Optimizing Cancer Care Delivery in 2022, Part 2

This blog is the last in a two-part series on value-based care transformation. Read the first post on re-evaluating oncology sites of care.

ACCC President David R. Penberthy’s 2022-2023 theme is focused on data and digital health tools to reduce disparities in care, as well as technology-based strategies to mitigate oncology workforce shortages and improve efficiencies in care delivery and the patient experience. Similarly, the April 6 ACCC workshop, “Remaining Focused on Value-Based Care Transformation,” featured discussions on how technology, like remote patient monitoring and artificial intelligence (AI), are helping cancer programs and practices stay financially viable, while offering high-quality care.

The Promise of Prescriptive Intelligence

Social determinants of health (e.g., access to transportation and/or treatment, health literacy, food insecurity) all affect cancer care outcomes. Recent estimates attribute 60 percent of health-related outcomes to the social, environmental, and individual behavior of patients—with the other factors driving outcomes being genetics (30 percent) and medical care (10 percent). Therefore, it’s critical for oncology professionals to know what risk factors their patients and communities face.

Traditionally, these data are often gleaned from individuals’ ZIP codes, but this approach doesn’t provide clinicians enough information about their patients. To get a 360-degree view of a patient, community insights must be mapped to each person’s unique experiences and risk factors. But how does a cancer program or practice go about attaining these data and using it most effectively? According to John Frownfelter, MD, FACP, chief medical officer at Jvion, Inc., AI can assist in automating the collection of these types of data but often requires a vendor contract or relationship.

Today’s cancer programs and practices need detailed data from each patient. But because our current model is incredibly labor-intensive, “the quality of the data collected rests on patients being transparent and knowing their own risk factors,” said Dr. Frownfelter. “The alternative is to use artificial intelligence, or another means of collecting these data, without [providers] having to manually collect it and without patients having to complete onerous assessments or questionnaires.”

So how does this technology work? First, patients are identified as part of a population of interest (i.e., Black women with metastatic breast cancer). Prescriptive intelligence then collects and/or purchases historical, non-clinical data and imports patients’ current clinical data into the provider’s electronic health record (EHR). “Having that data is just the first step. Next, the data must be interpreted, and that’s where AI comes in,” explained Dr. Frownfelter. AI analyzes the collected data to determine patients’ relative risk and make informed recommendations that can change their healthcare trajectory. “At Jvion, we pull in social determinants of health and generate AI insights that can be displayed in the Jvion web portal or through the partnering provider’s EHR,” he explained.

A Patient Story

Dr. Frownfelter when on to share this case study. Upon receiving a high-risk mortality alert from Jvion’s prescriptive intelligence platform, the oncology practice brought in its patient for assessment. Based on concerning labs, the patient was admitted to the hospital and diagnosed with a urinary tract infection and early urinary sepsis. According to Dr. Frownfelter, none of the patient’s past labs indicated mortality risk, but the AI picked up this risk via the vast amount of other data it collects, stores, and analyzes.

But that wasn’t the end of the story. “The patient was treated and discharged home,” said Dr. Frownfelter. “Four to six weeks later, she came back to her oncologist and said, ‘I am tired and ready to stop fighting.’” After deciding to stop treatment, the patient was enrolled in hospice and passed away peacefully.

“Our solution picks up patterns and changes in what’s happening with the patient,” explained Dr. Frownfelter. This software can track patients’ prescription refills, type and frequency of medication use, and activity levels. “These are all the other data around the patient that are less visible to clinicians, but which offer a true picture of what the patient is experiencing daily. That is the type of data that prescriptive and artificial intelligence can collect and analyze.”

Deconstructing the Mystery Box

Dr. Frownfelter explained that AI uses a clustering technique to arrange patients according to their similarities. “When you have millions of patients that are clustered along thousands of dimensions, you have what feels like an infinite number of possibilities,” he said. And these patients then map to others based on their similarities. This lets providers infer factors (i.e., risk factors) that were not previously measured. Risk identification is then just a matter of cut-off. Prescriptive intelligence (or AI) grades patients on a curve against the whole population of which they are a part. This could be a population of 10,000 patients or one million patients. “Then we take a reasonable cut. Say the top 10 percent [of patients] are high-risk, and then the next 10 to 15 percent [of patients] are medium risk,” explained Dr. Frownfelter. And if too many patients are included in this high-risk group, “We can cut that list and focus on a smaller number of patients so that we [clinicians] can actually do something about those patients. We can develop an intervention to improve the outcomes for these [patients] and decrease costs of care.”

As AI is adopted and deployed, measuring success becomes critically important. At Jvion, “We define and measure—with our clients—what success looks like,” said Dr. Frownfelter. “Metrics like hospitalizations, avoidable readmissions, better use of palliative care and hospice, and better pain management...can all be measured. Most cancer programs have baseline data in these areas, and we then measure and show if AI is helping them improve in those areas.”

For example, Jvion’s AI helped one cancer practice reveal hidden risks for mortality, resulting in increased palliative care consults and hospice referrals.  Analysis of these study data showed a 68 percent increase in palliative care consults and that hospice referrals increased by eight-fold. Though this dramatic increase can be attributed to redefined referral pathways and increased resources made by the cancer practice, Dr. Frownfelter shared that “the prescriptive intelligence insights [Jvion] provided changed clinician decision-making at the point of care. This has implications for the cost of care, patient satisfaction, and the patient and family experience.”

Oncology Is a Team Sport

Dr. Frownfelter concluded that AI helps cancer care teams see oncology as “a team sport.” AI gives oncology professionals the power to see patients truly as they are using sharable data, and “it [AI] brings alignment to the whole cancer care team,” he said.

You can learn more about AI’s potential in oncology and value-based care by registering for and watching the on-demand workshop recording.

This resource was made possible through the ACCC Alternative Payment Model Coalition that is supported by Merck & Co, Inc. and Takeda Oncology.