We often hear that we, in the United States, spend more on healthcare than other high-income countries, but get less for our money. A report from the Commonweath Fund, “Mirror, Mirror 2017”, is among many pieces of research to reach that conclusion. Just why is U.S. healthcare spending so high and performance so low? What are the realistic options for reformers? One of the report’s key charts provides an excellent framework for discussing what it calls “flaws and opportunities for better U.S. healthcare.”
Why is U.S. healthcare spending so high?
To understand why U.S. healthcare spending is so high, we can begin by asking, “High relative to what?” After all, as one of the wealthiest countries in the world, we spend more than others on a lot of things. The question is whether U.S. healthcare spending is higher than we would expect it to be even when we take our generally high standard of living into account.
Consider, for example, the following scatterplot of healthcare expenditures (HEX) per capita against GDP per capita. The eleven countries covered in the Commonwealth report are shown in red and other OECD countries in blue.* The United States is the furthest outlier in the chart, falling more than four standard errors above the trend line. That puts it well outside an interval of plus or minus two standard errors, which would be expected to contain 95 percent of countries.
However, GDP may not be the appropriate basis of comparison. According to the anonymous author of the blog Random Critical Analysis (whom I identify hereafter as RCA), it makes more sense to compare healthcare spending, not to GDP, but to a less familiar measure of living standards known as actual individual consumption (AIC).
AIC is the sum of household final consumption expenditure, final consumption expenditure of non-profit institutions serving households, and government expenditure on individually consumed goods and services. Government spending on healthcare, education, housing, and other goods and services is included in AIC because those items are consumed individually by consumers. However, spending that serves the public in general, such as national defense and police protection, is not included. As a measure of a country’s average standard of living, AIC is thus not sensitive to differences among countries in the way healthcare funding is divided between public and private sources.
The share of AIC in GDP varies widely within the OECD, ranging from 59 percent in Norway to 78 percent in Greece. At 73 percent, the U.S. ratio of AIC to GDP is near the top of that range. One reason that AIC varies is that households in some countries consume more of their personal income and save less. Trade deficits and surpluses are another reason. Since countries with deficits import more than they export, they can use the extra imports to support higher levels of both private and government consumption expenditures per dollar of GDP. Both low saving and persistent trade deficits tend to raise the share of AIC in GDP for the United States. It has the highest standard of living in the OECD as measured by AIC, even though it does not have the highest GDP per capita.
Not surprisingly, healthcare spending tends to increase as AIC increases. Not only that, but healthcare is a superior good, meaning that its share of total spending increases as AIC grows. My estimate for OECD countries is that each 1 percent increase in AIC is associated with approximately a 1.6 percent increase in HEX. RCA gets similar results both across countries and over time within individual countries. In the following chart, which uses log scales on both axes, about 80 percent of the variability in health expenditure is statistically explained by variations in AIC.**
Comparing this chart with the previous one shows that the United States, although still an outlier in healthcare spending relative to its OECD peers, is much less of one. When compared to AIC rather than GDP, U.S. healthcare spending lies just 1.5 standard errors above its predicted level, instead of 4 standard errors. That puts it well inside a 95 percent prediction interval.
For completeness, it is worth noting that the preceding chart says nothing about whether prices or quantities of healthcare are more important in producing higher-than-expected U.S. healthcare spending. That remains a matter of ongoing debate among healthcare economists.
One line of thought is reflected in a widely-cited paper with the provocative title, “It’s the Prices, Stupid.” Anecdotal evidence, such as $100 saline bags and million-dollar cancer pills, reinforces the idea that excess healthcare spending stems from unreasonably high healthcare prices.
However, there is also evidence pointing the other way. For example, in another post, RCA argues that high U.S. healthcare prices are partly explained by the so-called Baumol effect, according to which productivity increases faster in goods-producing sectors than in services as GDP rises. That causes wages and prices in service sectors, including healthcare, to rise relative to the average price level as an economy’s general standard of living increases. Corrected for the Baumol effect, U.S. healthcare prices are not as far out of line with average OECD prices as they would otherwise appear. RCA also cites direct evidence that quantities have increased, along with total spending. (For anecdotal evidence to that effect, see my own recent notes on dermatology clinics and urine drug testing.)
Whichever point of view one takes on the relative role of prices vs. quantities, however, the point stands that high spending can only in part be viewed as a “flaw” in our healthcare system. Yes, lack of competition, perverse incentives, and misguided or ineffective regulations undoubtedly are sources of inefficiency, but there are also more fundamental forces at work. On the demand side, there is strong evidence, over time and across a wide variety of national systems, that people are willing to devote an increasing share of their total consumption spending to healthcare as their standard of living rises. At the same time, the Baumol effect makes it inevitable that the price of personal services, including healthcare services, will rise as standards of living rise. Perhaps the real question, then, is not why we spend so much, but why we get such poor performance in return.
Why is U.S. healthcare performance so poor?
Health system performance, in the Commonwealth report, represents a weighted average of scores on many individual variables. These are grouped in into five categories: The care process, access, equity, healthcare outcomes, and administrative efficiency. The following is a brief summary of the findings in each category. A set of appendixes to the report provide full data on each of the underlying variables for each of the eleven countries.
It turns out that the United States does rather well, ranking fifth overall, in terms of the first category, “care process.” The rankings are based on data covering preventative care, safety, coordination, and engagement with patient preferences. Certainly, there are places where there is room for improvement, but this is not the area of the rankings where U.S. reformers should look for dramatic gains.
Instead, low U.S. performance scores can, to a much larger degree, be traced to last-place rankings for healthcare access and equity. Americans who lack insurance or face high out-of-pocket costs are far more likely than citizens of the other countries to report skipped checkups and serious problems paying medical bills, and are less likely to have a regular doctor or clinic. What is more, the gap between access to healthcare by those with below-average and above-average incomes is wider for the United States than for any other country in the Commonwealth study.
The United States also ranks last among the eleven countries in terms of healthcare outcomes. Broad measures of population health, including infant mortality, life expectancy, and the prevalence of chronic diseases in working-age adults are all worse in the United States than in any of the ten other countries in the Commonwealth study. U.S. amenable mortality rates are also much higher, and have been declining more slowly than elsewhere. (Amenable mortality is defined as premature deaths that could potentially be avoided given timely treatment.)
The difference in healthcare outcomes appears to be linked, at least in part, to problems in access and equity. That is suggested by the fact that on variables that measure outcomes for people who actually receive treatment for conditions like stroke, heart attacks, and cancer, U.S. performance is well above average. Not everyone has access to the best treatment, however, nor do people without insurance or with high out-of-pocket costs always follow through on treatment that is offered.
Still, it would be wrong to attribute poor U.S. health outcomes entirely to flaws in policy. Socioeconomic factors, such as obesity and lack of exercise, also contribute. One study found that traffic accidents, gun violence, and drug use account for as much as two-thirds of the mortality gap between the United States and countries such as Austria, Denmark, Finland, and Germany.
Administrative costs are the final performance category in the Commonwealth report. This is an area in which many American advocates of a single-payer system have high hopes for improving performance. They are right to point out that the U.S. ranks poorly in this area compared to the median for the Commonwealth study. However, even some European healthcare systems that have higher overall performance scores struggle with some aspects of administrative costs. In France (which is the worst-performing of the eleven in terms of administrative costs), and also in Switzerland, Germany, and the Netherlands (all with below-average scores), complaints about excessive time spent on insurance, paperwork, and reporting are common. The United States is perhaps less of an outlier in this area than one might think.
Implications for reformers
Our review of the reasons for high U.S. healthcare spending and low health system performance has clear implications for reformers, as suggested by the following modification of the Commonwealth chart:
- There are limits on what U.S. healthcare reformers can reasonably expect to achieve in terms of spending and performance. Those limits are represented here as a “policy option envelope,” which is pushed to the right, relative to other countries, by high U.S. levels of actual individual consumption, and downward by socioeconomic factors. As a result, it cuts to the right and below the positions occupied by other countries in the sample.
- It is wishful thinking to hope that simply copying one of the various systems used in Australia, the UK, the Netherlands, or elsewhere would automatically produce similar results in terms of cost and performance. There is no easy path—probably no path at all—by which the U.S. could leap into the central cluster of low-cost, high-performing countries in the chart. That does not mean we have nothing to learn from others. It simply means that what works well elsewhere is likely to cost more here, or produce less favorable results, or both.
- Although we could certainly improve overall healthcare performance by prioritizing greater access to the system we now have, doing so would, other things being equal, mean an increase in spending.
- We could, instead, prioritize spending cuts to Medicaid, Medicare, ACA subsides, and other programs, but doing so would worsen performance. That is especially true when performance is measured as it is by the Commonwealth study, where large weights are assigned to access and equity.
In listing these implications, I do not mean in any way to pour cold water on reform efforts. Rather, I think that greater realism makes it more important, and perhaps also easier, for reformers of all political persuasions to work together.
In particular, I would emphasize that we are far from being tight up against the performance-vs.-spending tradeoff represented by the policy option envelope in the chart. There are many potential efficiency-enhancing initiatives, for which it ought to be possible to get broad consensus, that would move us closer to the curve. Those include measures to mitigate perverse incentives that encourage patients to seek and providers to offer care that is neither clinically nor economically effective. They would also include reforms designed to increase competition, improve price transparency, better educate consumers, and push back against socioeconomic factors that undermine public health.
None of these should be controversial in principle, although many of them will meet resistance from special interests who profit from existing inefficiencies. As Uwe Reinhardt once put it, “Every dollar of health spending is someone else’s healthcare income, including waste, fraud, and abuse.”
Still, even with maximum efforts to enhance efficiency, we need to accept that healthcare will continue to consume a large share of both personal and government spending. In the end, the prospect of continued high healthcare spending is not a reason to resist reform, but rather, to do our best to make sure that we get a good return on what we do spend.
*The chart covers 34 OECD countries, excluding Luxembourg. Luxembourg is unusual in that nearly half of its labor force commutes daily from neighboring EU countries. The commuters contribute to Luxembourg’s measured GDP, but are not counted in its population. Moreover, most of them get their healthcare where they live, not where they work. For that reason, data on GDP per capita and health expenditures per capita for Luxembourg are not comparable to those of other OECD countries. In our chart, Luxembourg would be an extreme outlier in the lower-right hand corner. Including it would pull the trend line down and exaggerate the degree to which the United States is an outlier in the upward direction.
**One technical detail: Healthcare itself accounts for a substantial share of all consumption spending, so as a matter of simple arithmetic, HEX would increase as AIC increased even if healthcare spending were completely uncorrelated with consumption spending on other goods and services. To avoid bias, then, the scatterplot puts non-healthcare consumption spending (AIC minus HEX) on the horizontal axis, rather than total AIC. For a regression of HEX on AIC-HEX, the coefficient of determination (R2) is 0.80. As expected, that is a little less than the R2 of 0.87 obtained when total AIC is used as the independent variable.