Artificial Intelligence and Community Oncology

The Community Oncology Conference recently featured a panel discussion of organizations piloting an artificial intelligence (AI) program.

The discussion opened with John Frownfelter, the Chief Medical Information Officer at JVION, describing the challenges of cancer care today and value-based programs, in particular. Providing the highest quality of care at a lower cost in an increasingly complicated landscape is difficult. In order to achieve the goals of value-based programs, there is too much information and too many variables to succeed without the help of technology. While AI is a common technology in areas like robotics and imaging, it’s only in its infancy for healthcare, a sector historically slow to adopt new technologies. However, as oncology care providers juggle new requirements and expectations and as treatment regimens become increasingly more complex, some are considering its potential.

Frownfelter sees the oncology market as ripe for disruption since AI will help in ways humans can’t and the complexities of cancer care are expanding rapidly. AI can take large amounts of data and draw associations that would be difficult for a human to do. There is still a need to study and understand the conclusion, but it can be very valuable.

The Challenge

Cancer care providers are trained to understand information fully and to dissect scientific information to confirm treatment decisions on their own, so it’s difficult to accept a new technology that might not fully explain how a conclusion was reached. The potential benefit of a tool like AI is that it helps clinicians anticipate what might happen and therefore can guide interventions that will be most impactful, so it is beginning to gain traction.

The Pilot

Northwest Medical Specialties and The Center for Cancer and Blood Disorders are both participating in a pilot with JVION and have been surprised by the results. They set up the following vectors within the AI software and were then alerted when the risk for certain patients increased:

  • 30 Day Mortality; Patients at risk of mortality within 30 days of prediction
  • 30 Day Pain: Management Patients at risk of having severe/moderate pain within 30 days
  • 6 Month Depression: Patients at risk of having a depression diagnosis within 6 months
  • 6 Month Deterioration: Patients at risk of deterioration of ADL levels (at least 2 levels) within 6 months
  • 30 Day: Avoidable Admission Patients at risk of an avoidable IP admission within 30 days
  • 30 Day: ED Visit Patients at risk of an ED visit within 30 days
  • Readmission: Patients at multiple admissions within 3 months

Results: Northwest Medical Specialties

  • 6 month Deterioration: up to 30% reduction in loss of function/ADLs (ECOG)
  • 6 month Depression: 22% increase in depression diagnoses
  • 30 day Pain Management: 33% reduction in moderate and severe pain

Results: The Center for Cancer and Blood Disorders

  • 6 month Deterioration: 17% reduction in loss of function/ADLs (ECOG)
  • 6 month Depression: 33% increase in depression diagnoses
  • 30 day Pain Management: 28% reduction in moderate and severe pain


These early metrics are encouraging and, as Dr. Blau pointed out, they are also learning a lot throughout the pilot. When a patient ends up on the high risk list, they may need an intervention that you couldn’t have foreseen. When social determinants, pharmacy visits and a host of other data points are considered in the algorithm, it can pick something up that you might not have.

All of the panelists agreed AI could never replace the human element of cancer care, but it can be an important new tool, like lab results or medical history, to be factored in to the care provided.

Stay in the Know

Subscribe to the Navigating Cancer blog for the latest news, innovations, and best practices in oncology care.