• Chart of the day: Health care spending by age and country

    UPDATE: Chart deemed bogus. See comments. Rest of post altered to reflect that fact, including a big red “X” through the chart. I’ve left this post up to document the problems with the chart. If you use the chart and don’t mention those or refer readers to the comments below, you’re acting irresponsibly.

    Dan Munro has a post up at Forbes: The Year In Healthcare Charts. Though they’re all worth a look, this is the one that dilated my pupils:

    HCcostsbyAge-X

    This is, apparently, from a year 2005 study. There are many problems with the chart, as raised in the comments.

    @afrakt

    Share
    Comments closed
     
    • While looking at age can be interesting, I think using cohorts could be better…What do you think would occur if the study used cohorts rather than straight age?

    • I am not disputing the overall qualitative result but two things I could not find in the Kotlikoff study which worry me for interpreting:
      a) at older ages, need to check whether LTC costs included. Medicaid LTC costs appear included, not sure for other countries. Makes a difference at older ages.
      b) these appear to be averages. Medians would probably be closer between U.S. and other countries and I would conjecture that catastrophic expenditures boost these differences in means.
      This does not take anything away from the importance of this chart for moving forward in understanding cost differences. To disentangle disease prevalence from intensity of care I think we would need these profiles for people with the exact same health conditions. Very hard right now to get good micro data on total expenditures (there is some on out-of-pocket).

      • If they used OECD data, this should include long-term care. I’ve not used OECD data personally, but I know that it does include long-term care (both nursing facilities and community-based care). the McKinsey report used OECD data and referenced both institutional and community LTC.

    • One potential causative factor for the curve spike you mention—the much bigger focus on preventive care in the US.

      I practiced in the US, India, Poland and the UK at various parts of my medical training and there is an amazing and unreal level of focus in the US on:
      –routine physical checkups
      –multiple followup visits
      –routine screening for multiple diseases
      –lab panels (many of them theoretically unnecessary but encouraged by patients who wants active doctors, the physician community that wants to avoid “mistakes”, legal pressures, and payors)

      In contrast in other nations, preventive care seemed much more targeted (especially in Europe) and so much more effective than our approach where universal screening, etc. are encouraged. This certainly has downstream effects as the potential of incidental findings is much higher in older patients.

      • V,

        While I think your basic point is excellent, I think you have missed one of the key cost drivers. I once asked my doctor why I was having a blood test every two months – his response was “because your insurance will pay for it”. I think you overstate “patients wanting/demanding care” and under appreciate how much of our costs are driven by big pharmacy and insurance companies working the system to their mutual benefit.

    • I’d be interested in a look at how these selected countries compare with Japan. Their new government has pledged cutbacks on the amount of care spent on their aging population, wouldn’t population demographics have a noticeable effect on these curves?

      • FWIW, as a customer of the Japanese system, other than regular dental work (yes, it’s covered here), check ups, and flu shots, I had almost no interaction with the system before 55. Since 55 (well, 57), I’ve had a detached retina, high blood pressure, major back pain (fixed by a few weeks of physical therapy). It’s been a wild two and a half years. And a real lesson in getting old being by far the worst thing that could happen to you other than not getting old.

        So I strongly suspect that this effect is being suppressed/understated for other countries, at least as much as it’s being overstated for the US in the graph in this article.

        In terms of demographics, the US has the lowest ratio of retirees to population of the industrialized countries, and will continue to for the forseeable future. The US may never get to where Japan is today, and Japan will be getting very much worse very rapidly. So they’ll be asking for more money from the elderly. Fortunately, medical costs are 1/3 what they are in the US, so it will merely be painful, not excruciating. By the way, in the new government’s recent resounding electoral victory, they received _fewer_ total votes than they got in the thorough trouncing they received in the elections 3 years ago. (The opposition party imploded: Ichiro Ozawa left the party with 60 of his followers, every single one of whom was defeated.) The LDP is feeling very pleased with themselves and is returning to business as usual for Japan, Inc.; tax policies that favor the rich and corporations, major spending on infrastructure to prop up the domestic economy, and if it’s good for Toyota and Sony it’s good for country. The opposition is regrouping, and if the LDP is seen as overreaching, they may be out again. Maybe. Exciting times ahead here.

        To pull my weight with real data, the following PDF has a graph (figure 1, page 54) for Japan similar to the one in this article. (Note that “social security” doesn’t mean “Social Security”!)

        http://www.ipss.go.jp/webj-ad/webjournal.files/socialsecurity/2004/Dec/Fukawa&Izumida.pdf

    • The chart is rather misleading, as it is representing categorical data as continuous.

      From what I can tell from the working paper – http://sws1.bu.edu/kotlikof/Who%20is%20%20going%20Broke,%20August%2029,%202006.pdf – as well as from the chart in the linked story, the data is represented by age groups 50-64 and 65-69 (and other, of course)… and the jump occurs in the 65-69 group.

      The published paper in 2009 – http://www.kotlikoff.net/content/who%E2%80%99s-going-broke-comparing-growth-public-healthcare – does not have the same charts, but I will include it here for completeness (and I will hold my tongue on that 3D dataviz).

    • Maybe in other countries they accept the onslaught of aging better. I’m 56, and just this year had three injections for sciatic pain, and one shoulder surgery for a torn rotator cuff. That’s about $35,000 total right there for one year, and I’m thinking about having back surgery this coming year.

      I was surprised at the $24,000 cost of shoulder surgery. I looked online before I had it done, and it seems like it usually costs more like $15,000. The steroid injections in my back cost around $3,300 each in an outpatient clinic. They seem to offer only incomplete and temporary relief from degenerative disc symptoms.

      Two MRIs cost about $450 each. I don’t know if that’s a reasonable price. I have good insurance this year, paid 100% after a $500 deductible, with no rejected services. Never had that happen before.

    • I cannot verify this chart, which seems not to have been published other than in a newspaper article by one Paul Fischbeck, whose Internet database (DeathRiskRatings.com) I was unable to locate. Is there another source for this chart’s data?

    • This is a bogus chart. It takes a lot of searching to find the original data. The links reference Forbes, which references the Pittsburgh Post-Gazette, which references Prof. Paul Fischbeck. However, I could not find his publication (it may not have been published). The Post-Gazette also states that Prof. Fischbeck got his data from a 2005 study by Boston University economist Laurence Kotlikoff, though they misspelled his name.

      The original charts in the Kotlikoff study are for 10 countries, for different years from each country, and, most importantly, THE DATA IS ONLY FOR GOVERNMENT SPENDING. From the Kotlikoff study: “…for age groups under 65, the average values of government health expenditures used to form the U.S. profile are averages over the entire population at a particular age, including those not eligible for Medicaid and, therefore, receiving no benefits.”
      Accordingly, it does not include the lion’s share of healthcare spending for those under 65, skewing the chart.
      (see: http://www.nber.org/papers/w11833 for the original paper).

      The Kotlikoff paper did not chart them together presumably because the data was from different years, currencies, etc. It also mentions that of the 10 countries studied, the US was at one extreme, Austria, Germany, Spain, and Sweden at the other, and Japan, Norway, the UK, Canada, and Australia are in between. This chart conveniently removes the in between countries to make the US appear more extreme. I don’t know how Prof. Fischbeck chose to normalize the 5 cherry-picked countries for one chart, but it doesn’t appear he had data do do so accurately.

      This is the problem with policy debates today. It fails to even get accurate information at the start of the debate.

    • This chart is indeed eye-popping, and the underlying message of the chart could well be true, but I think some caution is warranted.

      As described in the newspaper article you linked to, this figure is based on calculations by Professor Paul Fischbeck of Carnegie Mellon using data compiled by Laurence Kotlikoff and Christian Hagist. It appears that the underlying Kotlikoff paper is this one: http://www.nber.org/papers/w11833. It was subsequently published in a Spanish journal with which I am not familiar.

      Having taken a look at this paper, I have two main concerns:

      (1) The underlying data on countries’ age-spending profiles were obtained by Kotlikoff and Hagist from 10 different country-specific studies. The authors do not report having done anything to confirm that the various studies all cover the same types of spending. In fact, as nearly as I can tell, they do not even acknowledge the possibility that the figures from different studies might not have comparable coverage.

      This could be a huge problem. Suppose, for example, that long-term care was considered “health care” in some of the studies but not in others. The relative spending of the elderly would then be dramatically understated in some countries relative to others, badly skewing cross-country comparisons. Due to time constraints, I cannot slog through the 10 different studies myself, but I think this merits taking the results with a grain of salt.

      (2) The chart shown covers only half as many countries as were included in the Kotlikoff and Hagist study. Because the details of Fischbeck’s calculations using the Kotlikof and Hagist data are not available anywhere as nearly as I can tell (the newspaper article linked to above provides no details, and the website referenced therein no longer exists), I have not been able to precisely replicate Fischbeck’s approach for the other countries.

      Looking solely at the data reported in Kotlikoff and Hagist, however, it appears to me that the selection of the five countries shown here modestly overstates the extent to which the US is an outlier. In particular, Canada is included in the Kotlikoff and Hagist work, but not here. But Kotlikoff and Hagist work indicates that the age-spending profile in Canada looks much more like the US age-spending profile (in shape, not in level) than the other countries listed.

      In sum, I agree that how the shape of age-spending profile varies across countries is an extremely interesting question. I am just not confident that we know the answer to that this graph provides a reliable answer to that question.

    • The chart is misleading. Kotlikoff et al.’s paper looks at PUBLIC spending only. There is very little public spending on adults in the US until they enroll inMedicare (mostly at 65, but quite a few disabled enrollees in the 55-64 age group). That’s what produces the dramatic leap in the US data. But the pattern does not hold at all once all spending is considered. See Louise Sheiner’s much more careful and comprehensive paper at http://www.federalreserve.gov/pubs/feds/2004/200414/200414pap.pdf. The US is right in the middle of the range in terms of the ratio of elderly to non-elderly spending, and on the low side for the oldest old. Of course, the levels at all ages are,ich higher in the US.

    • This raises some questions.

      Per capita healthcare costs are a function of a) # of units of services, b) the types of services (e.g. MRI/CAT scans, CABGs, prostatectomies, etc) and c) the average health status of the beneficiaries.

      We know that on c), we aren’t noticeably worse off than the other countries in the chart. Austin attributes the difference to more care.

      But if you read McKinsey’s excellent study of why the US has higher healthcare costs, it might be more complicated. McKinsey attributes most of our higher healthcare costs to higher total inpatient and outpatient total costs (the latter includes physician office visits, same-day hospital visits, ambulatory care center visits).

      On the inpatient side, we have fewer admissions and shorter lengths of stay than OECD countries. But our costs per bed day are double the OECD 2008 average. So, the MS-DRG prospective payment system for hospitals has worked pretty well to contain admissions and lengths of stay – but our cost base is still a lot higher. It’s not like the hospitals are profiting off this, though.

      Compared to OECD countries, we have a higher volume of outpatient care, indicating that we’ve effectively shifted many inpatient stays to outpatient visits. This hasn’t helped us in overall costs, though. It appears that hospitals have shifted their outpatient service mix towards the more profitable categories, like imaging. So, higher outpatient volumes than the OECD countries times higher reimbursement per outpatient procedure.

      For physician office visits, we have about the same number of visits per capita, but higher costs per visit.

      On drugs, we use fewer but more expensive ones. We spend way more on administration, but that’s a small slice of the total pie. On long-term care and durable medical equipment, we – get this – spend LESS than expected given our higher per capita incomes.

      http://www.mckinsey.com/insights/mgi/research/americas/accounting_for_the_cost_of_us_health_care

      In other words, I think it’s more than that we’re delivering more services to seniors in total. It’s mostly that we deliver more costly care. In the places where our service mix may be less costly than OECD countries (e.g. our shifting some inpatient utilization to outpatient utilization), Medicare hasn’t been able to capture as much of the savings as perhaps it should.

      And I have to say, at first, the graph really made me think again. We know that Medicare’s per capita cost growth has been lower than private insurance’s. But we have a much bigger disparity between the young and the old in per capita costs than the other countries here. I wonder why, and how you square the two data points?

      I do know that politics can constrain Medicare’s ability to manage costs. The prospective payment system for hospitals has worked quite well in aggregate. But Medicare can’t clamp down on outpatient payment or utilization rates without provoking a political response. It can’t clamp down on advanced imaging as effectively for the same reason. And Medicare can’t really clamp down on drugs – we have single payer, central administration for most inpatient and outpatient services, but the drugs are in the hands of the Part D plans and pharmacy benefit managers.

    • Back of envelope calculation, using census data for cohorts and a fortiori per capita spending per the chart, comes to greater than $900 billion in expenditures. Medicare spending in 2005- $524 billion. Another $375 billion in out of pocket, medigap and Medicaid?

    • The US chart is on page 23 of the PDF. The US chart shows no jump in age 50-64 expenditures versus age 20-49 expenditures. The URL for the PDF is http://www.nber.org/papers/w11833.pdf?new_window=1.

      “What happens to Americans when they turn 55? Don’t blame Medicare. It doesn’t kick in until age 65.”

      If the Laurence J. Kotlikoff and Christian Hagist paper is correct, nothing happens when Americans turn 55. However, when they turn 65 expenditures soar.

      Blaming Medicare would appear justified.

    • Originally posted by “Plugh” at Ygelsias’s site:

      “This is a bogus chart. It takes a lot of searching to find the original data. The links reference Forbes, which references the Pittsburgh Post-Gazette, which references Prof. Paul Fischbeck. However, I could not find his publication(it may not have been published). The Post-Gazette also states that Prof. Fischbeck got his data from a 2005 study by Boston University economist Laurence Kotlikoff, though they misspelled his name.

      The original charts in the Kotlikoff study are for 10 countries, for different years from each country, and, most importantly, THE DATA IS ONLY FOR GOVERNMENT SPENDING. From the Kotlikoff study: “…for age groups under 65, the average values of government health
      expenditures used to form the U.S. profile are averages over the entire population at a
      particular age, including those not eligible for Medicaid and, therefore, receiving no benefits.”
      (see: http://www.nber.org/papers/w11833). Accordingly, it does not include the lion’s share of healthcare spending for those under 65, skewing the chart.

      The Kotlikoff paper did not chart them together because the data was from different years, currencies, etc. It also mentions that of the 10 countries studied, the US was at one extreme, Austria, Germany, Spain, and Sweden at the other, and Japan, Norway, the UK, Canada, and Australia are in between. This chart conveniently removes the in between countries to make the US appear more extreme. “

    • This is the problem with kludgeocracy. We scaffold solutions on top of a foundation that is fundamentally broken.

      The original sin of medicare was admitting the government should help Americans receive medical care, but only supporting one group in the population. It should be no surprise that Americans spend the most money and get the worst health outcomes: we guarantee spending on those who are closest to death and ignore the rest.

      The chart might be a fake, but the underlying idea is true: Supporting the healthcare for one class of people (in this case the elderly) but not others will distort the allocation of resources away from an ideal outcome.

    • All,

      Interesting to note that this one graph was selected for wider distribution – out of 8 total in the original article.

      Couple of points:

      1) This isn’t meant to be a peer-reviewed, scientific analysis. Simply a broad-brush of data that is arguably a little aged (2005).
      2) The larger point – emphasised by some of the other charts – is the prevelance here in the U.S. to use high-cost, specialty care (often billed piecemeal) over low-cost primary and preventative care.
      3) Older populations everywhere will consume more healthcare services.

      Beyond that – I’m not sure we can infer much useful information from this one chart. It would be great if wider exposure to this debate were to prompt an additional – more current study. My wish for the New Year!

    • Makes it look like a Government problem. It also reminds me of my friend who waited until he got on medicare to have to be looked at for a problem with his legs. He was doing find but they operated on him with very bad results.

    • I just got the link to this web site. I am happy to provide answers to any question that any readers may have.

      As one of my many academic pursuits, I take publically available data and build easy-to-use interfaces so that the general public can access the data. One such project resulted in DeathRiskRankings.com which allows direct comparisons for annual death rates for US states and 19 European countries by age, gender, race, and 66 causes of death. The web site when released went viral with millions of hits.

      In support of analyses run on the data extracted from web site, I pulled together data on obesity and smoking rates and health care costs. I only used countries for which my web site had death statistics, thus the selected countries. The healthcare costs graph was part of a large presentation on the web site and its use. About a year after that presentation, I re-did the graph using better data. The graph shown is not from analysis. It appears to be a redrawn sketch of the original. I will be happy to post the most up-to-date version.

      I used total medical expenditures for each country using OECD data.
      http://www.oecd.org/dataoecd/43/23/40905066.pdf

      I used the age expenditure distribution data from the following article:
      “Who’s going broke? Comparing growth in Public Healthcare Expenditure in ten OECD countries,” C. Hagist and L. J. Kotlikoff, Hacienda Pública Española / Revista de Economía Pública, 188-(1/2009): 55-72

      I distributed total costs across the age categories based on demographic profiles for each country.

      US data for pre-60-year-olds is from http://www.healthcostinstitute.org/files/HCCI_HCCUR2011.pdf

      • Paul – I would be happy to update the chart with more accurate data – if you care to provide it.

        I did “redraw” the graph – from it’s original incarnation – and the intent was never to be scientifically accurate.

        The fact remains – we *should* know how we compare to other countries as it relates to healthcare spending by age. This group was clearly quick to dismiss the chart – but the data does support the underlying premise.

        60 Minutes did a piece in 2010 that also alluded to this:

        “Last year, Medicare paid $50 billion just for doctor and hospital bills during the last two months of patients’ lives – that’s more than the budget of the Department of Homeland Security or the Department of Education.”

        Again – happy to update the chart base on newer – more accurate data. Just let me know.

        • FWIW, my heretofore unstated objection is the chart was not clearly labeled. Had it said “government spending” I’d be happier. I still don’t buy the magnitudes. They just aren’t right. Could it be they’re averages of non-zeros?

          • … and I would question the data now being submitted as “more accurate.”

            Simple math – take the 41M Age 65+ – and multiply it by the Mean annual spending ($10, 264) and you get about $422B. That’s off by a factor of $100B (at least) and doesn’t take into account other significant influencing factors:

            2) Dual Eligible
            3) Private Insurance (small % comparatively – but certainly influencing). It’s not remotely reasonable to assume that 100% of the population over 65 gets their healthcare exclusively through Medicare.

            It’s easy to be dismissive with simple math when the problem is actually far more complex.

    • Well I am certainly surprised with all the activity that my slide has generated.

      First, I am not a healthcare professional (I do system-level risk assessments across a wide variety of topics like space shuttles, oil platforms, and wind farms), so I don’t have a background on the strengths and weaknesses of different healthcare-cost data sources.

      From a quick perusal of the comments on this and other sites, I find that there is no agreement as to what is wrong with the graph, aside from the comment that it only captures government spending (which implies that it should be a lower bound). I have tracked several of the links that have been proposed to provide results and have found it difficult to extract a single common truth.

      The medical expenditure panel survey has 2010 total expenses as being $1.26 trillion.
      http://meps.ahrq.gov/data_stats/tables_compendia_hh_interactive.jsp?_SERVICE=MEPSSocket0&_PROGRAM=MEPSPGM.TC.SAS&File=HCFY2010&Table=HCFY2010_PLEXP_%40&VAR1=AGE&VAR2=SEX&VAR3=RACETH5C&VAR4=INSURCOV&VAR5=POVCAT10&VAR6=MSA&VAR7=REGION&VAR8=HEALTH&VARO1=4+17+44+64&VARO2=1&VARO3=1&VARO4=1&VARO5=1&VARO6=1&VARO7=1&VARO8=1&_Debug=

      The national health expenditures from the Centers for Medicare and Medicaid Services has 2010 expenditures as being $2.60 trillion.
      http://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/NationalHealthExpendData/Downloads/tables.pdf

      Per capita results will vary considerably based on what assumptions are made. Which is the better starting point? What is the MEPS data not including that the MCS does?

      As a quick check, I would like to collect your estimates of the per capita healthcare costs for the US by age categories. Having readers post estimates could help focus the discussion on what the disagreements are.

      What do you think the annual per capita expenditures are for:
      5-17 year olds ________
      18-44 year olds ________
      45-64 year olds ________
      65-74 year olds ________
      75-84 year olds ________
      85 and older ________

      Major assumptions behind the estimates?

    • Healthcare costs are *especially* squishy – which is exactly my point in defense of the original chart.

      For example – “hospice” care alone (toward end of life) – is over $5,000 per year – per person (U.S. only).

      When you include ALL healthcare costs (over 65+) – I think the $45k/yr figure could easily be accurate (if not scientifically precise).

      FYI – $2.6T / year is TOTAL NHE (National Healthcare Expenditure) – not just 65+ (for 2010).

      As I’ve argued before, however, that’s actually closer to $3.5T when you add “out-of-pocket” consumer healthcare costs – and accrual amounts (10yrs+) on categories like SGR (Sustainable Growth Rate = $300B on the books). Will we write that off? Maybe – but we need to account for all healthcare costs – not just those that support an agenda or policy POV!

      It’s provocative to “X” out the chart (with crayon red to boot) as a FAKE – but that’s equally wrong.

      • The chart is not labeled accurately. Data I’ve pointed to don’t agree with it. Many people have a lot of problems with that. I can’t defend it. I won’t. You may, of course.

        Meanwhile, how about some official data or citations to back up the figures, chart labeling, etc.? I’m just not seeing a coherent story.