Artificial Intelligence Pilots in Community Oncology Yield Promising Results
Posted on April 9, 2019
A panel at the Community Oncology Alliance (COA) annual meeting (April 4-5, 2019; Orlando, FL) spoke to the evolution of artificial intelligence (AI) as well as the results of AI implementation in a few community practices.
John Frownfelter, MD, FACP, chief medical officer, Jvion, began the session by elucidating some data trends in health care’s current market. Demand for analytics is at an all-time high, he stated, and health care providers are interested in data to illustrate how they are performing against quality standards, identifying patients who are at high-risk of 30-day mortality, and determine whether interventions are working.
He then listed the definition of artificial intelligence as computers performing tasks that are usually assumed to require intelligence. An AI machine can accept information about a problem from its surroundings, generate insights based on this data, and determine the best course of action that will lead to a desired outcome, he explained. Among the top applications of AI within health care are robot-assisted surgery, fraud detection, and precision medicine.
While AI has been growing in popularity in many fields, adoption been a slow process in health care. Medical culture is the biggest barrier to integrating AI, Dr Frownfelter asserted, noting that clinicians often are skeptical of new technology and are hesitant to submit their trust.
Aaron Lyss, director of value-based care, Tennessee Oncology, spoke next about a pilot AI use in his practice. Tennessee Oncology participated in the Cardinal/Jvion partnership to provide cognitive analytics solutions for risk stratification. Lyss and colleagues compared AI-generated high-risk patient lists with patients already identified through care navigation risk triggers, measured predictive capacity of the AI technology, and measured improvement in care delivery through palliative/hospice referrals and utilization of anti-depressants.
Mr Lyss and colleagues found that the AI technology provided validation of how they assess high-risk patients and interventions in real-time. However, there were gaps between what AI could deliver and the current capacity of other clinical systems to make AI insights easily actionable in real-time. Furthermore, he noted that it was difficult to measure the impact of AI versus other care-delivery improvement initiatives aimed at the same outcomes.
Concluding the panel session was a couple more examples of pilot success stories of AI in community practices. Amy Ellis, director of quality and value-based care, Northwest Medical Specialties, shared pilot outcomes from Northwest as well as The Center for Cancer and Blood Disorders. In the Northwest pilot, the oncology specialty vectors included in the deep dive were 30-day mortality, 30-day pain management, 6-month depression, 6-month deterioration, 30-day avoidable admission, 30-day ED visit, and readmission. She noted that changes in mortality metrics and hospice/palliative care referrals were the most illustrative of the effects of the AI intervention. There was a 225% increase in the rate of hospice referrals per 1000 patients per month after Jvion implementation, as well as a 35.3% increase in the rate of palliative care referrals and supportive care consults per 1000 patients per month. Additionally, counts of patients with severe pain per month and patients with depression per month decreased substantially from baseline to 12-months post-Jvion implementation.
Similarly, changes in mortality metrics for hospice and palliative care referrals at The Center for Cancer and Blood Disorders were positive as a result of Jvion implementation. There was a 113.3% increase in the rate of hospice referrals per 1000 patients per month, as well as a 218.8% increase in the rate of palliative care referrals per 1000 patients per month.
View original article HERE.