading conclusions that could pose risks if not rigorously validated,” Kothari stresses.
While AI has the potential to expedite research, concerns about its impact on employment in various medical specialties arise. Will pathologists or radiologists face job reductions due to AI handling preliminary work? Kothari does not foresee significant job losses.
“While some roles may evolve, such as medical scribes, which I believe are at risk of becoming obsolete, it's essential to consider how these individuals can transition into new, valuable roles,” he states.
"This presents one of the most challenging aspects of the excitement surrounding AI—navigating the implications of its implementation. We need to devise solutions," he adds.
As AI becomes more prominent in healthcare, Kothari emphasizes the importance of fostering AI literacy among both providers and patients.
"My primary concern is ensuring discussions around AI are in-depth enough for everyone to comprehend its capabilities and potential," he notes.
"Patients are empowered to connect their health data to AI tools beyond regular clinical environments, which is immensely beneficial," Kothari explains. "While these technologies can be advantageous, understanding their limitations, challenges, and benefits is crucial for informed usage."
