Letting the Data Speak: A Fresh Look at Health Care Cost Growth

In this post I recast the visual display of international health care expenditures. For select OECD countries, this clearly shows the growth of average costs has been moderating while U.S. cost-growth has been accelerating. The graph methodology is discussed along with a caution about marginal thinking. A conjecture is presented as to why the OECD cost-growth is moderating followed by a couple thoughts for action.

What’s Usually Presented
There are many graphs published of international health care expenditures (HCE). Some remind me of multi-colored electric cables strung together, except for one cable that strays from the group. Others are simpler but their message is obscured with chart junk. Here’s a recent example of the electric-cable chart.

This chart shows HCE trending as a percentage of GDP for a group of OECD countries, with the U.S. as an outlier. If the purpose of such charts is to emphasize how out of whack U.S. HCE is, then they work. But they don’t contribute any new insight.

A Different Approach
I wanted to see the data from its beginning, 1960, to get a better sense of trend, as well as something less obtrusive than a bunch of multi-colored squiggly lines. Here’s my version.

US_OECDhcegdpIt shows a parallel trend from 1960 to almost 1980 followed by subsequent U.S. accelerated cost-growth. It also clearly shows average OECD (less the U.S.) cost-growth decelerating. To construct this graph, I started with the original 20 OECD countries, 18 European countries, Canada and the U.S.1 There were many missing values with only nine countries having data for all 51 years. I arbitrarily picked 25% missing as a cut off, which excluded five European countries.2 A visual inspection of the three sets of data (no missing values, fewer than 25% missing, and all 20 countries) indicated it didn’t practically matter: the three curves using 9, 15 or 20 countries were visually similar.3 I opted for the 25% cut-off. A simple unweighted average consolidated the OECD countries into a mean for each year, using up to 14 values depending on the number missing. This resulted in two data points, the U.S. and OECD, for each of the 51 years. These were then averaged and based on the years 1960-1980 were used to estimate a linear trend, which was then extrapolated out to 2010 as if the early trend continued past 1980. This overall linear trend became a visual reference line. (One reason I decided to use %GDP instead of per capita expenditures is the latter is curvilinear and so did not permit a simple intuitive reference line.) Lastly, local polynomials were fitted to the U.S. and OECD data for smoothing. All of these, the pair of data points for each year, the linear trend and the polynomials, comprise the above graph.

The following difference graph emphasizes the divergence. This graph is simply the difference of the two polynomials from the linear trend line.

DifferenceChartThese two graphs show that U.S. HCE has been accelerating since the late 1970′s. Contrariwise, they also show that OECD cost-growth has been moderating. A quick review of relevant literature indeed indicates this is happening.4 At least for me, though I had previously read some of that literature, the OECD deceleration hadn’t sunk in till I saw these graphs.

A Cautionary Note
I take it for granted that contemporary OECD countries generally have better non-medical determinants of health than we do, as well as a more extensive social safety net. For example, regarding children the U.S. ranks below average in preschool enrollment and high school graduation rates while having higher than average child poverty rates. Better education and lower child poverty are associated with better population health. But they come with costs. Germane to comparative graphs like these is those costs are not included in HCE. Think of the resources that go into other determinants of health and the safety net as fixed or sunk costs, while health care costs can be thought of as marginal costs. The OECD has invested more in fixed costs than the U.S. but that is not reflected when comparing the marginal, health care, costs. The fixed v. marginal cost disparity doesn’t explain the divergence shown in the graphs, but it does mean the HCE numbers are not apples-to-apples comparable.

What Explains the Divergence?
I can’t say for certain why there’s a divergence but I do have a reasoned conjecture. Broken down by structural, conceptual and empirical elements, it is this: Structurally, OECD countries have better non-medical determinants of health, such as lower poverty rates and greater educational opportunity. This is in part because they started fresh after the devastation of WW II. (About 36.5 million Europeans were killed along with significant property destruction, e.g., 40% of German homes were destroyed. This was then followed by the European Economic Miracle.)5 Conceptually, the OECD is more systems oriented. For instance, they emphasize a Health System—not a Health Care System but a Health System. The only times I’ve heard Health System mentioned integral to the U.S. was as a joke. The OECD has made a concerted sustained effort at cost containment. But it’s not just that. For example, in some of their writings they aspire towards stewardship.6 If nothing else, they think more long-term while we are almost synonymous with short-termism. Empirically, their chronic disease prevalence is significantly less than ours, treatment for chronic diseases account for more than 75% of U.S. health care costs, and chronic disease is largely preventable.7

Here’s a chart that illustrates the chronic disease contrast between the U.S. and selected OECD countries. (Some countries are different than the 14 used above.) U.S. chronic disease prevalence conspicuously exceeds that of the other countries.

EvChronicDisDiffIn summary the conjecture is: Improved non-medical determinants of health reduce the prevalence of chronic disease. That coupled with a concerted long-term effort to control costs has moderated OECD HCE cost-growth. I am obviously glossing over many other important contributing factors (e.g., primary physician utilization, different chronic care models). However, this conjecture strikes me as potentially useful for generating policy discussion and formation and should be pursued. It also helps explain the charts’ divergence.

A Couple Thoughts for Action
In light of the divergence-plots it would be fruitful to broadly explore why the OECD cost-growth is declining, not only why ours is increasing. What’s critical is to learn from other’s successes and to be clear about what we know and don’t, and act accordingly. Let’s assume I’m close with my conjecture, that non-medical determinants of health contribute to lower prevalence of chronic disease and thus reduced HCE. Then we should move toward implementing better determinants—not just behaviors but also improving the environments where those behaviors occur. Such an investment would be potentially cost effective by decelerating our cost-growth. If we don’t, we’ll likely pay anyway with higher HCE. I’m reminded of a precept attributed to Henry Ford: “If you need a machine and don’t buy it, then you will ultimately find that you have paid for it and don’t have it.”  I think that’s where we’re at now.

One other thought to share: If chronic disease accounts for more than 75% of our HCE, then we need a standard age-adjusted Chronic Disease Index. Such an index would facilitate discussion, analyses, assessment and international comparisons. If we can create a GDP index and an YPLL index (Years of Potential Life Lost), we surely can create a Chronic Disease Index.

A Final Thought
The emphasis in this post has been on costs and benefits. Our health and well-being as a nation was not mentioned but would significantly benefit from an investment in a better life for Americans. We could think of it as a Marshall Plan for the U.S.A.

Frank de Libero is an independent statistical consultant oriented toward policy and strategy. This post originally appeared on his blog, Letting the Data Speak.


  1. The original 20 members of the OECD: Austria, Belgium, Canada, Denmark, France, Germany, Greece, Iceland, Ireland, Italy, Luxembourg, The Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, United Kingdom, United States.
  2. The five countries with more than 25% missing data: France, Greece, Italy, Luxembourg, and Turkey.
  3. For Stata users who would like to explore the 20-country %GDP data themselves, the -do- and -dta- files I used to create the HCE graphs are available here: https://dl.dropboxusercontent.com/u/4207674/anHCEOECDGDP.do, and here https://dl.dropboxusercontent.com/u/4207674/oecdhcegdp4.dta
  4. See, e.g., Chapin White. Health Care Spending Growth: How Different Is The
    United States From The Rest Of The OECD? Health Affairs, 26, no.1
  5. The estimates, 36.5 million casualties and 40% homes destroyed, was from
    Judt, Tony. Postwar: A History of Europe Since 1945. New York: The Penguin Press,
    2005., pages 17 and 82 respectively.
  6. Searching the Internet on terms like, OECD stewardship health systems cost containment, will give you plenty of material. However, if you prefer not to search, I found the following useful (of what I’ve read so far): Figueras, Josep and M. McKee, eds. Health Systems, Health, Wealth and Societal Well-being. McGraw Hill Open University Press, 2012. Available here: http://www.euro.who.int/__data/assets/pdf_file/0007/164383/e96159.pdf.
  7. The idea that higher chronic disease prevalence would help explain higher U.S. cost growth is not new. The following article by Thorpe et al. argues that position but from a behavioral risk factor perspective: Thorpe, Kenneth E., D. Howard, K. Galactionova. Differences In Disease Prevalence As A Source Of The U.S.-European Health Care Spending Gap. Health Affairs, (October 2, 2007): w678-w686. Exhibit 2 of the article gives comparative prevalence statistics.


16 replies »

  1. Most of those OECD countries have a valid form of universal/single payer type of health insurance.Herein lies the answer to that “unknown factor” that appears to be so elusive to answer.Apart from a higher poverty rate, the other OECD countries face the same public health epidemics of obesity,increasing longevity and alcoholism.
    Thanks for your insightful review of the “stats”!

  2. As I grow older and become a perceived part of the problem of increasing the cost of healthcare I cannot but look around me and see the incredible need for preventive disease strategies. in the obesity area alone, millions could be saved if older and younger people were eating correctly and exercising. A good primary preventive program is a must.

  3. *Lovely* post, thanks. Answers some questions I’d been having about why costs are so much lower in the OECD.
    Like the guy used to say in the ad,
    “You can pay me now. Or you can pay me later.”
    Still true!

  4. Brian and Al, thank you for your affirmation about the graphs and the differences they show. You also raise two issues, the chronic disease prevalence estimate of 75% and reference to established vested interests (my words here).

    In preparing this post I wondered about the 75% estimate and tried to run it down. I went to the CDC site, followed their citation, only to find another assertion. As you know, this happens a lot related to health. Regardless, I went with the official CDC number. I thought about writing in my penultimate paragraph, “If chronic disease accounts for a major portion of HCE…” but then I thought, I’m using HCE as a percentage of GDP across multiple countries. How accurate is that? Sometimes you need to go with what you’ve got, what the majority uses, just to move forward. The final point was about the need for a standard index. Ultimately that would improve the accuracy of its prevalence as well as the DNA of what makes up chronic disease.

    Brian, you wrote about the industry controlling the process. Indeed! I think about that a lot, not just regarding health care. For instance, at the top of my reading pile is Jeffrey Winter’s “Oligarchy” (recommended, BTW). I want to know what determines us. Yours as well as mine, was the originating thought for including reference to WW II: The war scattered the established vested interests opening the way to a greater degree of social and economic justice. Only afterwards did WW II work in the narrative as a home-coming device, the Marshall Plan. (I certainly didn’t plan it that way. And, of course, I’m not advocating WW III or any other war.)

    The 12 comments so far have been personally rewarding. They have been thoughtful and authentic — and encouraging. Thank you all.

  5. Jane, It is not the lack of primary care, it is the absence of primary prevention that is the differece between these healthcare systems.
    If not the government or physicians, how can we promote prevention?
    Every science and medicine starts with measurement. Wellness and health should also start with valid measurements. When your health is measured, you have a clear direction to work from. The problem with available tests are that they measure disease not wellness or health.
    Linking Epigenetic Modification to Wellness!
    Wellness Testing provides the ability of the individual to link their actions to the harm they are causing themselves and their loved ones. The current lack of a test that provides ones personal health yardstick has been the primary roadblock to improved wellness and natural cures. This testing could identify your over worked liver long before it has had time to develop into a disease state. At this early stage, natural remedies such as herbals and other supplements may be taken.
    The assumption is – when people are provided with the appropriate knowledge (incentive), they will adjust their behavior to maximize their wellness (reward).
    Using new testing protocols that measure epigenetic biochemical changes, an individual can monitor if they are meeting their wellness goals or if their intervention is working for management of their disease. Wellness testing is not dependent on the intervention used. For example, the client could have taken a drug, used an herb, practiced yoga, or used group prayer as the intervention. Wellness monitoring can also show if your wellness is degrading over time even if you are asymptomatic.

  6. Simple economics in free market is supply and demand. Increasing supply or lowering demand reduces costs. At least since 1980 th eer ee has been continuous efforts by suppliers to control the market. Hospital consolidation and legalized insurance company collusion are two examples. The payment system is used to exercise these controls. Wealth cation has priority over patient care. Efforts to reduce costs and develop more effective means to manage patient care are useful. Unfortunately they will have miniml impact on actual health care costs. Revising the payment system will remove the wealth.creators ability to control the market. Eliminate payment for services and products in its stead payment is for access to products and services. The free market will be the invisible hand that drives down costs and simultaneously improves the quality of patient care.

  7. Yes–the statistical analysis makes a lot of sense to me, but when you enter the realm of speculation on this 75%-on-chronic-disease bromide you lost me. That is one of the most misleading figures in healthcare — see my post from a couple of days ago.

  8. Frank,

    Your central chart, showing the divergence between US and OECD HCE as a percentage of GDP, is a significant contribution. I also strongly agree that we undervalue the health impact roles of social welfare programs like early childhood education and anti-poverty subsidies.

    This said, relying on Ken Thorpe’s and the CDC’s figure that 75% of all US health care cost is founded in chronic disease probably isn’t a good idea. (See Al Lewis’ discussion – https://thehealthcareblog.com/blog/2013/04/14/is-patient-engagement-the-solution%E2%80%A6or-a-healthcare-urban-leg/) My colleagues and I actively involved in medical risk management (or as you referred to it, cost containment) have found significant unnecessary and inappropriate cost in rank-and-file non-chronic patients due to the system’s distortions – e.g., egregious unit pricing and overutilization – which both disproves the 75% figure and suggests that nearly all US care is inflated due to economic drivers as well as lifestyle determinants. It is important to consider both in your analysis.

    The larger point is that, because the industry has shaped policy through its lobbying influence, it is very difficult to dislodge these trends. Not only are you correct that a great focus on social welfare would attenuate the lifestyle costs we develop, but changing what and how we pay for care, possibly following the OECD countries’ lead, would yield far greater health care value.

    We won’t do that, though, because the health care industry control the process, and they like the economic result delivered by the current approach.

  9. I found a reference to the policy change in this article http://www.nationalaffairs.com/publications/detail/saving-medicare-from-itself, although what I remember reading awhile ago ewas a detailed analytical paper on the specific change. In any event, here is the reference:

    “In 1983, under Ronald Reagan, HCFA imposed a “prospective payment system” whereby hospitals and physicians would be reimbursed at a set rate for a specific diagnosis — Medicare’s first price controls.

    But after a few years, hospitals and physicians grew wise to the new system, and found ways to shift patients from poorly paying “diagnosis-related groups” to higher-paying ones — a practice called “upcoding.”

    The paper I read linked the increased growth rate in Medicare payments to this specific change.

  10. My recollection (I have been unable to track down the article) is that a Medicare reimbursement change during the Reagan Administration designed to control costs actually backfired and that is when costs (at least the Medicare side) accelerated. I’m continuing to search for the reference.

  11. This is a nice analysis and makes a good contribution. It’s too quick to draw the conclusion that disease is the main culprit. Whatever happened to “it’s the prices, stupid?” It’s still true that our prices are twice as high as other nations for the same services.

    I would love someone to show how to apportion the contribution to the total excess US cost by disease prevalence vs prices.

  12. What bird says above is false. If he’s referring to managed care, that didn’t really pick up steam until the late 1980s, and it wasn’t until the early 1990s that managed care (HMOs, PPOs, POSs) became dominant nationwide. US costs started accelerating long before that.

    BUT, bird really is confused about what managed care changed. Yes, it moved from a deductible/coinsurance model to a primarily copay model. But it did not bring about the attitude that people can “get what ever they want and see the doctor whenever they [want]” Famously, or infamously, managed care brought more restrictions on care in order to get covered by the health plan. In the old fee for service major medical world, people paid mostly out of pocket and didn’t need any prior approvals, prior authorizations or specialist referrals submitted to their health plan. Post 1993, most people did. The period 1994-1997 was actually the one flat part of the curve for the US. Then managed care retreated, and costs went up again. So, if you want to substitute “1998” for “1980” in your analysis, it has a semblance of truth, but only that. You still haven’t explained the 1980s, and still haven’t explained why other nations that have a copay based model did so much better on costs.

    But I’m not holding my breath on bird’s mea culpa.

  13. Frank,
    Thanks for a very informative analysis here. I think your argument could be strengthened by showing the trends in the prevalence of chronic conditions rather than a single year cross-section as shown. We know that the prevalence of chronic conditions has been increasing for some years.

  14. Until both we and the OECD nations recognize the underlying drivers of health outcomes we are not going to solve this growing problem

    To acertain what really drives health outcomes in all developed nations see http://www.unnaturalcauses.org

    Dr. Rick Lippin

  15. in 1980 patients were conditioned to be able to get what ever they want and see the doctor whenever they for a 10 dollar copay, we created a generation of entitled worriers, then the shift to fee for service came and utilization continued.

  16. Frank, terrific, smart and inventive analysis here. Bravo! I would add one more important difference between OECD and US health systems: most of the higher performing health systems have a strong primary care backbone. In the US, our primary care infrastructure of PCPs has dwindled due to lower pay and valuation, and greater emphasis on incentives for specialty and volume-based care. The emerging primary care platforms of retail health, pharmacy-as-primary care deliverer, more NPs and PAs, and worksite clinics can help fill in the gaps. But the risk is greater fragmentation and dis-continuity of care. Thanks again for doing the heavy lifting on the stats…the “differences” chart is enlightening. JSK