Designing rational spending targets and having small sample sizes are 2 main challenges payers and partners face in the shift toward alternative payment models (APMs), said Ravi B. Parikh, MD, MPP, assistant professor of medical ethics and health policy, assistant professor of medicine, University of Pennsylvania.
Ravi B. Parikh, MD, MPP, assistant professor of medical ethics and health policy, assistant professor of medicine, University of Pennsylvania, discusses the partnership between University of Pennsylvania and Tennessee Oncology, as well as the barriers commerical payers and practice partners face in the shift to alternative payment models (APMs) in oncology.
Parikh and his co-authors published an article in the March issue of The American Journal of Managed Care®, "Oncology Alternative Payment Models: Lessons From Commercial Insurance."
Can you discuss some of the work the University of Pennsylvania is doing with Tennessee Oncology?
We have a one-of-a-kind partnership with Tennessee Oncology, which is a large community-based oncology network in Tennessee and Georgia, where we're focused on using analytics to drive value-based cancer care. Specifically, we know that specialty palliative care, which is a specialty that focuses on management of symptoms and burden of serious illness, is underutilized in cancer care when we know it has a lot of downstream benefits to quality of life and maybe even survival.
What we're doing is we're partnering with Tennessee Oncology at the University of Pennsylvania to design a really one-of-a-kind clinical trial to use advanced analytics in electronic health records to identify patients who may benefit from earlier palliative care and flag those to clinicians in order to prompt default or opt out referrals for palliative care so that more patients can access this really valuable service, oftentimes concurrent with their cancer treatment. We think what that's going to do is demonstrate that predictive analytics and these behavioral nudges or prompts for value-based behavior can be scaled in a really unique way in a community oncology setting that's never been done before.
How can commercial payers and practice partners overcome barriers they face in the shift from fee-for-service to alternative payment models in oncology?
I'll highlight maybe 2 or 3 barriers. The first is designing rational episodes for cancer care. In other conditions, like post hospitalization or diabetes, it's a little easier to identify who your base population is. It's those folks who have a diagnosis code or those folks who have had a recent hospitalization, or whatever it is. In cancer care, it's a little different, because cancer can mean about 20 or 100 different diseases, and it's really hard to identify some of that from claims-based data.
Commercial alternative payment models have found some challenges in really trying to identify who are the right patients and what type of cancers people do have in order to set rational spending targets and quality targets. We propose in the article actually generating better data exchanges between practices with the electronic health record and payers, so that we can more rationally define true episodes of cancer care and get more specific, which is the goal in the precision oncology era.
Another potential challenge with commercial alternative payment models, particularly in comparison to Medicare APMs, is small sample sizes or small numbers. It becomes really difficult to set targets and look at variability in some of those spending targets when you have certain practices that only see 1 or 2 breast cancers or 1 or 2 multiple myelomas in a given year.
What we've argued for in that article is trying to focus, as a proof of concept for commercial APMs, in looking at cancers that have sufficient numbers and sufficient volume and being really conscious in what cancers—particularly for common cancers or cancers that involve a disproportionate share of spending—that we include in some of these commercial APMs so that we can sort of walk before we run and not bite off too much more than we can chew with how we how we define some of the targets. So those are just 2 of a few different suggestions we have.