Spillovers from Medicare Cancer Care Models

In the most recent issue of the Journal of Clinical Oncology, Mullangi et al examine whether the Medicare Oncology Care Model led to spillover and practice pattern changes within participating clinics for non-Medicare patients.  This is a fascinating question.  A spillover occurs when a change in practice occurs that impacts people who are not targeted by the intervention.  We know spillovers are quite common in Medicare Advantage.  Once a county reaches sufficient Medicare Advantage enrollment (usually just under 20%) traditional Medicare costs decrease as well.  The observed logic is that doctors typically practice as if they are indifferent to insurance type so once they get told by a big enough payer to do something in a certain way in order to get paid, they’ll do that for basically everyone.

The Oncology Care Model is a bundled payment project.  It gives oncologists a set of cash to care for a cancer patient for six months. Initial research found clinical cost savings but not programmatic cost savings as these models require a lot of bonuses and administrative complexity to run.   But the initial research only looked at costs within traditional Medicare Fee for Service.  This paper looks at a broader range of payers.

So what did it do and how did the researchers try to figure this out:

This observational study used administrative claims from a large national payer, yielding 157,189 total patients with commercial insurance or MA with solid malignancies who initiated 229,376 systemic anticancer therapy episodes before (2012-2015) and during (2016-2021) the OCM at 125 OCM-participating practices (a subset of total OCM practices) and a 1:10 propensity-matched set of 860 non-OCM practices. We used difference-in-differences analyses to assess the association between the OCM and total episode spending…

The OCM was associated with adjusted spending decreases for both high-risk (–$6,756 USD [95% CI, –$10,731 USD to –$2,781 USD], P = .001) and low-risk (–$4,171 USD [95% CI, –$7,799 USD to –$543 USD], P = .025) episodes. OCM-associated spending reductions were strongest for outpatient (–$5,243 USD [95% CI, –$8,589 USD to –$1,897 USD], P = .002) and infused/injected anticancer drug (–$3,031 USD [95% CI, –$5,193 USD to –$869 USD], P = .006) spending. There were no associations between OCM participation and changes in hospital or ED utilization nor quality of care.

WHAT?

The short version is that costs went up everywhere BUT for the practices in the Oncology Care Model, costs for commercially insured patients went up far less and far slower than patients who were treated by docs who were not in the OCM.  There were no differences in quality of care that would make us wary of these cost savings.

Methodologically, this is a reasonable way of asking this question.

Spillovers are fascinating (the R&R I’m sending back this afternoon is a spillover paper) as most policies are not evaluated on their spillovers even if we have reason to believe that spillovers can drive a substantial portion of total results.  Furthermore, once we have evidence of spillovers happening, it changes the policy creation and modification incentives — do we change something in the primary program just to modify the spillover effect?  That is a tough question.

I think I’ll be thinking hard about this paper on my next couple of long walks.