Thread by @onceuponA

I hadn't read last week's Lancet paper on M4A, but thought I would today since Bernie cites it in his newish-mostly-old set of pay-fors. thelancet.com/journals/lance…

Now that I've read it, I wonder—quite sincerely—how this article got past peer review.

The thing that jumped out to me most glaringly was the stipulation that utilization will ONLY increase among the 24% of Americans who are currently uninsured or underinsured (the yellow box has the underinsurance criteria from the cited Commonwealth Fund)

The implication here is that going from modest cost-sharing to zero cost-sharing will have no impact on utilization. None.

This is not a tenable assumption. The RAND HIE is dated, but it's not obsolete—it's backed up by a whole canon of health economics research now.

The report also suggests that universal coverage will decrease avoidable ED visits and adminission.

The best relevant evidence we have (from Oregon's randomized Medicaid expansion) is that coverage actually increases use of the ED, including for nonemergent care.

This isn't necessarily surprising. The ED is open after-hours, and the hospital may be more accessible by public transit than random PCP practices. Patients who face non-financial barriers to care—of which there are many!—may rationally increase ED use. theincidentaleconomist.com/wordpress/the-…

Now, there was quasi-experimental evidence on the under-26 population that found modestly decreased use of the ED, but it strikes me as dicey to extend findings for this very particular population (young adults with access to parents' private insurance) to the full population.

That paper is here (healthaffairs.org/doi/full/10.13…) and I wrote it up here (theincidentaleconomist.com/wordpress/insu…)

Here's an assertion I just don't understand: the withdrawal of cost-sharing reductions for marketplace plans increased premiums... for people in employer-sponsored coverage?

There is no citation.

Now turning to the insurance and mortality results. Let's start by agreeing that the existing research on this is various shades of gray, and we don't have a great understanding of the "true" mortality effect of insurance, or how that varies based on things like health status.

There has been a lot of important recent work in this area, exploiting the way the ACA expanded coverage.

However, to reach their top-line number (68,531 lives saved annually) the authors rely on a single paper from 2009. ncbi.nlm.nih.gov/pmc/articles/P…

This isn't because the authors are unfamiliar with other work. They even plot various studies on a continuum of effect size. The study they used for their estimate reports the largest effect size.

They go on to call heir calculation "highly conservative in a number of aspects."

I would like to take this moment to plug two very good studies on mortality effects of insurance (Medicaid) expansion that were *not* cited in this paper:

Medicaid expansion, generally: www-personal.umich.edu/~mille/ACAMort…

For ESRD patients specifically: jamanetwork.com/journals/jama/…

Less important, but still worth noting: Traditional Medicare is not known as some exemplar of coordinated care. We've been muddling through this whole ACO thing for years trying to do better.

Yes, one payer would do something to improve continuity relative to multiple payers over time.

But we're still going to have issues related to, for example, interoperability of EHRs—perhaps even more so when there are no networks hemming people into shared HIT ecosystems.

(brief interruption for meeting with a student)

I deliberately raised all of these issues without mentioning the assertion of reimbursement at Medicare rates and no higher. That, and use of the VA formulary, are easy targets to critique, but they're policy considerations, not misreadings of the literature.

I'm not sure if Sanders leaning on this plan means he endorses 100% Medicare rates (his legislation is silent on this extremely consequential detail).

MedPAC says hospitals had an aggregate Medicare margin of -9.9% in 2017. medpac.gov/docs/default-s…

And if we're talking about policy considerations now, it appears that this plan does not account for the introduction of a universal long-term care benefit—a monumental omission.

The paper's applicability to the present debate is also complicated by the fact that it talks about "annual" savings without stipulating a spending growth rate (are we freezing rates? tying them to general inflation?) or doing any projection over ten years per policy norms.

Oh back to the literature for a hot second, I forgot about the fraud stuff.

We're just extrapolating from Taiwan's experience with moving to single-payer because reasons. (This is not how generalizability works.)

I'm not trying to be a pain in the ass, it just so happens that I am a pain in the ass.

This debate is important, and we're going to be having it for a long time yet. It's important that evidence put forward in service of this debate can stand up to rigorous evaluation.

Anyway, I will conclude by saying that this paper also provoked a very important question this morning, and I'm still unclear on the answer.

https://twitter.com/onceuponA/status/1232353626766467073

Appending this to the thread. We can have a discussion about capacity constraints, but then we have to get comfortable with the R-word.

And if these constraints *are* binding, it's going to erode the purported mortality and health benefits of coverage expansion.

https://twitter.com/onceuponA/status/1232689256986030085

I think there are real supply-side concerns (and that none of the candidates want to talk about them).

I also think that hospitals will find ways to change supply inputs to eke out those extra margins over the medium and long-term.

Attaching this, too. Ascribing all utilization increases to uninsured and underinsured populations is inconsistent with the literature, even if overall metrics look similar—it's cherry-picked evidence that troubled me most and that I aimed to highlight.

https://twitter.com/jodilliu/status/1232781275309670409

That selective application of existing research—not whether 100% Medicare rates is a reasonable policy decision, or how much we should expect administrative costs to fall—is where I would have expected peer review to intercede.