It is commonly claimed that the American health care system is inefficient compared to other countries’ systems, and therefore major changes for the U.S. system are in order. For example, Paul Krugman, in a March 28, 2008 post on his New York Times “Conscience of a Liberal” blog, wrote, “Everyone knows that the US spends much more on health care than anyone else, without getting better results.” In a speech before the American Medical Association in 2009, President Obama said that “we are spending … almost 50 percent more per person than the next most costly nation. And yet … we aren’t any healthier. In fact, citizens in some countries that spend substantially less than we do are actually living longer than we do.” Donald Berwick, a health care expert who was the administrator of the Centers for Medicare and Medicaid, wrote in 2008 that “[d]espite spending on health care being nearly double that of the next most costly nation, the United States ranks thirty-first among nations on life expectancy (and) thirty-sixth on infant mortality.”

These arguments about efficiency of health care systems are based on the economic concept of a production function or relation. This is called “household production” because households use health care and other inputs to produce health. Personal characteristics such as education, income, pollution, lifestyle, culture, and possibly genetic differences are important inputs into the production of health. Health is unobservable, so indicators, most commonly life expectancy and infant mortality, are used as proxies for health. These indicators are used because of wide availability and a belief that they are reasonably well measured. But they ignore the quality of life. This is a problem because much health care is intended to improve quality of life rather than to reduce mortality.

The claim of U.S. inefficiency flows primarily from an overly simple view of the production of health: the idea that health is produced only by health care or that the other inputs do not differ much by country. This view ignores other inputs that affect health and that vary from country to country. Economists, including me, have pointed out that there are many problems with this and with the data definitions and measurements that are used in international comparisons. My colleagues Stephen Parente and John Hoff and I are continuing work in this area. All the health care systems have inefficiencies and distortions, but the bottom line remains: the U.S. health care system is probably no less efficient than the systems of other developed countries.

Scott Atlas’s informative new book In Excellent Health takes up much of this argument in a nontechnical way, from the viewpoint of a physician. He illustrates his arguments at a thought-provoking level of clinical detail. Many of the analyses are known to some researchers in the relevant areas of economics, demography, or medicine. This book provides an excellent, accessible summary for specialists and nonspecialists alike.

WHO index and rankings | In this book, Atlas first tackles the 2000 World Health Organization index and rankings of health care systems, where the U.S. system was ranked 37th in “overall performance” and 15th in “overall attainment.” The index underlying the ranking was based on subjective factors that are not directly related to the comparative efficiency of health care systems. For example, financial fairness and health distribution constitute 50 percent of the rankings, while actual health only counts for 25 percent. Further, missing data for many measures of many countries were filled in by judgments of what the WHO calls “informants.” This analysis is useful and nicely explained.

Life expectancy and confounding external factors | Atlas then critiques the use of life expectancy as a proxy for health output. First, it is heavily dependent on infant mortality, which I discuss below. Beyond that, Atlas notes the importance of other factors that affect life expectancy that are not related to the productivity of health care.

Atlas lists 25 factors that are external to the health care system but that strongly influence life expectancy. As he points out, it is a serious mistake to assume that these factors are somehow controlled by, or even much influenced by, the health care systems. Atlas nicely illuminates this point by noting that the life expectancy at birth of Americans of Asian and Pacific Island background is almost as high as the Japanese (the highest country level), 81.5 versus 81.8 years, while the U.S. average over all ethnic groups is 77.2 years. This potent image calls to mind Victor Fuchs’s earlier comparison of age-specific mortality of residents of Nevada versus Utah. The differences were spectacular, even though the health care systems were virtually identical. The excess mortality for Nevada was about 60 percent for both males and females aged 40–49 and about 38 percent for infants. Fuchs’s and Atlas’s messages are the same: lifestyle and other external inputs into the production of health are vitally important.

Unfortunately, the influence of these other inputs on health confounds an attempt to compare the efficiency of the U.S. health care system to other countries’ systems. Lifestyles are generally much less healthy in the United States than other developed countries. Consider Atlas’s discussion of three categories of these external factors: accidents, suicides, and murder; obesity; and smoking.

One might think that health care matters for deaths from accidents, but I believe that this is a minor issue. The main issue, after an initially serious but not-yet-fatal accident, is the speed with which the individual reaches a hospital. Indeed, it is called the “golden hour.” The time to treatment is largely explained by population density. Studies of traffic fatalities, such as those by Michael Morrisey and David Grabowski, achieve very good explanatory power across U.S. states without using any health care variables at all. Atlas reports on an adjustment that standardizes all countries to the Organization for Economic Cooperation and Development average death rates from these external causes. In other words, it answers the question: What would be the life expectancy of these countries if they all had the average OECD death rates from these external causes? The result shows that the United States was ranked low in raw life expectancy, but much higher in standardized life expectancy.

Obesity, conventionally defined as having a body mass index above 30 (e.g., 209 lbs. for a 5 ft., 10 in. person) is both harmful to life expectancy and also raises health care costs. Obesity itself is an intermediate product, being produced by underlying cultural attitudes and lifestyle choices such as exercise, diet, and even urban design. None of these causes are importantly influenced by the health care system. The United States is by far the most obese country in the developed world, with 34 percent of the population considered obese. (In comparison, the United Kingdom is second, with 24 percent, and Canada is third, with 16 percent. Is speaking English a risk factor?) Atlas cites a finding that obesity affects life expectancy with a substantial lag of about 25 years for the full effect. William Comanor, Richard Miller, and I found that controlling for obesity (with a time lag of about 10 years) accounts for a bit more than half of the difference between U.S. life expectancy and what the life expectancy would be if the United States had average OECD countries’ apparent productivity. In other words, if one doesn’t control for obesity, the United States looks relatively inefficient, but simply adding even an imperfect control variable for obesity eliminates a bit over half of the apparent difference.

Smoking is obviously detrimental to life expectancy. One might think that this is not an issue for comparing the United States to other developed countries because U.S. smoking rates in the past 10 or 20 years are not high by international standards. Indeed, David Squires, in a recent Commonwealth Fund report, makes that argument. But analyzing this relationship requires a longer horizon. Atlas reports a surprising fact: For 50 years, ending in the 1980s, Americans smoked more than consumers in any other developed country. Indeed, there were long periods when 70 percent of adult Americans smoked. Obviously, the change in smoking rates since the 1980s has been dramatic. This somewhat distant history of heavy American smoking is relevant for current life expectancy. The ill effects of smoking operate with a very long lag. The relationship between smoking and lung cancer is strongest at a lag of 21 years and is still nontrivial at a lag of 35 years! Indeed, demographer Samuel Preston and his coauthors have shown that prior smoking still has a strong influence on U.S. life expectancy, even in recent data. By removing smoking-related deaths from 2003 data, female life expectancy at age 50 for the United States moves from near the lowest in the developed world to the middle of the pack.

Measuring infant mortality | Next, Atlas considers the weakness of the other commonly mentioned health outcome measure: infant mortality. There are two major problems with using this measure. First, perhaps surprisingly, measures of infant mortality are not comparable across countries. Second, infant mortality is highly sensitive to external factors, especially to the lifestyle of the mother.

Taking the measurement issues first, Atlas notes that different countries have different practices and standards for whether a fragile, very high risk birth is recorded as a live birth versus a still birth. Recording a fragile birth as a live birth raises measured infant mortality. Atlas shows that in the United States, these births are more likely to be recorded as live births and that the variation among the developed countries is large and quantitatively important. Apparently, there has been only limited progress in standardizing how births are recorded, even in the developed countries. Deviating from the WHO’s definition, it is still the case that many other countries define a live birth by birth weight, length, gestational age, and even actual survival time. In a recent article in the BMJ, K. S. Joseph and coauthors find that the number of reported births at less than 500 grams (1.1 pounds) in the United States is 16.9 per 10,000, while in Ireland and Luxemburg it is 0.0, in Belgium it is 0.4, and in Norway it is 1.9. The highest European country, England and Wales, reports only 6.2. Joseph and coauthors attribute most of the variation to differences in birth registration, which they say “compromises the validity of international rankings based on perinatal, infant, or child mortality.” As a result of these registration differences, the WHO recommends that international comparisons be limited to babies who weigh 1,000 grams or more. Joseph and coauthors note that both the United States and Canada are ranked higher in such a comparison than in raw infant or neonatal mortality. Recording differences have a large effect on infant mortality because very high risk babies account for a large proportion of infant mortality. The definitional and recording differences artificially inflate U.S. infant mortality.

Lifestyle and infant mortality | Turning to lifestyle, Atlas argues convincingly that it is even more powerful for infant mortality than for life expectancy, an argument that my coauthors and I have also made. Much of the effect of lifestyle on infant mortality is summarized by birth weight or gestational age. (Low birth weight is closely correlated to low gestational age.) Low birth weight babies are much less likely to survive. For example, babies weighing less than 2,500 grams (5.51 pounds) are 20 times more likely to die than the average-sized baby, with the odds getting dramatically worse for smaller babies. The United States has a higher percentage of preterm or low birth weight babies than any other developed country. Certain lifestyle choices are especially likely to lead to low birth weights and high infant mortality, particularly teenage motherhood, smoking, and obesity. The United States leads the developed world in teenage motherhood, over 40 per 1,000 girls, which is almost double the UK’s rate, four times France’s, and almost 10 times Switzerland’s.

Another factor that has gotten less attention is differences in treating infertility. Fertility treatment leads to more multiple births, which are far riskier than single births. The mortality rate for twins is about five times that of singletons, while for triplets it is about 12 times. Because of aggressive infertility treatment, the United States leads the world in births of three or more.

There is a natural way to adjust for some of the lifestyle effects on infant mortality and arrive at a superior measure to compare health care system productivity. That is, one could examine infant mortality for specific birth weights or gestational ages. One of the simplest ways to do this is to calculate what the U.S. infant mortality would have been if the United States had the same distribution of gestational age or birth weight as some other countries. Atlas reports several calculations of this type, showing that the U.S. infant mortality, adjusted in this manner, is quite low, comparable to Canada’s, Sweden’s, and Norway’s. In sum, because of definitional and measurement differences and the powerful confounding influence of external factors, especially lifestyle, infant mortality is a poor measure of health system output.

In passing, Atlas notes that life expectancy is strongly influenced by infant mortality. It is also influenced by mortality at young ages, which is dominated by accident, suicide, and violence. This is the reason that life expectancy at birth is particularly inappropriate for comparisons of the efficiency of health care systems. Life expectancy at later ages, such as 40 or 60, is somewhat less contaminated by external factors and has been studied to some extent. But it is not the main emphasis one sees in broad policy discussions.

Measuring health care spending | The main goal of the international comparisons is to compare costs to benefits. Atlas does not discuss the health care spending side in a comparative context. But there are more problems there and they also tend to make the U.S. system look less efficient than it really is. To measure health care system efficiency, one needs a measure of health care resources used. Then one can compare the productivity of the health care systems. The most common and comprehensive approach is based on health care spending. Spending in the domestic currencies of the various countries is translated into a common currency (usually U.S. dollars) using some exchange rate. This is normally done using the overall, economy-wide purchasing power parity (PPP) exchange rate. The PPP exchange rate adjusts for differences in average prices across countries so that purchasing power is identical across countries. In principle, $1,000 exchanged at the PPP rate would enable purchasing the same bundle of goods in all countries. Using this economy-wide exchange rate makes sense only if the relative price of health care is approximately the same everywhere. But this is not the case. In particular, American health care is relatively more expensive, so the overall PPP exchange rate gives the incorrect impression that the United States uses more health care resources than other developed countries.

The mismeasurement caused by using the economy-wide PPP exchange rate is large and quantitatively important. This is discussed in several places, including by my coauthors and me, by Mark Pauly, and by David Squires. There are two ways to deal with this problem. First, one can use the PPP exchange rate that is specific to medical care to measure real resources used in health care. A few comparisons illustrate the magnitude of the differences. Using the health PPP exchange rate instead of the economy-wide one moves Denmark from 57 to 73 percent of U.S. spending. The biggest mover is France, which moves from 61 percent to 113 percent, higher than the United States. Alternatively, one can look at physical measures, such as the number of physicians and other health workers per capita or the number of visits or hospital stays. Doing either analysis shows the United States is not an especially high user of real resources in health care. Higher U.S. prices appear to be caused primarily by higher salaries and incomes for American physicians and, probably more importantly, for nurses and technicians.

The result of using the economy-wide purchasing power parity exchange rate is to overstate the resources going into U.S. health care, making it appear on the surface to be less efficient. But there is also another, more subtle issue in the mismeasurement of the costs of care: the hidden costs of health care in non‑U.S. countries.

Hidden costs | In most other developed countries, health care prices are controlled below the level necessary to clear the markets. This is especially common in single-payer systems like those of Canada and Japan. The result is a great deal of nonprice rationing. Some of the nonprice rationing is based on professional judgment, roughly similar to that occurring in competing managed care plans in the United States. It is probably reasonably efficient. But much of the rationing is accomplished by consumers waiting for services, which leads to large hidden costs of health care. This general point has been made before and has even become a political and legal issue in some countries. Atlas documents this in valuable micro detail. For example, the wait time for cataract surgery in the United States is essentially zero, but the mean wait time in Europe is 3.5 months. Waiting causes direct harm to consumers’ well-being and raises medical risks, including the risk of permanent vision loss. These waiting time costs of health care systems are not on any budget. They are difficult to track accurately because some patients never go on formal waiting lists, either because the waiting is not formalized or because they are discouraged from obtaining the care at all.

Thus, access in some systems is not as good as it appears. Having a service covered formally by a system is no guarantee of access to care. This problem also occurs in the American Medicaid program, where prices paid are set so low that the majority of physicians will not treat Medicaid patients. On average, Medicaid pays only 72 percent of what Medicare pays for the same service. In California it pays only 56 percent; in New York only 47 percent. Research by Chapin White shows that expanding the SCHIP program for children has increased insurance coverage, but has not increased utilization and has reduced access by some measures. Unlike Medicaid, the nonprice rationing problem is system-wide in some other counties. Atlas shows that for many different diagnoses, Americans obtain appropriate care more often than those in many other countries. The delay and poor access to care resulting from rationing by waiting harms health outcomes, but delay and poor access tend to be concentrated on issues that are not life threatening; therefore, they do not seem to have large effects on mortality.

Conclusion | Atlas’s book is an excellent contribution to the study of international differences in health care productivity. Written in a readable, clear style, it covers many of the problems of measurement and external causes of health, often at the micro level of the individual diagnosis or service.

Atlas’s findings raise an important question: Why are U.S. lifestyles so unhealthy? I suspect that the answers to that question are bound up in complex issues of culture and history, and we may never have a satisfying answer. But whether or not we make much progress on that question, it is important that we don’t make policy based on misunderstanding.

Readings

  • “A Comparison of Two Approaches to Increasing Access to Care: Expanding Coverage versus Increasing Physician Fees,” by Chapin White. Health Services Research, February 2, 2012.
  • “A Road Map for Strengthening Comparisons of International Health System Performance,” by H. E. Frech III, Stephen T. Parente, and John Hoff. Health Policy Outlook (AEI), forthcoming.
  • “Distinguishing between Heterogeneity and Inefficiency: Stochastic Frontier Analysis of the World Health Organization’s Panel Data on National Health Care Systems,” by William Greene. Health Economics, Vol. 13 (2004).
  • Explaining High Health Care Spending in the United States: An International Comparison of Supply, Utilization, Prices, and Quality, by David A. Squires. Commonwealth Fund, 2012.
  • “Gas Prices, Beer Taxes and GDL Programmes: Effects on Auto Fatalities among Young Adults in the US,” by Michael A. Morrisey and David C. Grabowski. Applied Economics, Vol. 43, No. 25 (2011).
  • “Influence of Definition-Based versus Pragmatic Birth Registration on International Comparisons of Perinatal and Infant Mortality: Population-Based Retrospective Study,” by K. S. Joseph. BMJ, Vol. 344 (2012).
  • International Differences in Mortality at Older Ages: Dimensions and Sources, edited by Eileen M. Crimmins, Samuel H. Preston, and Barney Cohen. National Academies Press, 2010.
  • “Is the United States an Outlier in Health Care and Health Outcomes? A Preliminary Analysis,” by William S. Comanor, H.E. Frech III, and Richard D. Miller Jr. International Journal of Health Care Finance and Economics, Vol. 6, No. 1 (2006).
  • “The OECD’s Study on Health Status Determinant: Roles of Lifestyle, Environment, Health-Care Resources, and Spending Efficiency: An Analysis,” by H.E. Frech III. AEI Working Paper #145, February 6, 2009.
  • “The Triple Aim: Care, Health, and Cost,” by Donald M. Berwick, Thomas W. Nolan, and John Whittington. Health Affairs, Vol. 27, No. 3 (2008).
  • “U.S. Health Care Costs: The Untold True Story,” by Mark V. Pauly. Health Affairs, Vol. 12, No. 3 (1993).
  • Who Shall Live? Health, Economics, and Social Choice, by Victor R. Fuchs. Basic Books, 1974.