Saturday, July 29, 2006

Life Expectancy and Infant Mortality are Unreliable Measures for Comparing the U.S. Health Care System to Others

Excerpts from here. See the original for tables and references

How does the United States health care system fare when compared to the rest of the industrialized world? This is an important question. Accurately measuring our health care system relative to those of other nations can yield insight into the types of health care policies America should pursue. New York Times columnist Paul Krugman has expressed the view that the U.S. health care system is inferior:

The United States spends far more on health care than other advanced countries. Yet we don't appear to receive more medical services. And we have lower life expectancy and higher infant mortality rates than countries that spend less than half as much per person. How do we do it?

Life expectancy and infant mortality are two measures that are widely cited, yet seldom questioned. This is unfortunate, because life expectancy and infant mortality tell us little about the efficacy of a health care system. This paper examines the deficiencies of using life expectancy and infant mortality to measure a health care system. It also examines the question: How should we measure a health care system?......

Any statistic that accurately measures health-care systems across nations must satisfy three criteria. First, the statistic must assume actual interaction with the health care system. Second, it must measure a phenomenon that the health care system can actually affect. Finally, the statistic must be collected consistently across nations.

Under the first criterion, the phenomenon being measured must be one in which the individual actually has contact with the health care system. More specifically, he must have contact with a health care professional, be it a doctor, nurse, lab technician, etc. A statistic measuring the rate of cancer survival satisfies this criterion, since diagnosis and treatment of cancer requires health care professionals. By contrast, a statistic measuring the rate of car accidents would not satisfy such a criteria since health care professionals are not essential to identifying car accidents.

Some statistics may assume interaction with the health care system, but the phenomena they measure are not ones on which the health care system can have any meaningful impact. Take, for example, the rate of cancer incidence. While this statistic assumes interaction with the health care system (an incidence of cancer cannot be known without the diagnosis of a health care professional), there is little a health care system can do about the rate of cancer. Rather, cancer incidence is affected by factors such as genetics, diet, lifestyle, etc., over which the health care system has no control. Thus, to be an adequate measure of the effectiveness of a health care system, a statistic must measure a phenomenon that health care professionals can actually affect.

Finally, a statistic must be collected consistently across nations. While this seems simple in theory, in practice it is quite complicated. Nations use diverse definitions of health phenomena. This leads to some nations excluding a segment of their populations from the collection of a statistic while other nations include those segments. In such circumstances, cross-national comparisons are largely meaningless. Thus, for health care systems across countries to be meaningful, there should be little to no variation in how statistics are collected. As shown below, both life expectancy and infant mortality are poor measures of a health care system because each fails to satisfy at least one of the above criteria.

Life Expectancy

Life expectancy is a poor statistic for determining the efficacy of a health care system because it fails the first criterion of assuming interaction with the health care system. For example, open any newspaper and, chances are, there are stories about people who die "in their sleep," in a car accident or of some medical ailment before an ambulance ever arrives. If an individual dies with no interaction with the health care system, then his death tells us little about the quality of a health care system. Yet all such deaths are computed into the life expectancy statistic.

Life expectancy also largely violates the second criterion - a health care system has, at most, minimal impact on longevity. One way to see this is to reexamine the table constructed by the Center for Economic and Policy Research. The interpretation that the Center for Economic and Policy Research wants readers to derive from Table 1 is that the United States would be better off with a system of universal health care. However, a careful examination of that table yields a more accurate interpretation: There is no relationship between life expectancy and spending on health care. Greece, the country that spends the least per capita on health care, has higher life expectancy than seven other countries, including Belgium, Denmark, Finland, Germany, Netherlands, the United Kingdom and the United States. Spain, which spends the second least per capita on health care, has higher life expectancy than ten other countries that spend more.

More robust statistical analysis confirms that health care spending is not related to life expectancy. Studies of multiple countries using regression analysis found no significant relationship between life expectancy and the number of physicians and hospital beds per 100,000 population or health care expenditures as a percentage of GDP. Rather, life expectancy was associated with factors such as sanitation, clean water, income, and literacy rate. A recent study examined cross-national data from 1980 to 1998. Although the regression model used initially found an association between health care expenditure and life expectancy, that association was no longer significant when gross domestic product (GDP) per capita was added to the model. Indeed, GDP per capita is one of the more consistent predictors of life expectancy.

Yet the United States has the highest GDP per capita in the world, so why does it have a life expectancy lower than most of the industrialized world? The primary reason is that the U.S. is ethnically a far more diverse nation than most other industrialized nations. Factors associated with different ethnic backgrounds - culture, diet, etc. - can have a substantial impact on life expectancy. Comparisons of distinct ethnic populations in the U.S. with their country of origin find similar rates of life expectancy. For example, Japanese-Americans have an average life expectancy similar to that of Japanese.

A good deal of the lower life expectancy rate in the U.S. is accounted for by the difference in life expectancy of African-Americans versus other populations in the United States. Life expectancy for African-Americans is about 72.3 years, while for whites it is about 77.7 years. What accounts for the difference? Numerous scholars have investigated this question. The most prevalent explanations are differences in income and personal risk factors. One study found that about one-third of the difference between white and African-American life expectancies in the United States was accounted for by income; another third was accounted for by personal risk factors such as obesity, blood pressure, alcohol intake, diabetes, cholesterol concentration, and smoking and the final third was due to unexplained factors. Another study found that much of the disparity was due to higher rates of HIV, diabetes and hypertension among African Americans. Even studies that suggest the health care system may have some effect on the disparity still emphasize the importance of factors such as income, education, and social environment.

A plethora of factors influence life expectancy, including genetics, lifestyle, diet, income and educational levels. A health care system has, at best, minimal impact. Thus, life expectancy is not a statistic that should be used to inform the public policy debate on health care.

Infant Mortality

At first glance, infant mortality appears to be a good measure of a health care system. First, it assumes interaction with a health care system since most babies born in the industrialized world are born in a hospital or other health care facility. It also satisfies the second criterion of assuming that health care professionals can affect the outcome, since doctors and nurses have a direct impact on the survival chances of a newborn. If infant mortality were accepted as an adequate measure based on those two criteria alone, then the U.S. health care system is one of the least effective in the industrialized world. This can be seen by constructing a table using the data on infant mortality utilized in the report from the Physicians for a National Health Program. Table 2 shows that on infant mortality, the U.S. ranks below all nations save New Zealand.

But infant mortality tells us a lot less about a health care system than one might think. The main problem is inconsistent measurement across nations. The United Nations Statistics Division, which collects data on infant mortality, stipulates that an infant, once it is removed from its mother and then "breathes or shows any other evidence of life such as beating of the heart, pulsation of the umbilical cord, or definite movement of voluntary muscles... is considered live-born regardless of gestational age." While the U.S. follows that definition, many other nations do not. Demographer Nicholas Eberstadt notes that in Switzerland "an infant must be at least 30 centimeters long at birth to be counted as living." This excludes many of the most vulnerable infants from Switzerland's infant mortality measure.

Switzerland is far from the only nation to have peculiarities in its measure. Italy has at least three different definitions for infant deaths in different regions of the nation. The United Nations Statistics Division notes many other differences. Japan counts only births to Japanese nationals living in Japan, not abroad. Finland, France and Norway, by contrast, do count births to nationals living outside of the country. Belgium includes births to its armed forces living outside Belgium but not births to foreign armed forces living in Belgium. Finally, Canada counts births to Canadians living in the U.S., but not Americans living in Canada. In short, many nations count births that are in no way an indication of the efficacy of their own health care systems. The United Nations Statistics Division explains another factor hampering consistent measurement across nations:

...some infant deaths are tabulated by date of registration and not by date of occurrence... Whenever the lag between the date of occurrence and date of registration is prolonged and therefore, a large proportion of the infant-death registrations are delayed, infant-death statistics for any given year may be seriously affected.

The nations of Australia, Ireland and New Zealand fall into this category. Registration problems hamper accurate collection of data on infant mortality in another way. Looking at data from 1984-1985, Eberstadt argued that, "Underregistration of infant deaths may also be indicated by the proportion of infant deaths reported for the first twenty-four hours after birth." Eberstadt found that in the U.S. and Canada more than a third of all infant death occurred during the first day, but in Sweden and France they accounted for less than one-fifth. Table 3 shows that the pattern still holds today.

Inconsistent measurement explains only part of the difference between the U.S. and the rest of the world. Were measurements to be standardized, according to Eberstadt, "America might move from the bottom third toward the middle, but it would be unlikely to advance into the top half." Another factor affecting infant mortality Eberstadt identifies is parental behavior. Pregnant women in other countries are more likely to either be married or living with a partner. Pregnant women in such households are more likely to receive prenatal care than pregnant women living on their own. In the U.S., pregnant women are far more likely to be living alone. Although the nature of the relationship is still unclear (it is possible that mothers living on their own are less likely to want to be pregnant), it likely leads to a higher rate of infant mortality in the U.S. In summary, infant mortality is measured far too inconsistently to make cross-national comparisons useful. Thus, just like life expectancy, infant mortality is not a reliable measure of the relative merits of health care systems.

Conclusion

Life expectancy and infant mortality are wholly inadequate comparative measures for health care systems. Life expectancy is influenced by a host of factors other than a health care system, while infant mortality is measured inconsistently across nations. Neither of these measures provides the United States with conclusive guidance on health care policy, let alone serve as reliable evidence that a system of universal health care "should be implemented in the United States."

Do measures that would permit accurate cross-national comparisons of health care systems exist? The most exhaustive source of cross-national data is the Organization for Economic Co-operation and Development (OECD). Yet the OECD notes that in most cases its data is not "internationally comparable" because "there is a lack of international agreement on the most promising indicators and many definitions of each indicator that could be adopted."

To rectify this problem, the OECD and the Commonwealth Fund have embarked on a collaborative effort to develop comparable measures across nations. Called the "OECD Health Care Quality Indicators Project," it is taking the "first steps towards a comprehensive reporting system for quality of care in OECD member countries." A recent report updating the progress of this project looks promising. For example, one standard that an indicator must meet is its "susceptibility to being influenced by the health care system." The researchers pose important questions on this regard, including, "Can the health care system meaningfully address this aspect or problem?" and "Does the health care system impact on the indicator independent of confounders like patient risk?" In other words, these statistics will assume interaction with a health care system and measure phenomena that a health care system actually affects. Furthermore, the aim of this project is to assure that data is collected consistently across nations, so that national policymakers have "the opportunity to compare the performance of their health care delivery systems against a peer group"

While the project researchers have chosen many indicators that measure phenomena that are actually affected by a health care system, comparability issues across nations remain. For example, one indicator measures the fatality rate within 30 days of those diagnosed with acute myocardial infarction (heart attack). However, the report notes that some "countries are able to track patients after hospital discharge, [while] some are not."

Hopefully such difficulties can be resolved as the project progresses. In the meantime, policymakers, pundits and reporters should stop referring to life expectancy and infant mortality as meaningful comparative measures of health care systems.

***************************

For greatest efficiency, lowest cost and maximum choice, ALL hospitals and health insurance schemes should be privately owned and run -- with government-paid vouchers for the very poor and minimal regulation. Both Australia and Sweden have large private sector health systems with government reimbursement for privately-provided services so can a purely private system with some level of government reimbursement or insurance for the poor be so hard to do?

Comments? Email me here. If there are no recent posts here, the mirror site may be more up to date. My Home Page is here or here.

***************************

No comments: