America’s health-care system differs from its counterparts in other affluent nations in a number of ways: greater fragmentation among payers and price-setters, stronger incentives for overuse of advanced diagnostic and treatment technology, higher administrative costs, less access to care for some. We might therefore expect it to perform less efficiently — to achieve poorer health outcomes for a given amount of expenditure (see here, here, here).
The following chart is sometimes viewed as evidence in favor of this hypothesis. The chart plots life expectancy at birth by per capita health expenditures as of 2007. Twenty affluent nations are included. Among these countries the U.S. spends by far the most money on health care and yet has the lowest life expectancy.
The inference is problematic, however, because America differs from the other countries in a number of ways that may affect health outcomes. It has a higher murder rate. It has more obesity. The U.S. population is more spatially dispersed than those of most other countries, so rural residents may live farther away from medical providers. Given these and other differences, how confident can we be that health spending is less effective in the U.S. than elsewhere?
Here’s a better way to compare. This chart shows trends in life expectancy by trends in health spending from 1970 to 2008.
The United States still stands out, and in a big way. Our gain in life expectancy per additional health spending is much smaller than in other countries, particularly after the early 1980s when we reached expenditures of about $2,500 per person (in 2005 dollars) and life expectancy of around 74-75 years.
The advantage of analyzing country differences in change is that it takes constant nation-specific factors out of play. It’s not a foolproof analytical strategy, but it reduces the likelihood of mistakenly inferring causation from correlation.
What we need to be wary of is life expectancy depressors that may have increased more or decreased less in the U.S. than in the other countries. Are there any? Not smoking: our rate of decline is in the middle of the pack. Not homicide: it’s decreased more here than elsewhere. Probably not spatial dispersion: Americans began moving back into cities in recent decades. One possibility, though, is obesity. Not only is it more prevalent here; it’s also increased more.
This kind of analysis is by no means conclusive. Life expectancy and total spending are highly aggregated indicators; it’s important to also examine more fine-grained measures of health-care effort and outcomes (see here, here, here). But to the extent we treat the aggregate patterns as informative, a comparison of changes over time, rather than of levels, is likely to be our most valuable guide.
Update: Second chart now corrected, thanks to commenter Roger Chittum.
It appears that something changed dramatically in the mid to late 70s. Could that be the expansion of for-profit businesses into medicine taking advantage of government funding of Medicare, Medicaid, etc.? How would you measure that? For-profit hospitals, doctor groups, drug companies, medical equipment and supplies, nursing homes, etc.? Would this not have happened in countries with universal health care, at least not to the same extent?
Everything began changing in the mid-to-late 70s, these charts merely swell the parade of other graphs showing the same timing. Incomes stopped rising along with productivity in the late 70s, minimum wage buying power peaked in the late 70s and dropped within 8 years to about half of what it had been – and stayed there. Unions membership and effectiveness ditto.
What happened in the late 70s? Middle-class Americans were the richest people in the world, and had built a vast store of wealth in the form of property, pensions, education, savings, civil rights, general health, and lots more — and this wealth was a big, fat target. In the 70s, myriad effective strategies to tap this wealth came online and consolidated over the following decades. Like rats gnawing a way into the grain elevator, a host of predators carted off as much as they could carry, but also spoiled or wasted far more than they actually took.
If a foreign nation had somehow impoverished America to this extent it would be seen as an act of war. But instead, we fell prey to the hunger of home-grown rats, who have used part of their plunder to buy off the cats.
And it all started in the late 70s.
Noni
Excellent graph, it really makes it easy to rule out smoking and homicide as primary drivers of health care’s cost expansion.
Just stunning! But I can’t quite understand what is being plotted. Is it indexes of some kind? Taken at face value, the graph seems to be saying that the US and other nations were spending nearly nothing for health care in 1970. After printing it out and scaling it with a ruler, it looks like less than $200 for the US in 1970. It can’t possibly be right that we increased inflation-adjusted per-capita healthcare spending by a factor of 25-30, but that’s what the chart seems to say, doesn’t it?
BTW, using the same ruler technique, it appears the US curve began to roll over in 1972 and had flattened out completely by 1976. I can imagine reasons why costs might have begun to escalate in that time frame–inflation starting in 1971, oil shock in 1973, even worse inflation during Ford Admin, ubiquitous COLAs, etc.–but why would the gains in life expectancy level off then for Americans only?
Can we get a link to the original data?
@Doug You have the right idea, perhaps use a timeline of policy/law changes in concert with this graph.
Also, comparing this with a time plot of obesity over time may show if the trends match.
@Doug You have the right idea, perhaps use a timeline of policy/law changes in concert with this graph. See if when a major policy shift caused the change.
Also, comparing this with a time plot of obesity over time may show if those trends match.
Roger: As late as 1979, the U.S. was spending only $500 per capita on health care (in 2008 dollars).
Lane,
From this link at DHHS, https://www.cms.gov/nationalhealthexpenddata/02_nationalhealthaccountshistorical.asp, I get the following health care expenditures per capita per year:
1960: $147
1970: $356
1979: $972
1990: $2,853
2000: $4,878
2009: $8,086
Those are all very clearly in nominal dollars because the GDP numbers are clearly in nominal dollars and no inflation adjustment is stated. Per these data, the increase in health care expenditures as a percent of GDP is from 5.2% in 1960 to 17.6% in 2009, an increase of 3.4 times, not 25-30 times.
On these data, 2009 expenditures are 22.7 times 1970 expenditures, in nominal dollars, not so far from my rough estimates with the ruler. But these are not inflation adjusted dollars as claimed in the posted graphic. Does this have something to do with PPP adjustment? The plot is still stunning–especially the sharp knuckle in the rate of increase of life expectancy in 1972-76–but it doesn’t seem to portray the costs it says it does. Can we get to the data behind it? And where does the 500 2008 dollars in 1979 figure come from in your comment? Looks like there is a wild disagreement between our sources.
Roger: You’re right; I’m wrong. I’m doing some recalculating and will post the corrected version shortly.
Now corrected.
The combination of the expansion of for profit providers and fee for service incentives caused the increase in costs in this country. Other countries had universal health care which prevented both of those factors from being able to expand costs nearly as much. Most insurance companies in this country operated on a cost plus basis and did very little to control costs until very late in the game and the managed care and DRG systems were not very effective. Health care increasingly became a business as the nonprofits were pushed out or forced to operate similarly. There is a huge consultancy business showing providers how to beat the cost control systems.
“The United States still stands out, and in a big way. Our gain in life expectancy per additional health spending is much smaller than in other countries, particularly after the late 1980s when we reached expenditures of about $2,500 per person (in 2005 dollars) and life expectancy of around 75 years.” Does this require an edit with the change in the graph: Late 1970s instead of 1980s?
The corrected chart is in 2005 dollars, and life expectancies are life expectancies at birth, right?
Thanks for this. It tells the story very powerfully.
Actually early 1980s. Note that the health spending data are five years earlier (t-5), though the pattern looks similar with no lag.
Yes, per capita health expenditures are in 2005 dollars and it’s life expectancy at birth.
One thing that separtates the US from all the other rich countries is our diversity (the others are all far more homogenous). It’d be interesting to separate the US chart by race and see if it changes.
My understanding was that it is rather agreed upon that the lower US longevity rate is due to specific social factors, not the healthcare system. Here are some links…
http://www.nber.org/papers/w15213.pdf
http://repository.upenn.edu/cgi/viewcontent.cgi?article=1012&context=psc_working_papers
http://www.bbc.co.uk/news/world-us-canada-14070090
However, healthcare costs are way out of control.
Fantastic blog!
@bluto: while the US is more diverse than other rich countries that hasn’t really changed fundamentally in the last 40 years (yes, we have seen some change but nothing as dramatic as that sudden tick over in the early 80s as shown in the graph).
Also have to wonder about obesity. No denying we have a negative trend there but it seems that might be connected to a larger problem in US regulation: it seems the market in the US is skewed to produce cheap food that’s bad for you and to produce expensive health care that’s not particularly good for you.
Excellent, provocative post.
Inflection point in the percentage of those in the US with health coverage was 1980, give or take a few years. It had been growing, but began to decline as more people fell out of the system than entered it. Not to exclude other factors, but I strongly suspect that may be a significant contributor to what you’ve identified.
The coverage inflection was driven by economics (both micro and macro). I did a quick review here: http://bit.ly/n1eeqZ
Like my initial reaction, JWR focuses on why life expectancy did not increase more. But as I looked at it more carefully, and as I think Lane says, US life expectancy did improve about as much as the worst of the other 19 nations. We just did it at enormously greater cost. In an email, Lane says the kink in the US curve occurred 1981-83. What permanent change occurred then?
My best hypothesis is that the kink occurred at the end of the hyperinflation beaten by Volcker’s Fed and the beginning of the Great Moderation but that the healthcare industry was somehow able to keep inflating its incomes per unit of output faster than CPI. How could that happen? Well, who or what would stop these politically powerful actors from using their market power? Insurers are happy to handle more cash. Employers can’t seem to get any advantage of competition between insurers. Patients are terrible negotiators. Every year Congress enacts the “doc fix” to waive the law that limits Medicare reimbursements to the baseline plus inflation.
In the Great Moderation, those of us who had been accustomed to wage and revenue increases that exceeded the inflation rate in most industries realized that would no longer be happening, and it didn’t. But, I speculate, in healthcare the expectation of rising real incomes for producing the same outcomes just never got stamped out. Anybody have data on this?
I would question the correlation between life expectancy and healthcare (regardless of cost).
Based on my experience Diagnostic and Preventative Healthcare in the US has not advanced at all – very primitive. Our god complex doctors are focused only on procedures (i.e. broken bone / cast) and prescribing useless tests and drugs. Summary, doctors and resulting healthcare in the US sucks.
What about the fact that the US has a lot more immigration, especially from low income countries?
Ideally, we would like to use a dataset with only people who have been in the US for their entire life.
@bluto: You can’t find a more diverse country than Canada right now. In major urban areas, nearly 50% of the population consists of immigrants who have settled in Canada. These diverse communities are well served by the health care system (fully publicly funded, I might add), at significantly lower per capita cost. Perhaps broad socio-economic disparity may be a better indicator? Perhaps a better understanding of how the political landscape and powerful interests groups affect health care policy (both difficult to measure) provides a clearer explanation?
I’ve read elsewhere that when homicide and traffic deaths are subtracted from ALL countries, U.S. life expectancy is the highest. Can anyone verify that? If it’s true, the U.S. lag in life-expectancy trend shown in the second graph could simply be the law of diminishing returns; each additional dollar buys a smaller increase in life expectancy?
Another possible explanation for the graph is that the U.S. spends a lot more on R&D than other countries. Foreign countries may have longer life expectancies by free riding off of U.S. medical innovation without having to pay the costs.
Did you try to include immigration in the calculation? The explosion of immigrants since the 70s has been well documented. Many of the huddled masses come in at the bottom of the income distribution. Poor folk don’t live as long. Healthcare costs are, of course, too damn high, but I think obesity + immigration explains a lot more of the difference than some sort of war on the middle class.
So if we capped health care costs for medicaid and medicare then we could bring our spending in line with other countries and we might improve outcomes because doctors wouldn’t be spending millions on worthless tests to avoid lawsuits.
It is not obesity or diversity. Other countries in this list have nearly equally high levels of obesity (e.g. Australia) and equal or greater levels of diversity. Neither of those characteristics correspond to the stark difference highlighted here.
It is also certainly not the first time we’ve seen this. It has been reported ad nauseum in most of the high-impact medical journals for quite a while.
No need for this much talking.
Everybody who lived for some time in Europe perfectly knows how the US health system is doomed.
One of the ways the UK spends healthcare money is on obesity prevention – obesity is not an external factor but part of the equation.
Fascinating analysis.
Is the 2nd graph available showing the evolution between the US and BE, which is one of the 20 ‘affluent’ nations?
Thanks
It illustrates US spending is out of control. But, the chart does not together lower US life expectancy and higher rates of obesity. I saw a study last year of Hispanic immigrants. The second generation is the heaviest population in the country and has all the related health consequences. But, there parents who came to the US and did physical labor and tended to eat a more vegetarian diet (beans) turned out to be quite healthy in their old age.
The effects of obesity are great and the way we grow and produce food in the US is not good. If their were better statistics on weight and the effect on health (i.e. length of life), then it might explain statistics better. Nonetheless, it wouldn’t take a genius to conclude that spending on medical care cannot help anyone that doesn’t help themselves by shedding the weight.
The original discussion in the post, as well comments left by readers, all tend to treat obesity as some exogenous factor, presumably categorized under “lifestyle choices that can be controlled” or some such grouping. But an effective healthcare system deals with obesity, too. In other words, even if 100% of the deviation shown by the US came from obesity (and it certainly does not), we could still say that healthcare spending is considerably less effective than other countries’ spending. The same goes for a health system that fails to provide treatment to rural areas. These should be considered endogenous to the healthcare system and treated as such in our analyses.
I think you could clarify the graph a bit, and remove one confounding factor, by using as the x-axis “Total expenditure, % GDP” — which is conveniently available in the OECD dataset.
If it’s not obvious, the reason this is better is that a richer country is obviously going to spend more in $ terms than a poorer country, but this points to higher overall wage levels rather than inefficiency.
The answer is absolutely self-evident:
In US healthcare is provided in relation to the ability to pay for care/insurance (or the ability to maintain a job that often requires an education). Poor health is much more prevalent amongst those with little resources and poor education. While the rest of the world provide healthcare to the people that need it most and benefit most from it, the US health-system instead provides the best and most expensive care to the richest and healthiest.
Also, given the same resources, a good insurance is more affordable for the healthy. In short, US healthcare system allocates the most resources where they are needed the least. The essence of capitalism, ladies and gentlemen.
Two factors best explain root causes: Our lifestyle and healthcare spending in the last year of life.
75% of all healthcare costs are spent on preventable chronic disease (CDC). We eat 500 + more calories a day than in the early 80’s. Only 3 out 10 of us get adequate physical activity. 68% are overweight. In 2010, all 50 states have an obesity rate above 20%. In 2000, zero states were above 25%. I could go on. To contrast, check out the Amish with a 4% obesity rate despite a high calorie, high fat diet. I don’t want to live like the Amish, but we can do better than we are doing! Time to step up to the plate (or away from it) America. We are also doing a great job exporting our lifestyle to other nations.
Spending in the last year of life constitutes 25% of all Medicare spending. As a physician, most of this seems to be on fruitless procedures in the ICU because either the doctor or patient or family doesn’t know what a good death looks like. I’m not an advocate by any means of euthanasia or death panels. I’m a huge fan of patients, doctors, and families sitting down and making informed decisions about the last year of life and how it should be spent. This takes time, trust and relationships – not policy, ulitization review, laws, etc. I’m also a big fan of hospice. I think we are all abit too squeemish about talking about the dying process and how to do it well. The economic cost of this squeemishness is also catching up to us!
Earlier this year the CIA issued a study saying that the biggest road block in the U.S. to better health outcomes was access. Not smoking. Not obesity. Access. When you have 50 million citizens without health insurance and another 50 million with basically worthless insurance you end up with an access problem and therefore, near third world outcome numbers. Even families with so-called “good” insurance can face $8000 a year deductibles which research shows is a major impediment to seeking health care.
Face it. Our non-health care, non-system is expensive, ineffective, inefficient and immoral. Other countries spend half as much and cover EVERYONE. But then, they aren’t owned by insurance companies, Big Pharma and the hospitals.
If we are going to use life expectancy and/or per-capita spending as the primary measures of healthcare system desirability, we need to think things through.
With life expectancy, to honestly call it a measure of *healthcare* effectiveness, we need to exclude deaths that do not touch the healthcare system (e.g. homicides and accidental deaths) from the life expectancy statistics. Is it possible to redraw the graph with that adjustment?
Regarding per-capita costs, if my Aunt Edna can get a hip replacement and cataract surgery here (at some cost), but cannot somewhere else (so $0), she would affect a higher per-captia expenditure here. But are lower per-capita expenditures *really* what we want in such a case? If not, how do we make the stats reflect only the *undesired* spending increases (e.g. higher costs for an identical product or service here versus the average under some other type of system)?
Analyses that disregard these factors are irrelevant at best and deceptive at worst, IMO.
“Twenty affluent nations are included”
How exactly were they chosen? I see Asl Aus Bel Can Den Fin Fr Ger Ire It Ja NZ Nor Por Sp Swe Swi UK US and one other I can’t read.
This doesn’t seem to be a list of the *most* affluent countries, like Qatar, Singapore, Kuwait, UAE,etc.
Were they cherry-picked to make the US look bad? Do the graphs of poor countries look the same?
December 2014 update – The disparity between health care costs and life expectancy has only worsened since the date of this article and its tables. See the later comparison of OECD member countries (developed countries) and the bottom line is that the $3 trillion we spend only seems to yields $2 trillion in relative healthcare outcomes. It’s an unfunny joke.