How AI can improve patient care in radiation oncology

As AI technologies continue to evolve, they may be able to make a significant impact on patient care by reducing the amount of time physicians spend sorting through paperwork and documentation. A new analysis published by the Journal of the American College of Radiology examined this potential at length, focusing on how it applies to radiation oncology.

“The burden of clinical documentation on physicians has substantially increased in recent years due to multiple factors including the introduction of the electronic health record (EHR), the elimination of in-house transcriptionists, and value-based payment programs requiring the reporting of quality metrics,” wrote Join Y. Luh, MD, department of radiation oncology for Providence St. Joseph Health in Eureka, Calif., and colleagues. “Physicians spend an estimated 34% to 78% of their work day creating notes and reviewing medical records in the EHR, costing an estimated $90 to $140 billion in physician time per year.”

So how can AI help? According to Luh and colleagues, these solutions can help automate much of the documentation physicians face, “freeing physician time from clerical tasks, reducing burnout, preserving privacy, and organizing medical data into searchable and useable element.” EHRs, for instance, could seek out every last bit of needed context from a wide variety of places and piece them together into one helpful story that can be read in an instant.

While such breakthroughs may be years away at this point, the authors noted, the potential is certainly there to make a huge impact on patient care.

“An AI enabled auto-scribing platform can allow physicians to focus on patient care and minimize computer time, while generating accurate, evidence-based clinical notes that have increased analytical value,” the authors wrote. “Beyond clinical documentation, AI can assist with scheduling, coding, billing, third-party authorization, order entry, assessment of value, outcomes prediction, and radiation treatment planning.”

Luh et al. explained that certain barriers do exist before the “AI-powered EHR” and other changes becomes a reality. Voice recognition software isn’t where it needs to be, for instance, to turn “broad conversational and clinical jargon” into efficient, easy-to-read documentation that explains a patient’s situation. Improvements in natural language processing and the computing power available to most providers would also be needed.

“Additionally, the distribution of an individual’s care over a fragmented health care system has historically necessitated records transfers between institutions, often via faxed or scanned documents,” the authors wrote. “For an algorithm to effectively query and analyze these prior health records, they must first be converted from an image to text (eg, using AI tools for optical character recognition), and then processed into discrete data fields with clinical meaning.”

Something else to consider, the team pointed out, is that some physicians simply like to review a patient’s chart and go through that process, so much so that it is unclear if there would be a downside to using an AI-powered EHR.

“Will use of AI-based technologies dull clinician skills or insight?” the authors asked. “How might the exponential increase in data volume due to enhanced audio (and potentially video) recordings of clinic visits be stored in a sustainable but secure fashion? How should private features be recorded or otherwise protected (eg, pelvic examinations, faces of family members or other bystanders)?”

It’s questions such as these that will surely keep researchers busy for years to come. But even though some specifics remain unclear, it remains true that advances in AI have the potential to change radiation oncology—and healthcare as a whole—forever.