AI Breaks New Ground in Radiology Value-Based Care Delivery
Ten years ago, the notion of value-based healthcare as a disruptive model in healthcare delivery was but a gleam in the eye of the Affordable Care Act. Since then, medical imaging stakeholders have been seeking a way forward on the new value-based landscape. In an RSNA plenary session on Monday titled “Radiology in the Value-based Healthcare Arena: Player or Payer”, insights were revealed.
Presenter Dr. James A. Brink, radiologist-in-chief at Harvard Medical School, is a longtime advocate for radiology’s role in delivering quality at an affordable cost. Ironically, according to Brink, breakthroughs are emerging from a sector of radiology that not too long ago was viewed with suspicion by practitioners: artificial intelligence.
Brink presented several examples of emerging AI solutions helping referring physicians reduce unneeded scans and formulate treatment plans for better outcomes, framing these developments as key to the journey toward value-based care delivery.
“Radiology is at the epicenter of medical care and delivery and … we are very well-suited to coordinate care transformation even outside of our immediate specialty … to move from stewardship to leadership,” said Brink, citing a recent study published by colleagues in the Journal of the American College of Radiology.
He stressed that if radiology doesn’t embrace value-based healthcare delivery, other specialties and other organizations will be looking to lead the way. He reminded the audience that in 2017 another RSNA presenter urged radiologists to create measures of accurate and timely diagnosis to make the jump to value-based delivery.
He presented recent efforts to eliminate low-value imaging as a jumping-off point to promote high-value imaging. It requires, Brink said, measurable policies and interventions to overcome bias and knee-jerk attitudes. This is especially critical, given that workloads have gone up for radiologists by 20-30% due to the pandemic.
Brink cited evolving AI applications to support point-of-care clinical CT decision support and adherence to guidelines for incidental lung nodules. It increased appropriate concordance from about 40-50% to over 90%. The radiologist's findings and recommendations can also be inserted directly into the report to the referring physician.
Similar advancement has been recently realized in high-value imaging through automation and AI. Brink highlighted a decision support tool in development for lumbar spine MR. A similar project is underway to better manage patients with pancreatic abnormalities.
Improving the patient experience is also prime for AI applications. Brink cited a platform by a colleague who correlated the geography of the Boston area unemployment and income data to help predict which patients might miss care and treatment due to social determinants of health, such as lack of transportation. Another program is applying AI to achieve better follow-up with lung cancer screenings and consultation.
“We [radiologists] are a key member of the healthcare continuum,” Brink concluded. “We need to work to quantify our effect on outcomes.”