In this week’s JAMA Health Forums, Xu et al* assessed the role of provider networks in Medicare Advantage against non-network Traditional Medicare on the rate of plausibly preventable hospital admissions. Medicare Advantage insurers are allowed to build networks. These networks are a subset of clinicians in a region. Traditional Medicare does not place cost-sharing restrictions or incentives to steer patients to any particular Medicare participating clinician. The key question is whether or not this matters. They measure admission rates for five disease categories and find that it likely does matter (although more and better causal work is needed!) as going to the hospital is both expensive and quality of life lowering if it the admission was avoidable.
What do they find?
When controlling for the primary care clinician, the RR of avoidable hospital stays in MA vs TM changed by 2.6 percentage points (95% CI, 1.72-3.50; P < .001), suggesting that compared with TM beneficiaries, MA beneficiaries saw clinicians with lower rates of avoidable hospital stays.
The controlling for physician is important. They look at clinicians who are serving both TM and MA patients and find that the physician does not appear to treat those patients differently. Instead there seems to be selection occurring in who the MA plans contract with:
The first decile in the graph are the doctors who send the fewest patients to the hospital for avoidable conditions. These docs have a Medicare panel (we can’t see their Medicaid, ACA, uninsured or commerical patient counts in these data) that is heavy on MA patients and lighter on TM patients. The worst performing docs in the 10th decile treat comparatively more TM patients than MA patients by a substantial margin.
This is suggestive that MA is doing a decent job of contracting with docs who keep patients out of hospitals and/or steering patients to docs who keep patients out of the hospital when possible. We know from other contexts that there is usually a lot of value of chopping off the worst 5% to 10% of a distribution for outcome measures and it seems that MA does this. However, we need to think about what this would mean for access to care, equity and workforce development as MA is now over 50% of all Medicare covered lives.
* Disclosures: I am on an R01 grant application with Dan Polsky that is currently under review. I am playing e-mail ping pong this morning with Kelly Anderson (no relation) and others on a preliminary data/feasibility grant application. If our idea is feasible, we will turn the concept into a big grant and a lot of writing together over the second half of this decade.
Avoidable hospitalizations and provider networksPost + Comments (12)