Medicaid in Oregon: Does it Really Matter?

Elias Isquith

Elias Isquith is a freelance journalist and blogger. He considers Bob Dylan and Walter Sobchak to be the two great Jewish thinkers of our time; he thinks Kafka was half-right when he said there was hope, "but not for us"; and he can be reached through the twitter via @eliasisquith or via email. The opinions he expresses on the blog and throughout the interwebs are exclusively his own.

Related Post Roulette

16 Responses

  1. NewDealer says:

    I read the New Republic article because that is the kind of guy I am,

    My guess is that conservatives are focusing on the part of the study that showed physical health did not improve for the people on medicaid. They simply do not care about the financial aspects or the mental health issues.Report

    • Barry in reply to NewDealer says:

      Did not improve to a statistically significant amount. Which, given the sample size, was probably impossible.

      Kevin Drum has looked at the study:

      http://www.motherjones.com/kevin-drum/2013/05/what-we-know-oregon-medicaid-study

      http://www.motherjones.com/kevin-drum/2013/05/no-oregon-medicaid-study-did-not-show-no-effectReport

      • trizzlor in reply to Barry says:

        Drum is a smart guy but he’s weaseling here:

        One fair thing would be to simply say that it’s inconclusive, full stop. It tells us nothing about the effect of Medicaid access on diabetes, cholesterol levels, or blood pressure maintenance.

        Not quite, the study does provide us with an upper-bound on what Medicaid can achieve after two years in this cohort. That upper-bound may not be tight (and a comparison to other treatments would demonstrate that) but it’s not nothing.

        From a Bayesian perspective, the Oregon results should slightly increase our belief that access to Medicaid produces positive results for diabetes, cholesterol levels, and blood pressure maintenance.

        This is just bullshit. Just because you are Bayesian doesn’t mean you stop believing in statistical significance. If Drum wants to conclude that the non-significant decrease in three traits that he highlights – in combination – should have a significant impact then he needs to show us the Bayes factor he computed. Considering his evidence is cherry-picked, it’s not even yet clear what the direction of effect is.Report

        • Barry in reply to trizzlor says:

          “Not quite, the study does provide us with an upper-bound on what Medicaid can achieve after two years in this cohort. That upper-bound may not be tight (and a comparison to other treatments would demonstrate that) but it’s not nothing.”

          Offhand, I don’t see that; please explain (BTW, I’m a statistician, so please write accordingly).Report

          • trizzlor in reply to Barry says:

            Offhand, I don’t see that; please explain

            I’m basing this entirely on the confidence intervals the study reported on their measures, which provide us with lower and upper bounds on where the true value can “reasonably” (according to some definition) be expected to lie. So, for example, they report Systolic blood pressure to have a net change of -0.52 mm Hg (-2.97 to 1.93) so I can be reasonably confident that (assuming normality, independent samples, in this cohort, in this time-frame, etc.) the true effect does not lie below -2.97.

            I can, for example, compare that to a study of dietary patterns on blood pressure which found a diet that reduced Systolic blood pressure by -5.5 mm Hg (-7.4 to -3.7) in eight weeks and be reasonably confident that the this eight-week diet is significantly more effective at reducing blood pressure than two years of Medicaid. Prior to the Medicaid study, I would not have been able to draw that conclusion.Report

            • Barry in reply to trizzlor says:

              OK (with the caveat that the dietary study was on people in stage 1 or below).Report

            • Michael Drew in reply to trizzlor says:

              So… what alternate policy does this lead you to? Public service announcements? Dietary counseling? If the latter, how do we get dietary counseling to people with low incomes? How do we get people to follow the advice, or at least what do you propose to try? It seems to me, and I may be getting your upshot wrong, that you’re treating Medicaid as though it is and can be only one thing. If we identified diet (and exercise?) as the most important and cost-effective health interventions to improve the health of poor people, you still need a way to implement whatever your actual intervention is. It seems to me that providing poor people with regular appointments with nutritionists and (an option you mock below) physical trainers would be a reasonable intervention to undertake at least as a trial. So what’s the payment mechanism? It seems likely and reasonable that if this were to happen, it would and could happen under Medicaid.

              The point is that your view that we can “be reasonably confident that this eight-week diet is significantly more effective at reducing blood pressure than two years of Medicaid” is a false choice. As I point out below, “Medicaid” isn’t actually a thing from the standpoint of analyzing models of heath intervention – it’s something like fifty different things. More accurately, for our purposes we ought to be treating it as something even more broad, like, “the idea that government provide basic and/or prudential health coverage (where what basic and/or prudential means is the subject of study and deliberation, not any one ex-ante definition) to poor people.” If high blood pressure is better treated through diet (and/or exercise) than medication (which I assume is the intervention most prescribed to patients in the Oregon study), that is something that actually can be addressed through Medicaid – if only via affording recipients access to GPs who will attempt to make the determination on a case-by-case basis. But if your point is that diet is a better treatment for high blood pressure than medication – or that some combination is best – then in my view the right thing to do in response to that is not to say that those interventions are better than “Medicare,” but to say that what we need to do is to look for ways to put that knowledge to use in delivering care to patients. That applies to all patients, regardless of what the payment mechanism is for their care, but it may mean looking for ways to ensure that that insight is particularly not neglected in the care of patients whose care is paid for by Medicaid.

              The other thing to note here is that you are really comparing apples to apple carts here, comparing study of the effectiveness of a course of treatment for one particular health condition to the effectiveness of a payment mechanism for improving that condition. It’s okay to do that, but we should account for it in our expectations for results. If we’re going to compare one intervention called “insurance” (whether Medicaid or other insurance) to another “a diet fairly reliably followed by all study participants under the ‘diet’ designation, who were presumably paid to be reliable about it,” we need to understand that a) the insurance intervention is not targeted only that this condition, also that the effect for this condition isn’t be the only health benefit that it is intended to impart, and b) unlike a controlled study where participants’ adherence to the diet can be monitored fairly tightly, insurance has a limited ability to actually ensure behavior changes, whether taking medication, or, even more difficult, effecting consistent changes in diet and exercise habits. The consistency of changes in behavior effected in the context of of a study of the effect of diet and exercise on health outcomes are just never going to scalable to any overall model of health maintenance. Although the size of the Oregon study group probably compares to the size of the group in the study you mention (in fact I’m guessing the study you mention had a larger number of patients with high blood pressure), the treatment was limited to just whatever the Oregon medicaid program consisted of at that time. Lab studies can tell us what might, under ideal circumstances, be the best treatments for various conditions, but quite often their findings will have only limited relevancy for craftin and evaluating health coverage programs at the aggregate (whether private or public insurer) level. Apples and apple carts. Which is not to say that there is no lesson to take. But the lesson might be “incorporate diet interventions into high blood pressure treatment” (a lesson applicable to private insurers as much as public ones), not that “diet is better than Medicaid for treating high blood pressure. What’s your scalable means for delivering that intervention, and why isn’t Medicaid a compatible vehicle for doing so?Report

    • trizzlor in reply to NewDealer says:

      They simply do not care about the financial aspects or the mental health issues.

      The argument I’ve heard from conservatives is that if the goal is to alleviate issues of mental-health or financial suffering then there may be better and more direct ways to do that than by providing everyone with extensive health-insurance. I’ve yet to see a arguments supported by data that outline precisely what that policy would look like, but I think it’s at least a fair question.

      As a side note, does anyone have an idea of what the average effect-size is of other treatments on the metabolic traits they measure. For example, if we plugged in the average effect from two-years worth of modest exercise into this study, would it show as significant? I’ve heard that even well-regulated diets take at least a year to show observable impact, so I’m not all that surprised that a two-year study isn’t coming up positive.Report

      • greginak in reply to trizzlor says:

        Reducing the financial risk/suffering from uninsured illness was always….always….one of the reasons for Uni HC. The sample sizes are actually really small and it is over a shortish time so that the many positive finding ( and there were many positive results) didn’t reach statistical significance isn’t that much of a surprise. Increased access to health care is solidly linked to improved health outcomes. If any of the conservatives who are harping on the lack of statistical significance want to put there snoot where their money is they can give up there HI, stop going to the doc, stop taking their kids to the doc and let us study their health outcomes over years.Report

        • trizzlor in reply to greginak says:

          Increased access to health care is solidly linked to improved health outcomes. If any of the conservatives who are harping on the lack of statistical significance want to put there snoot where their money is they can give up there HI, stop going to the doc, stop taking their kids to the doc and let us study their health outcomes over years.

          But this leaves out the cost side of the equation. For example, having a personal chef and physical trainer may be linked to improved health outcomes (and no one who has them would want to give them up) but that alone doesn’t justify providing every low-income home with those benefits.

          It’s conceivable, for example, that the same investment could provide better health-outcomes with a mix of the subsidized catastrophic insurance Republicans are always pushing for and regular physical/mental counselling. Just because conservatives tend to ask this question insincerely doesn’t mean we shouldn’t be prepared to answer it.Report

          • greginak in reply to trizzlor says:

            What care we should or shouldn’t offer is a really good question. Most people seem to agree protection against bankruptcy/ financial ruin is kind of good thing. But of course preventing illness or managing it before it has a massive impact on our lives is better for us. So if we want to prevent a major financial hit we should be taking care of ourselves. So preventive care/ access to docs when we need it helps prevent needing catastrophic injury insurance.

            Like i said, what level of care to offer is a good question without any special bright line to say this is to much or this is not enough. It was just a few years ago when mental health trt was often not covered by insurance or at a much lower rate. It took gov action to get it covered more fully. So why just offer counseling? If mental health is worth taking care of, why isn’t a yearly check up or regular lab work? I’d say there are two answers; one lies at some intersection of what is considered typical trt, and found to be cost effective. Two is people should be empowered to take care of themselves, so having easy access to medical care, advice and trt’s gives them the power to do what they choose to. So offering a broad range of common medical care to everybody gives people more choices and freedom.Report

      • Michael Drew in reply to trizzlor says:

        The two years thing seems like what most limits the meaning of this study to me. Say you or I were suddenly uninsured. Statistically, how much would we expect our health to measurably deteriorate as a result over the course of just two years? I would suggest that we would expect the effect to be small, and likely not “statistically significant.” I would further expect measurable health gains from people recently insured to be if anything a bit more muted even than deterioration from loss of insurance. These are of course just intuitions, but the whole point here is that the study has meaning only as relates how it compares to the right set of expectations for the program. Other than intuition, what else ought we be using to set those expectations? (That’s an honest question; I’m quite open to answers to it.) As an initial matter, my view would be that expectations for measurable health improvements from two years of Medicaid coverage ought to be exceeding limited absent reason to have higher expectations. In my view that’s not a reason to think that getting people health coverage through Medicaid isn’t a worthwhile policy aim. If two years of new private insurance coverage showed similarly limited health gains for people, would we say that that would change our basic viewpoint to one that thinks that having health coverage is not worthwhile or cost-effective?Report

        • Barry in reply to Michael Drew says:

          “Other than intuition, what else ought we be using to set those expectations? (That’s an honest question; I’m quite open to answers to it.) As an initial matter, my view would be that expectations for measurable health improvements from two years of Medicaid coverage ought to be exceeding limited absent reason to have higher expectations.”

          The important thing to remember is that this was a policy thing by the Oregon government; researchers are analyzing it for what they can get because they don’t have the funding (or the lack of ethics) to run really large-scale randomized experiments.Report

      • Barry in reply to trizzlor says:

        “The argument I’ve heard from conservatives is that if the goal is to alleviate issues of mental-health or financial suffering then there may be better and more direct ways to do that than by providing everyone with extensive health-insurance. I’ve yet to see a arguments supported by data that outline precisely what that policy would look like, but I think it’s at least a fair question.”

        Except for that very fact – that these people drop the ‘better and more direct ways’.

        Remember that much of the current Obamacare was derived from the ‘alternate’ plan pushed by the GOP (and drawn up by the Heritage Foundation) back in 1993. When the Clinton plan failed, the GOP dropped it’s ‘alternative’, because it wasn’t a real alternative; the GOP never planned on trying to enact it.Report

  2. Michael Drew says:

    The other thing to say here is that, as a result of federalism, this study tells us still less about “Medicaid” than it might if Medicaid were in fact a unified, centralized federal program (aka bugaboo). In fact, it can’t tell us anything about “Medicaid” – there isn’t any such thing for the purposes of studying public health care/health coverage intervention models. There are fifty state Medicaid programs, all with their own (or at least, varying to some degree) setups and program specifics. This study can only tell us what one state’s Medicaid program was able to do for a certain cohort – not all that large a cohort: 1,500 – of new enrollees over the course of two years. If that (see: statistical significance debate).

    Because federalism.Report