• The most important chart in health policy

    Like it or not, the ACA will be judged, at least in part, by the degree to which it contributes toward reducing the rate of growth in health care spending. There are reasons to be optimistic that the law will help to “bend the cost curve”: as Henry Aaron put it, the ACA contains “virtually every idea for cost control that any analyst has come up with.”

    There are reasons for pessimism too: the success of the law depends on the body that has the power to destroy it, Congress. If Congress is unified, or reasonably so, in preserving and strengthening the cost control mechanisms in the law, the chance they will work increases considerably. If, however, Congress, or even a significant minority of its members, wants to delay, weaken, or repeal components of the law, success in cost control is far less likely. It’s no insignificant point that the law may be declared a failure by the very individuals who fought to undermine it. Would that really be a failure of the law?

    If, by action of politicians or the market, the health care spending curve is not bent, one might argue that this reflects our collective desire, our revealed preference. The history of health care spending in the US is consistent with the hypothesis that we view health care as a luxury good, one on which we spend more of our wealth as that wealth grows. A new paper in Health Economics by Robert Woodward and Le Wang, titled The Oh-So Straight and Narrow Path: Can the Health Care Expenditure Curve Be Bent?, illustrates that point (see their chart below). The great question Woodward’s and Wang’s paper suggests to me is that if health is a luxury good, why should we expect the health spending trajectory to change?

    The chart shows that on a log-log scale, the relationship between per capita national health expenditures and per capita GDP is ram-rod straight and has been since 1935.* Based on several statistical tests, Woodward and Wang conclude that there are

    no significant changes in either the intercept or the slope for the years of any of the major reforms of the period, including the introduction of Medicare and Medicaid in 1966, in the years following Nixon’s Wage and Price Controls and the health planning legislation of the early 1970s, in the years following Medicare’s introduction of the Prospective Payment System in 1984, or after the widespread adoption of managed care in the 1990s.

    With such overwhelming evidence from our past that, despite our efforts, this national health spending curve has not bent, why do we think the future will be much different? It is not unreasonable to think it will not be. But, if we presume the curve will not be bent, what then are reforms able to do? I think the answer is twofold. One is that they might shift the level of spending, but not the rate of growth. If the level is shifted gradually over time it will appear as if the growth rate has changed. The long view is likely to show that it has not.

    The other thing reform could do is change what we receive for our spending. If observable measures of population health change through the change in payment incentives without a change in spending, reform advocates will still claim victory, as they should. In a narrow sense, this is a bending of the curve–the “cost per quality” one. Spending more of our wealth to confer better care on more people is not without value.

    Woodward and Wang note that we may be receiving other things for our additional spending: “hope, uncertainty‐reducing information, and amenities.” None of these may actually improve health, but they are still valued consumption components of health care. In other words, the view that health care is entirely an investment in health is, perhaps, too narrow. We may rationally value additional spending on it if, by doing so, we buy hope, peace of mind, and other comforts that we value more than the dollars they consume.

    Politics and entrenched interests have a lot to do with how health care dollars are spent. But, as Woodward and Wang suggest, so does the fact that we have come to expect more for our health care dollar (or someone else’s) than just health. Bending the curve will require breaking our expectations. So far, I have not seen good evidence that we are willing to accept what that means.

    * The authors sent me this figure, which, due to an error in the publication process, is not the one in their paper but is the one that should have appeared there. There is no qualitative difference between the two figures or the conclusions drawn from them. (I think the only difference is axis labeling.)

    UPDATE: Revised figure from authors.

    Share
    Comments closed
     
    • Austin
      A perfect adjunct to this paper is this Steve Schroder citation from April, I believe you made mention of it as well in another post:

      http://archinte.ama-assn.org/cgi/content/extract/171/8/722

      CONCLUSION
      “In addition to the medical cost containment strategies tried in the past, recent efforts include dissemination of electronic health records (to reduce wasteful duplicate services), curbing fraud and abuse, paying for performance, and comparative effectiveness research—all reasonable things to do. Yet, given our past performance, it seems naïve to assume that these latest efforts will be any more successful than their predecessors. In the long run, reining in costs will require mobilizing political forces that can withstand the inevitable claims of rationing sure to come from the industries currently benefiting from the 17% of the economy spent on health care, and from consumers who have come to expect unlimited access to what they feel they need. Until there exist sufficient countervailing forces so that a comprehensive, multipronged strategy could be implemented, politicians and health policy experts will continue to embrace tepid and ultimately ineffective solutions that may sound good in theory but will fall short in practice.”

      Brad

    • I dont understand how this graph can be so straight. makes me wonder what they are measuring if medicare enactment did not change the slope of the graph.

      one thing to remember is that the health care industry is like the oil industry. it is actually a lot more than just a visit to the doctor and how you tally up health care expenditures is propaganda fodder. Since there is a chain of of economic transactions (each with its own mark-up) this is the kind of graph you’d expect. A larger gdp would therefore generate a larger fraction of health care expenditures.

      • Log-log plot means NHE growing exponentially with GDP. http://en.wikipedia.org/wiki/Log-log_plot

        There are some wiggles. It’s the long-term trend that is so strikingly straight (exponential growth factor constant).

        • Would’t you still expect a bigger jump in per capita expenditures between 1960 and 1970 because of Medicaid and Medicare than would have been predicted by the increase in GDP per capita, unless part of those programs paid for what people previously had been payng for themselves and/or the number of people eligible for those programs was initially to small to produce a dramatic jump in expenditures?

          • NHE, GDP, and population data are publicly available, so anyone could reconstruct this figure.

            https://www.cms.gov/nationalhealthexpenddata/02_nationalhealthaccountshistorical.asp
            http://research.stlouisfed.org/fred2/series/GDP
            http://www.census.gov/popest/archives/

            Authors citations for the figure:

            Gibson RM. 1979. National Health Expenditures, 1978. Health Care Financing Review 1: 1–36.
            Hartman M, Martin A, Nuccio O, Catlin A, and the National Health Expenditure Accounts Team. 2010. Health spending Growth at a historic low in 2008. Health Affairs 29: 147–155.

          • It is surprising indeed that Medicaid and Medicare had no permanent effect on the slope or intercept of the log-log curve. My sister was a young MD in Ann Arbor at the time and continues to tell stories about the rush in ER visits following Medicaid’s implementation.

            But we know that such stories, our intuition, and even our training are not always totally right.

            Indeed a detailed look at the period from 1960 to 1975 suggests more of a slowdown in 1966 (perhaps related to institutional uncertainty about the governmental regulations) than an increase in 1967 to1969. (I’m using NHE, GDP, and POP data from the NHE2009 and NHEGDP2009 files from CMS at https://www.cms.gov/nationalhealthexpenddata/02_nationalhealthaccountshistorical)

            The fundamental question is whether the comparative statics explanations we’ve all adopted (insurance increased so demand increased) of the long term dnamics (rising expenditures as a percent of GDP) is adequate for the task. Most of us have assumed the comparative statics are adequate. These data imply that there are other dynamics at work.

    • thank you austin for that reference. it’s interesting that a curve, y=x gives a straight log-log graph with a slope of one. the curve above has a slope of 1.5. The population of the US grew by more than 50% over that time period. So actually, less is being spent on health care per capita. Also, The same growth can occur if more people work in health care as a fraction of the population and this number is a large multiple of what it was before medicare.

      • The curve above is log(NHE per capita) vs log(GDP per capita). The slope, as reported in the paper is 1.388, which is the per capita health expenditure-income elasticity.

    • The graph above could also result from health care becoming a larger part of our economy as other parts (manufacturing) migrate abroad. If our entire Economy was health care, the graph above would have a slope of one. The curve will eventually bend over as health care grows . Problem solved .

      • We all know that healthcare is becoming a larger part of the economy. But there are interesting questions associated with extrapolating the 1.388 health expenditure elasticity into the future. Actually this exponential growth model causes NHE/cap to cross GDP/cap sooner than an extrapolation of the average annual 0.25% increase in the NHE/GDP ratio.

        Health policy analysts tend to portray GDP as if it were a fixed pie. In this mental framework, spending more on health means spending less on everything else.

        Our Health Economics Letter mentions, but does not fully develop because of space limitations, an alternative mental framework where NHE is a growth engine for GDP. If one considers the historical relationship between log(NHE/cap) and log[(GDP-NHE)/cap], the slope is even steeper. But GDP is then the sum of NHE plus GDP-NHE. In this context, if NHE grows faster than the rest of GDP, the growth in GDP will be influenced more and more by the growth of NHE. Moreover, NHE will never catchup to GDP and spending on the rest of GDP will never be adversely affected by the growth in NHE.

        • With due respect, this is a very weak analysis to say the least. Health policy analysts don’t portray GDP as a fixed pie – it’s wrong. What they claim is that that growth rate of health care expenditure is greater than that of GDP means health care expenditure is rising as a ratio of GDP – remeber GDP can still grow. Second, as for your implication that the higher growth of health care expenditure is good for GDP since it’s a larger part of the latter, it’s good as long as other sectors in the economy grow at least as fast as does the health sector. How ever, this is a wild dream as sources of productivity in other sectors are quickly exploited leaving growth at modest or insignificant rate; in contrast, due to climate change, global warming, mobility, aging population, new cases of disease, health care expenditure is more likely to grow faster.

          Best regards

          • When accused of thinking of GDP as a fixed pie, I agree that health policy analysts deny having such a simplistic thought. But they then go on to define the rising proportion of GDP spent on health care as a crisis mostly because it’s a rising percentage. But the rising percentage is only a problem if the increase is achieved at the cost of some other sector, an assumption implicitly reflecting “fixed pie” thinking.

            I also disagree with your second point. I see nothing that’s necessarily “bad” if health care expenditures are growing more rapidly than the rest of GDP, at least if the rest of GDP is generally “healthy” by at least some measures.

            I, at least, am unable to predict where or whether productivity increases will occur. But while productivity in any one sector or technology may experience diminishing returns, our recent economic history reflects a series of productivity gains in diverse (and unanticipated?) sectors. Nevertheless, the (GDP-NHE) model was only meant to illustrate a not-entirely-unreasonable alternative to the implicit “fixed pie” thinking.

            • Sir,
              I still disagree with you about your idea of health analysts’ assumption that GDP is a fixed pie. Implicitly or explicitly, what they claim and what you stated don’t match. On the second point, it was my bad to wrongly state productivity differences between the health sector and the rest. As a correction, the productivity growth in the health sector falls short that of other sectors like manufacturing sector. This means that the growth of health care cost is higher relative to others. Though I don’t wholly buy the Baumol’s disease – the supply-side matter – which you may look at, it offers some good insights about health care cost. But, more importantly, the implication of this second point is closely related to the first point that you claim that health analysts implicitly assume a fixed GDP.
              With best regards

    • Quotes from paper relevant to observations that there is no visible sign of Medicare implementation in the graph.

      “Of course, regressing per capita NHE against per capita GDP creates an obvious endogeneity problem as NHE has now become a significant part of GDP. Although the DOLS is consistent even in the presence of this endogeneity problem, we supplement our basic result by considering the relationship between per capita NHE and per capita GDP without NHE. In this context, the DOLS and Johansen method expenditure–income elasticity estimates increase slightly to 1.445 (P≤0.0001) and 1.349 (P≤0.0001), respectively.”

      “The statistical stability of these income–elasticity relationships can be graphically illustrated in two ways. First, simple extrapolations of trends observed in the years before Medicare and Medicaid are surprisingly accurate. For example, if one extrapolates the relationship between 1950 and 1960 to the per capita GDP in 2008, the error in predicted per capita NHE in 2008 is 8.6%. Or if one extrapolates the trends defined by the relationship between per capita NHE and per capita GDP in 1929 and 1965 to the per capita GDP in 2008, the error in predicted NHE in 2008 is 11.8%. Although the selection of other years does not always produce equally close estimates, the accuracy of these estimates is striking given that per capita NHE in 2008 was 52 times that of 1960.”

      “Second, we used truncated data from 1960 to 1979 to estimate the equation ln(NHE/cap) = f[ln(GDP/cap)] and then compared the forecasted ln(NHE/cap) with the actual values (Figure 3). In none of the years following 1960 does any actual value fall outside the 95% confidence limits. And although 1955, 1940, and 1935 do fall outside the confidence limits, the model’s prediction for 1929 does work. Additionally, the average percent errors for the 1960 and 1979 periods are only 2.7% and 3.4%, respectively.”

    • “One is that they might shift the level of spending, but not the rate of growth. If the level is shifted gradually over time it will appear as if the growth rate has changed. The long view is likely to show that it has not.”

      OK, clue me in, here. How does the level change, “gradually over time” or otherwise, if the growth rate only “appears” to change? I was under the impression that if the level changes more or less in one period than another, there must be a different rate of change. Whatever you were trying to say about levels and rates, you haven’t said it right.

    • Any predictions that assumes that exponential growth in anything will continue indefinitely is always going to be wrong sooner or later.

    • …what we receive for our spending.

      Employment, or more broadly speaking aggregate demand, is another benefit from increased spending on healthcare. No other sector of the economy can easily create millions of semi-skilled jobs over the next decade.

    • I just looked at OECD data for 1970-2009, 28 countries
      and ran the regression ln(NHE/cap) = f[ln(GDP/cap)
      Here are the slopes:

      Turkey 1.595
      Portugal 1.488
      Chile 1.405
      United States 1.399
      Belgium 1.370
      Switzerland 1.368
      France 1.346
      Mexico 1.338
      Iceland 1.335
      Spain 1.331
      Austria 1.320
      Italy 1.304
      Greece 1.302
      New Zealand 1.270
      United Kingdom 1.261
      Australia 1.243
      Norway 1.235
      Korea 1.224
      Canada 1.215
      Japan 1.215
      Netherlands 1.181
      Finland 1.164
      Luxembourg 1.162
      Germany 1.111
      Sweden 1.094
      Denmark 1.073
      Ireland 1.044
      Israel 1.011

    • i think that NHE vs GDP is a useless statistic because it does not inform how we can improve our health or our economy. At best, it is a propaganda tool for hysteria and obfuscation. Such statistics are equivalent to asking about the color blue when looking at the earth from a satellite.

      To assess the value of where we spend our dollars, a more useful and relevant statistic might be NHE for intrinsic diseases like cancer, kidney stones or accidents vs immediately correctible diseases like cardiovascular disease (obesity, tobacco, inactivity). This is an assessment that predicates that value is based on correctible health.

      But today, we often hear of the value of where federal money is spent as the number of jobs created. Even using this metric, a better graph might be Jobs created vs NHE. Drilling down into the underlying factors of such a graph would help us understand the dependent and independent components of NHE that lead to jobs and those that do not(and perhaps eliminated)

      I might add (at the risk of being off topic) that jobs are not really created by money. (Although they can be, this is artificial) Jobs are truly created by perceived value which is negotiated between the provider and the supplier. Government money is best spent on providing the venue for the transaction rather than the transaction itself. Consumers need to be able to put health care expenditure back into the context of their lives like buying a house or buying a car.

      • I agree that observing the close correlation (R-sq = .999) between the log of NHE/cap and the log of GDP/cap provides no evidence on how to improve health or health care expenditures.

        I strongly disagree that the observation is a propaganda tool for hysteria and obfuscation. Just the opposite. Since the early 70’s, we’ve been listening to Chicken-Little types making careers out of predicting the sky is falling because NHE/GDP ratio might get as high as 10%. The observation that the relationship has been a constant 1.39 or so in the US since 1960 should induce a sense that it might be time to more calmly amplify our list of causal factors.

        I agree with the value of jobs “truly created by perceived value” and note that our publication suggested several hypothesized sources of perceived value from health care services apart from improved health. These include luxury amenities, risk-reducing knowledge that doesn’t improve health, and hope.

    • There is a pretty simple explanation to trends in health care.

      1. The wealthier we are, the more we can spend – who wouldn’t extend their life if they could afford it.
      2. Technology advances don’t usually decrease costs because we tend to simply consume more medical serivces.
      3. The labour productivity of some services is static – nurses might be able to monitor patients better, but now they monitor more things, leaving the same patient nurse ratio.
      4. The longer we live, the longer we are old, and being old means lots of helath problems. We all die someday, so prevention of one disease only leads to another one later on.

      I have discussed these things before at my own site.
      http://ckmurray.blogspot.com/2011/06/smoking-decreases-health-costs-to.html
      http://ckmurray.blogspot.com/2008/07/health-care-economist-environmentalist.html

      For me, the question in the US (I’m Aussie) is why procedures cost so much more than anywhere else in the world? Are doctors unions to blame in keeping places at medical school lower than necessary? Is the insurance structure incentivicing overuse of services? Etc.

      Why the US does not simply borrow an existing working model for health care from Europe astounds me.

    • I can tell a story for why Medicare and Medicaid didn’t have a significant impact on total spending in this graph. We were rapidly growing our economy in the 60s at the same time we were rapidly growing HC costs, and some of the new Care/Caid expenditures simply displaced old expenditures that would have happened anyway.

      What I have a harder time understanding is how the curve can be so flat in the 90s during the prime managed care years of about 1993 to 1999. Though we had a rapidly expanding economy in the mid-late 90s, the rate of health care cost growth as a ratio of GDP did very significantly decline. We even had no growth at all in the ratio for a year or two. Why isn’t this showing up, even as a blip in the graph? Could it be that the use of logs is obscuring a real difference? Log graphs do tend to lessen the visual impact of changes in two variables in relation to each other and flatten out curves.

    • @cmurray

      There are no doctors unions that can be identified publicly here in the US. Tuition (~50k annually) is an example of the outrageous pricing structure that is at the tip of the medical industrial complex. It seems modeled on the military industrial complex where certain corporations have managed to pervert that part of the economy as well. It seems that the common sense of microeconomics is irrelevant to large industry that operates in the realm of macroeconomics. But is off topic.

      The point I made earlier is that a significant part of GDP includes nhe and that is why the line is straight. But it doesn’t matter because it’s a tautology .

      • You are correct that in the usual presentation of the subject (which we’ve not modified in this aspect) NHE is included in the GDP. You are also correct that this becomes more of a potential problem as NHE becomes a larger percent of GDP. But that does not quite make it a tautology.

        With NHE included in GDP, the health expenditure income elasticity is 1.39ish and the R-square is 0.999.

        With the NHE excluded from GDP, the health expenditure income elasticity is 1.45 and the R-square is 0.998.

        The graph looks remarkably similar. I don’t seem to be able to insert it myself, but Austin will be able to insert a copy.

    • As I just wrote on Google+, I think Woodward’s and Wang’s key point is getting lost in the focus on the details of the graph:

      “At its heart, high and rising health care spending may be more culturally or psychologically driven than economically or politically. Sure, economics and politics matter, but one has to ask why? And why for so long and with essentially the same outcome? Maybe we have to look more at ourselves, within ourselves, than at the system.”

      G+ link: https://plus.google.com/118292000513305131142/posts/Ssz1hrD7NZC

    • It is simply untrue to say that the PPACA contains “..virtually every idea for cost control.” The core of our enormous US costs is a fractionated, multi-payer system, fraught with more billers than clinicians. Our US health expenses pay for astonishing layers of profit-taking, marketing, advertising, and commerce.

      The PPACA will not bend any cost curves, because it is built upon what we already have: a failed multi-payer private finance system. Failed at providing care, that is, while continuing to be successful at controlling legislation in the US.

    • Why must we measure health care spending in absolute terms? We do not measure oil in “national crude oil expenditures per year,” we measure in “average price in a given year.” Obviously, it’s harder to break down healthcare into units in the way a tangible commodity can be broken down, but the point is often mentioned only parenthetically.

      Rising *marginal* cost of health *outcomes* is the crisis, as we spend far more than other countries, have a lower life expectancy, and our spending is rising much faster than life expectancy, disease survival rate, etc. Total spending is a red herring. You clearly pick up on this in the blog, but we all ought to be yelling about it at the tops of our lungs, because I’m not sure our lawkmakers understand this poitn.

    • I don’t think this graph tells us about luxury at all. If healthcare was a commodity, then yes, increasing spending would suggest just that.
      But the driver of GDP is scientific progress. And, surprise, the driver of healthcare advances is also scientific progress.
      You get to pay for a lot more through the years, including the wages for people like nurses and doctors, who happen to not be willing to be gradually forced to reduce their purchasing power, so their salaries are anchored to yours…

    • However, the U.S. is still an outlier in the sense that we spend more than one would expect considering wealth alone:

      That is true but the question remains, is it because we started from a higher base or because we had greater growth? Also if healthcare is funded through taxes the expense of the tax needs to be added into spending.

    • It is therefore readily concluded that health care expenditures move independently of the country’s GDP. Over and above the fact that it illustrates that health care expenditure inevitably grows outside of other economic influences.