Manhattan Institute for Policy Research.
search  
 
Subscribe   Subscribe   MI on Facebook Find us on Twitter Find us on Instagram      
 
   
 

Medical Progress Report

No. 4 July 2007


Why Has Longevity Increased More in Some States than in Others? The Role of Medical Innovation and Other Factors

Frank R. Lichtenberg, Columbia University and National Bureau of Economic Research

Executive Summary

It is no surprise that Americans are living longer today than in previous generations. A typical baby born in 1900 was expected to live to about age 45. Today, life expectancy at birth is about 78. Less well known, however, is the fact that the gains in life expectancy have not been uniform across the country. In his new study—the first of its kind—Columbia University researcher Frank Lichtenberg set out to find out which states are the leaders, which ones are the laggards, and why.

Lichtenberg began by constructing life-expectancy estimates of residents in all fifty states using data from the National Center for Health Statistics. He found that in 2004, on average, residents of Hawaii (81.3 years) and Minnesota (80.3 years) lived six or seven years longer than residents of Mississippi and Louisiana (74.2 years).

In addition, he found that while nationwide life expectancy increased by 2.33 years from 1991 to 2004, the increase varied greatly among the states. Certain states—New York (4.3 years), California (3.4 years), and New Jersey (3.3 years)—led the way, while others–Oklahoma (0.3 years), Tennessee (0.8 years), and Utah (0.9 years) trailed the national average by significant margins.

Lichtenberg then set out to examine why this “longevity increase gap” exists by measuring the impact of several factors that researchers agree could affect life expectancy. He found that, although some obvious suspects—obesity, smoking, and the incidence of HIV/AIDS—played a role, the most important factor was “medical innovation.”

Specifically, Lichtenberg found that longevity increased the most in those states where access to newer drugs—measured by mean “vintage” (FDA approval year)—in Medicaid and Medicare programs has increased the most. In fact, about two-thirds of the potential increase in longevity—the longevity increase that would have occurred if obesity, income, and other factors had not changed—is attributable to the use of newer drugs. According to his calculations, for every year increase in drug vintage there is about a two-month gain in life expectancy. These represent important findings given the fact that the costs of prescription drugs continue to receive a great deal of attention in the ongoing debate over health-care policy, while their benefits are often overlooked.

Lichtenberg also estimated impacts on productivity and per-capita medical expenditure. He concluded that states adopting medical innovations more rapidly had faster labor productivity growth, conditional on income growth and other factors, perhaps due to reduced absenteeism from chronic medical ailments. He also found that states that use newer drugs did not experience above-average increases in overall medical expenditure, which contradicts the common perception that advances in medical technology inevitably result in increased health-care spending.

There are two ways to improve the average quality of U.S. health care. One way is to give best-practice care to people who are currently receiving less than best-practice care (e.g., to ensure that all heart-attack patients take beta blockers after they are released from the hospital). The other way is to improve best-practice care by shifting the technological frontier (e.g., to develop new ways to monitor, treat, and even prevent heart disease). This study indicates that the development and use of new medical goods and services, which shift the technological frontier, have been responsible for many recent gains in the health and longevity of Americans.

SUMMARY OF FINDINGS

Variation in Life Expectancy Gains

• From 1991 to 2004, nationwide, life expectancy at birth increased 2.33 years; life expectancy at age 65 increased by 1.29 years.

• The states with the largest increases in life expectancy were the District of Columbia (5.7 years), New York (4.3 years), California (3.4 years), New Jersey (3.3 years), and Illinois (3.0 years).

• The states with the smallest increases in life expectancy were Oklahoma (0.3 years), Tennessee (0.8 years), Utah (0.9 years), Alabama (1.0 years), and West Virginia (1.0 years).

• In the eight states with the smallest increases, life expectancy increased by 0.31–1.16 years. In the eight states with the largest increases, life expectancy increased by 2.60–4.33 years.

Factors Affecting Life Expectancy

• Growth in obesity and, interestingly, growth in income were both inversely related to (and presumably reduced) the growth in life expectancy.

• If obesity and income had not increased, life expectancy at birth would have increased by 3.88 years from 1991 to 2004, instead of the actual 2.33-year increase. Thus, 3.88 years is the “potential increase in life expectancy at birth.”

• Of the 3.88-year potential increase in life expectancy at birth, medical innovation (i.e., the increase in Medicaid and Medicare drug vintage) accounted for 2.43 years (63%). The declines in AIDS incidence and smoking accounted for 0.23 and 0.12 years (6% and 3%), respectively. About 1.1 years (28%) of the potential increase in life expectancy at birth is unexplained.

• If obesity and income had not increased, life expectancy at age 65 would have increased by 2.15 years from 1991 to 2004, instead of the actual 1.29-year increase. Thus, 2.15 years is the “potential increase in life expectancy at age 65.”

• Of the 2.15-year potential increase in life expectancy at age 65, medical innovation (i.e., the increase in Medicaid and Medicare drug vintage) accounted for 1.19 years (55%). The declines in AIDS incidence and smoking accounted for 0.07 and 0.12 years (3% and 5%), respectively. About 0.8 years (36%) of the potential increase in life expectancy at age 65 is unexplained.

Medical Expenditure Impact

• Increases in income, education, smoking, and the incidence of AIDS tend to increase per-capita medical expenditure; expanded health coverage reduces it.

• States that had the greatest increase in drug vintage did not experience above-average increases in overall medical expenditure. While use of newer drugs has increased some types of medical expenditure, it has reduced other types, and the expenditure reductions approximately offset the expenditure increases.

• Although use of newer drugs does not appear to have increased annual medical expenditure, it probably has increased lifetime medical expenditure slightly as the use of newer drugs increased life expectancy at birth by 2.43 years. But the implied cost per life-year gained is quite low.

Productivity Impact

• States with larger increases in Medicaid drug vintage had faster productivity growth, conditional on income growth and other factors.

• The increase in Medicaid drug vintage is estimated to have increased output per employee by about 1% per year. Much of this may be attributable to increased hours worked per employee.

About the Author

Professor Frank Lichtenberg currently serves as the Courtney C. Brown Professor of Business at the Columbia University Graduate School of Business as well as a research associate of the National Bureau of Economic Research. His work has focused on how new technologies affect the productivity of companies, industries and nations. Dr. Lichtenberg’s studies have ranged from the impact of pharmaceutical innovation to the consequences of leveraged buyouts for efficiency and employment. This research has earned numerous fellowships and awards, including the 1998 Schumpeter Prize and a 2003 Milken Institute Award for Distinguished Economic Research, as well as grants by the National Science Foundation, the National Institute of Standards and Technology, Merck and Co., the Fulbright Commission, and the Alfred P. Sloan Foundation. He has worked for several U.S. government agencies, including the Department of Justice and the Congressional Budget Office, as well as taught at Harvard University and the University of Pennsylvania.

Dr. Lichtenberg received a BA in history from the University of Chicago and an MA and PhD in economics from the University of Pennsylvania.

Abstract

The rate of increase in longevity has varied considerably across U.S. states since 1991. This paper examines the effect of medical innovation (changes in drug vintage), behavioral risk factors (obesity, smoking, and AIDS incidence), and other variables (education, income, and health insurance coverage) on longevity using longitudinal state-level data. This approach controls for the effects of unobserved factors that vary across states but are relatively stable over time (e.g., climate and environmental quality); and unobserved factors that change over time but are invariant across states (e.g., changes in federal government policies). We also analyze interstate variation in productivity (output per employee) growth and in the growth of per-capita medical expenditure (total and by type).

States in which the vintage of both self- and provider-administered drugs grew faster than average had above-average increases in life expectancy, whether or not we adjust for state-specific changes in the distribution of disease. Life expectancy grew more slowly in states with larger increases (or slower declines) in AIDS, obesity, and smoking rates. States with high income growth had smaller longevity increases.

States with larger increases in Medicaid drug vintage had faster productivity growth, conditional on income growth and the other factors. The increase in Medicaid drug vintage is estimated to have increased output per employee by about 1% per year. Much of this may be attributable to increased hours worked per employee.

Increases in income, education, smoking, and the incidence of AIDS tend to increase per-capita medical expenditure; expanded health insurance coverage reduces it. States in which drug vintage has increased the most have not had above-average increases in overall medical expenditure. While use of newer drugs has increased some types of medical expenditure, it has reduced other types, and the expenditure reductions approximately offset the expenditure increases. Although use of newer drugs does not appear to have increased annual medical expenditure, it probably has increased lifetime medical expenditure, but the increase in lifetime medical cost per life-year gained from using newer drugs has been quite low.

The estimates indicate that the growth in obesity and the growth in income both reduced the growth in life expectancy. If obesity and income had not increased, life expectancy at birth would have increased by 3.88 years. The increases in Medicaid and Medicare drug vintage account for 2.43 years (63%) of the “potential increase” in life expectancy. The declines in AIDS incidence and smoking account for 0.23 and 0.12 (6% and 3%), respectively, of the potential increase in life expectancy. About 1.1 years (28%) of the potential increase in life expectancy at birth is unexplained. Differences in drug vintage explain some of the interstate variation in life expectancy, but the fraction of cross-sectional variance explained is smaller than the fraction of aggregate time-series variance (growth) explained.


Introduction

During the twentieth century, U.S. life expectancy at birth increased by almost thirty years (63%), from 47.3 years in 1900 to 77.0 years in 2000. (See Figure 1.) Nordhaus (2002) estimated that “to a first approximation, the economic value of increases in longevity over the twentieth century is about as large as the value of measured growth in non-health goods and services” (p. 17). Murphy and Topel (2005) observed that “the historical gains from increased longevity have been enormous. Over the 20th century, cumulative gains in life expectancy were worth over $1.2 million per person for both men and women. Between 1970 and 2000 increased longevity added about $3.2 trillion per year to national wealth, an uncounted value equal to about half of average annual GDP over the period.”

The rate of increase in longevity has varied considerably across states. Figure 2 shows the increase in life expectancy at birth during the period 1991–2004[1], by state. In the eight states with the smallest increase, life expectancy increased by only 0.31–1.16 years. In the eight states with the largest increase, life expectancy increased by 2.60–4.33 years. This paper seeks to help answer the question, why has longevity increased more in some states than in other states?

Longevity is likely to depend on a number of factors, including access to health care and medical innovations, exogenous changes in disease incidence (e.g., the appearance of new diseases such as HIV/AIDS), income, education, and behavioral risk factors (e.g., obesity and smoking).

A recent study by the Harvard School of Public Health emphasized the impact that ethnicity, through genetic predispositions, plays in determining longevity and how different concentrations of various ethnic groups throughout the United States affect the disparity in longevity. By using a longitudinal, state-by-state approach, we control for factors such as ethnicity, demographics, and environmental quality that vary across the states but generally remain constant or change very slowly over time. This approach also allows us to control for factors that do change over time but do not vary across the states (e.g., changes in federal government policies, scientific discoveries, and the Dow Jones industrial average).

In addition to interstate variation in longevity growth, we will analyze interstate variation in productivity (output per employee) growth and in the growth of per-capita medical expenditure (total and by type, e.g., expenditure on physicians, prescription drugs, and hospital care). In particular, we will examine how medical innovation (use of newer medical products) has affected the level and structure of health expenditure.

The overall conceptual framework of the paper is depicted in Figure 3.

Previous literature suggests that technological innovation in general—and new goods in particular—plays a key role in economic growth. In Section I, we briefly survey this literature, discuss the measurement of medical innovation, including adjustment for state-specific changes in the distribution of disease, and consider why the rate of innovation may vary across states. Section II describes the econometric models that we will estimate.

Section III describes the data sources and presents descriptive statistics. Empirical results are presented in Section IV. Implications of the estimates are discussed in Section V. Section VI presents a summary and conclusions.


I. Innovation: Literature Review and Measurement Issues

While longevity is probably influenced by a number of factors, medical innovation—the use of new medical goods and services—is likely to play a preeminent role in explaining longevity growth. Economists believe that the development of new products is the main reason that people are better off today than they were several generations ago. Grossman and Helpman (1993) argue that “innovative goods are better than older products simply because they provide more ‘product services’ in relation to their cost of production.” Bresnahan and Gordon (1996) state simply that “new goods are at the heart of economic progress.” Jones (1998) argues that “technological progress [is] the ultimate driving force behind sustained economic growth” (p. 2) and that “technological progress is driven by research and development (R&D) in the advanced world” (p. 89). Bils (2004) makes the case that “much of economic growth occurs through growth in quality as new models of consumer goods replace older, sometimes inferior, models.”

The best way to measure utilization of medical innovations (embodied technological change) is to measure the mean vintage of medical goods and services used. The vintage of a good is the year in which the good was first used. For example, the vintage of the drug atorvastatin (Lipitor) is 1997—the year that the drug was approved by the FDA. We seek to test the hypothesis that, ceteris paribus, people using newer, or later vintage, medical goods and services will be in better health and will therefore live longer. This hypothesis is predicated on the idea that these goods and services, like other R&D-intensive products, are characterized by embodied technological progress.[2]

A number of econometric studies (Bahk and Gort, 1993; Hulten, 1992; Sakellaris and Wilson, 2001, 2004) have investigated the hypothesis that capital equipment employed by U.S. manufacturing firms embodies technological change, that is, that each successive vintage of investment is more productive than the last. Equipment is expected to embody significant technical progress because of the relatively high R&D intensity of equipment manufacturers. The method that has been used to test the equipment-embodied technical change hypothesis is to estimate manufacturing production functions, including (mean) vintage of equipment as well as quantities of capital and labor. These studies have concluded that technical progress embodied in equipment is a major source of manufacturing productivity growth.

Although most previous empirical studies of embodied technical progress have focused on equipment used in manufacturing, embodied technical progress may also be an important source of economic growth in health care. One important input in the production of health—pharmaceuticals—is even more R&D-intensive than equipment. According to the National Science Foundation, the R&D intensity of drugs and medicines manufacturing is 74% higher than the R&D intensity of machinery and equipment manufacturing. Therefore, it is quite plausible that there is also a high rate of pharmaceutical-embodied technical progress.

Measuring vintage

The general definition of vintage we will use is:

In principle, we would like to measure the vintage of all drugs, all other medical goods and services, and even all other products and services. Unfortunately, this is not possible.

We will measure the mean vintage of outpatient prescription drugs paid for by the state’s Medicaid program and the mean vintage of drugs administered by providers (e.g., chemotherapy) to Medicare beneficiaries. The number of prescriptions paid for by Medicaid is very large: according to the Medical Expenditure Panel Survey, in 1997, Medicaid paid for about 201 million prescriptions—11% of all U.S. prescriptions. Moreover, we show in the Appendix that the extent of utilization of new drugs in the Medicaid program is strongly correlated with the extent of utilization of new drugs in general: the vintage of non-Medicaid (and all) prescriptions tended to increase more in states with larger increases in the vintage of Medicaid prescriptions.

Drugs administered by providers are quite different from self-administered drugs, and Medicare pays for a substantial fraction of the former. In 2004, Medicare paid providers $7.6 billion for performing 522 million pharmaceutical procedures.[3] Medicare data on the frequency of use of non-pharmaceutical services (e.g., lab and surgical procedures) are also available. However, because of asymmetries in FDA regulation, determining the vintage of non-pharmaceutical medical services is far more difficult than determining the vintage of pharmaceutical products and procedures.

Since we will not control for the vintage of non-pharmaceutical medical services, and the latter may be correlated with drug vintage, the drug vintage coefficients that we estimate may to some extent reflect the effect of other medical innovation as well as the effect of drug innovation. The coefficients could also reflect the effect of nonmedical innovation—for example, consumer use of information technology. We will attempt to control for the latter by estimating models that control for the percent of state residents who use a computer at home.


Adjusting for state-specific changes in the distribution of disease

If there have been state-specific changes in the distribution of disease, and drug vintage is correlated with disease severity (e.g., newer drugs tend to treat less severe diseases), the coefficient on drug vintage could be biased. However, we can eliminate any potential bias by constructing an alternative (fixed-weighted) index of drug vintage.

Consider the following simplified model of life expectancy:

The change in life expectancy is directly related to the change in drug vintage and inversely related to the change in the percent of patients with the high-severity disease.

Suppose that drugs for the low-severity disease (nervous system disorders) tend to be newer than drugs for the high-severity disease (cardiovascular disease), so that there is an inverse correlation across states between : states with smaller increases in mean severity will have larger increases in drug vintage. In this case, failure to control for changes in severity will result in overestimation of the effect of drug vintage on life expectancy.

We will control for the incidence of one highly severe disease—AIDS—but unfortunately, data on the incidence of other diseases, by state and year, are not available. Therefore direct measurement of mean disease severity (or the percent of patients with high-severity diseases) by state and year is not feasible. However, provided that the distribution of drugs utilized, by therapeutic class, is closely related to the distribution of patients, by disease, we can eliminate any potential bias in the vintage coefficient by using the following fixed-weighted index of drug vintage:

Changes over time in the fixed-weighted index V’ are entirely due to within-therapeutic class changes in drug vintage, not at all to between-class changes, that is, shifts in the distribution of drugs by therapeutic class. In contrast, changes in the standard vintage index are due to between- as well as within-class changes in vintage.

We will construct fixed-weighted indexes of drug vintage using data from the Veterans Administration’s National Drug File (U.S. Dept. of Veterans Affairs, 2007) on the therapeutic class of each product. The VA drug classification is hierarchical and comprises more than 500 classes and subclasses. We will classify drugs at the highest level of the VA classification system, which has thirty-two classes. Table 1 shows data on the distribution and vintage of Medicaid prescriptions in 1991 and 2004, by major VA therapeutic class. In 2004, two classes of drugs (central nervous system medications and cardiovascular medications) accounted for half of Medicaid prescriptions. The share of Medicaid prescriptions that were central nervous system medications increased from 19% in 1991 to 29% in 2004. The mean vintage of central nervous system medications increased much more than the mean vintage of cardiovascular medications (16.5 years vs. 6.5 years). However, for the nation as a whole, the fixed-weighted vintage index increased more from 1991 to 2004 than the standard index (11.4 years vs. 9.4 years).

We will estimate models using both the standard index and the fixed-weighted index of drug vintage. Performing this sensitivity analysis is useful, but eliminating the effects of shifts in the distribution of drugs by therapeutic class on vintage is not necessarily appropriate. If the rate of innovation varies across diseases/drug classes, states may benefit from innovation by changing the distribution of drugs consumed, by class, as well as by using newer drugs within drug classes.


Potential reasons for variation in the rate of increase of drug vintage

The rate of increase in drug vintage may vary across states because of both interstate differences in the types of diseases afflicting the population and differences in the drugs used to treat given diseases. Suppose that

Even if the increase in the mean vintage of drugs to treat each disease is the same in every state, differences in the fractions of state residents who have various diseases will result in interstate variation in the increase in the mean vintage of drugs.[4]

The relative incidence of various diseases does vary across states. This is illustrated by Figure 4, which plots the state-level incidence rate (cases per 100,000) of colon and rectum cancer against the incidence rate of prostate cancer for males in 2002. The correlation across states between these two incidence rates is not significantly different from zero (p-value = 0.61).

Moreover, because of medical practice variation, the increase in the mean vintage of drugs to treat any given disease is likely to vary across states. Medical practice variation is a well-documented phenomenon: there are 2,514 citations for this term in the PubMed database.[5] The Dartmouth Atlas of Health Care Project (Wennberg, 2006) has demonstrated “glaring variations in how health care is delivered across the United States.”

Skinner and Staiger (2005) argue that medical practice variation may be partly due to variation in the frequency and likelihood of informational exchanges through networks or other social activities, which may in turn be related to average educational attainment and other measures of social capital. They compared the adoption of several important innovations during the twentieth century, ranging from advances at mid-century in hybrid corn and tractors, with medical innovations in the treatment of heart attacks at the end of the century. They found a very strong state-level correlation with regard to the adoption of new and effective technology, and this correlation held across a variety of industries and time periods. These results are suggestive of state-level factors associated with barriers to adoption. These barriers may be related to information or network flows, given that farmers, physicians, and individual computer users often conduct their business in small and isolated groups and therefore are most vulnerable to potential information asymmetries.

Interstate differences in government health-care policy also contribute to practice variation. In the last few years, some state Medicaid programs and private managed-care plans have restricted access to certain drugs, especially newer, more expensive drugs. One important type of restriction is a “prior authorization” requirement: a prescription will not be dispensed without prior authorization by program officials. Lichtenberg (2005d) examined the effect of access restrictions on the vintage of drugs used by Medicaid enrollees. The sample included fifty brand-name drugs in six important therapeutic classes: antidepressants, antihypertensives, cholesterol-lowering drugs, diabetic drugs, osteoporosis/menopause drugs, and pain management medications. The extent of access restrictions varied considerably across states. Twelve states did not restrict any of the fifty drugs. Five states restricted over 47% of the drugs, and one—Vermont—restricted forty-three of the fifty drugs. The vintage of Medicaid prescriptions increased more slowly in states that imposed more access restrictions.[6]


II. Econometric Model

We will investigate the effects of drug vintage, behavioral risk factors, and other variables on life expectancy, productivity, and medical expenditure by estimating models of the following form:

In principle, there is some risk of feedback, or reverse causality, from life expectancy to some of the explanatory variables, especially mean income and education. Ceteris paribus, increases in life expectancy lead to an increase in the fraction of the population that is elderly. As shown in Figure 5, mean income and education of elderly people are significantly lower than those of non-elderly people. Hence unobserved shocks that increase a state’s longevity could reduce its mean income and education, causing a downward bias in the coefficients of these variables. However, the share of the population that is elderly need not be increasing faster in states with larger increase in life expectancy; these states could have higher birthrates or higher net immigration rates.

In practice, the share of the population that is elderly is increasing faster in states with larger increase in life expectancy, but the relationship is not very strong. By using estimates of this relationship and the age profiles shown in Figure 5, we obtained estimates of the feedback effect of life expectancy on income and education, via population age structure. These calculations indicated that the downward biases in the income and education coefficients in the longevity equations would be extremely small.


III. Data Sources and Descriptive Statistics

Life expectancy. The government does not publish data on life expectancy by state, so we constructed estimates using data on the number of deaths by age group, year, and state of residence from the Multiple Cause-of-Death Mortality Data from the National Vital Statistics System of the National Center for Health Statistics.[8] Each record in the microdata is based on information abstracted from death certificates filed in vital-statistics offices of each state and the District of Columbia. The average number of records (deaths) per year is about 2.3 million. We also used population data from the Centers for Disease Control (CDC) Wonder Bridged-Race Population Estimates.[9] As shown in Figure 6, the population-weighted means of my state estimates of life expectancy are quite similar to the National Center for Health Statistics (NCHS) national estimates.

Productivity and per-capita income. These data were obtained from two Bureau of Economic Analysis Regional Economic Accounts databases: the Gross Domestic Product by State database;[10] and the State Annual Personal Income database.[11]

Per-capita medical expenditure. The Centers for Medicare and Medicaid Services (CMS) Health Accounts by State database[12] provides data on the following categories of health expenditure, by state and year (1980–2005): Total Health Care Expenditure, Hospital Care, Physician Services, Other Professional Services, Dental Services, Home Health Care, Prescription Drugs, Other Non-Durable Medical Products, Durable Medical Products, and Nursing Home Care.

Vintage of Medicaid prescriptions. The mean vintage of Medicaid prescriptions is defined as follows:

Data on were obtained from the ndc_denorm table in the Multum Lexicon database.[15] There are currently more than 2,100 active ingredients in this database. Table 2 shows the top twenty-five active ingredients contained in 2004 Medicaid prescriptions, ranked by number of prescriptions.

Data on vinta were obtained from the Drugs@FDA database, produced by the FDA Center for Drug Evaluation and Research.[16] This database includes several tables. The product table enumerates properties of the products included in each application, including their active ingredient(s). The supplements table provides the approval history for each application, including dates of approval. We define vinta as the earliest approval date of any product that contains active ingredient a.

Vintage of Medicare drug treatments. Medicare is a health insurance program for people aged 65 or older, people under age 65 with certain disabilities, and people of all ages with end-stage renal disease (permanent kidney failure requiring dialysis or a kidney transplant). All Medicare enrollees are covered by Medicare Part A (hospital insurance). Most Medicare enrollees elect to pay a monthly premium for Part B. Medicare Part B helps cover doctors’ services and outpatient care. It also covers some other medical services that Part A doesn’t cover, such as some of the services of physical and occupational therapists, and some home health care. Part B helps pay for these covered services and supplies when they are medically necessary. In 2004, about 39 million Americans were enrolled in Medicare Part B.

Prior to January 1, 2006, when Medicare Part D was established, Medicare did not pay for most outpatient drugs, but the Medicare Part B (medical insurance) program did pay for drugs administered by health-care providers, for example, chemotherapy.

The Medicare drug vintage measure is similar to the Medicaid drug vintage measure, with one exception. For reasons discussed below, the Medicare index is expenditure-weighted rather than quantity-weighted:


Data on expend_medicare_drugdit were obtained from annual Physician/Supplier Procedure Summary (PSPS) Master Files produced by CMS for each of the years from 1991 to 2004. Each file is a 100% summary of all Part B Carrier and DMERC Claims processed through the Common Working File and stored in the National Claims History Repository. The files are large; the 2004 file has more than 12 million records. The file enables us to compute total submitted services and charges, total allowed services and charges, total denied services and charges, and total payment amounts, by Medicare carrier and procedure. In most cases, there is a one-to-one correspondence between a carrier and a state, so we can measure utilization and expenditure, by procedure and state.

As discussed in the technical documentation for the PSPS Master Files, Medicare carriers often make erroneous reports of service counts, but not of expenditures:

Service counts for drugs should be reported using pricing units, e.g., J0120: Injection, Tetracycline up to 250 mg. In this example, 250 mg = 1 pricing unit or service. If the injection were for 500 mg, then the pricing unit or service would be equal to 2, i.e., 500mg / 250mg = 2 pricing units or services. Many carriers are reporting the milligrams in the service count and MTUS Fields, e.g., 250 mg instead of 1 pricing unit. As a result the number of services are inflated, thereby deflating the average allowed charge.[17]

As shown in Figure 7, these reporting errors appear to cause spurious fluctuations in aggregate Medicare drug treatment service counts but not in expenditures. Therefore, while we believe that a quantity-weighted vintage index is preferable to an expenditure-weighted index, we will use an expenditure-weighted index of Medicare drug treatments because of errors in reporting service counts.

Data on eda were obtained from the ndc_denorm table in the Multum Lexicon database.

Table 3 shows the top twenty-five active ingredients contained in 2004 Medicare drug treatments, ranked by total services count. Comparison of Tables 2 and 3 indicates that the drugs administered by providers to Medicare beneficiaries are quite different from outpatient drugs used by Medicaid beneficiaries.

Demographic characteristics and behavioral risk factors. Data on body mass index (BMI), current smoking participation, health insurance coverage, and educational attainment were obtained from the Behavioral Risk Factor Surveillance System (BRFSS),[18] which is the world’s largest telephone survey. The BRFSS was established by the CDC in 1984 and was designed to collect state-level data. By 1994, all states, the District of Columbia, and three territories were participating in the BRFSS.

Data on the incidence of AIDS (the number of AIDS cases reported by state and local health departments) were obtained from the CDC’s AIDS Public Information Data Set.[19] This data set contains counts of AIDS, by demographics; location (region and selected metropolitan areas); case definition; month/year and quarter-year of diagnosis, report, and death (if applicable); and HIV exposure group (risk factors for AIDS). The data set covers the period 1981–2002. As noted above, the measure of AIDS incidence that we will include in our model of life expectancy will be the number of AIDS cases reported per 100,000 population lagged by two years. Using this measure allows us to have the sample period end in 2004 rather than 2002. Also, Lichtenberg (2006) provides evidence that even before highly active retroviral therapy was introduced in the mid-1990s, life expectancy of AIDS patients at time of diagnosis was 3.7 years, so overall life expectancy may depend on lagged AIDS incidence more than it depends on contemporaneous AIDS incidence.[20]

Table 4 shows population-weighted sample means of the variables included in eq. (1), by year. Table 5 shows sample means, by state. Figure 8 shows the increase in the fixed-weighted drug vintage index 1991–2004, by state.


IV. Empirical Results

Estimates of eq. (1) based on the standard index of Medicaid drug vintage are shown in Table 6. Estimates of eq. (1) based on the fixed-weighted index of Medicaid drug vintage are shown in Table 7. Overall, the two sets of estimates are fairly similar. We will discuss the estimates based on the fixed-weighted index.

The dependent variable in column 1 of Table 7 is life expectancy at birth. The coefficients on both Medicaid and Medicare drug vintage are positive and highly significant (p-value < .0001). This indicates that states in which the vintage of both self- and provider-administered drugs grew faster than average had above-average increases in life expectancy. The coefficients on the three behavioral risk factors (aids, bmi_gt25, and now_smoke) are all negative and significant. Life expectancy grew more slowly in states with larger increases (or slower declines) in AIDS, obesity, and smoking rates. The coefficients on educational attainment and health insurance coverage are not statistically significant. The coefficient on per-capita income is negative, and significant: states with high income growth had smaller longevity increases, ceteris paribus. This may be consistent with findings by Ruhm (2000, 2002, 2003, 2004, 2006, and forthcoming).

The dependent variable in column 2 of Table 7 is life expectancy at age 65. The signs and significance of these coefficients are similar to those in column 1. Below, we will use these coefficients to assess the contributions of medical innovation and changes in risk factors and income to longevity growth from 1991 to 2004. But first, we will review the estimates of the productivity and medical expenditure regressions in Table 7.

The dependent variable in column 3 of Table 7 is real gross state product per employee. The coefficient on Medicaid drug vintage (but not on Medicare drug vintage) is positive and highly significant (p-value < .0001). States with larger increases in Medicaid drug vintage had faster productivity growth, conditional on income growth and the other factors in eq. (1). The increase in Medicaid drug vintage is estimated to have increased output per employee by about 1% per year. Much of this may be attributable to increased hours worked per employee. Based on a study of disease-level household survey data from 1982 to 1996, Lichtenberg (2005c) concluded that pharmaceutical innovation reduced the number of work-loss days per employed person by 1.0% per year.

Productivity growth is likely to depend on non-pharmaceutical as well as pharmaceutical innovations. Moreover, Skinner and Staiger (2005) found a very strong state-level correlation with regard to the adoption of new and effective technologies, and this correlation held across a variety of industries and time periods. Therefore, the coefficient on Medicaid drug vintage in the productivity regression may be overestimated: it may be capturing the productivity effect of other, unmeasured innovations.

Measuring the adoption of most innovations by state and year is not feasible, but there is one important innovation whose diffusion can be tracked: use of personal computers in the home. In six of the years from 1994 to 2003, respondents to the Current Population Survey indicated whether they used a computer at home. As shown in Figure 9, the percent of people using computers at home increased from 25% in 1994 to 62% in 2003. The rate of increase varied considerably across states.

We did not include the computer-use measure in our basic model, because doing so would require a 57% reduction in sample size. However, we assessed the sensitivity of our estimates to controlling for computer use. We found that changes in Medicaid drug vintage were uncorrelated across states with changes in computer use, both unconditionally and controlling for income, education, and other factors. When computer use is included in the longevity and productivity equations, its coefficient is not significant in any equation. Controlling for computer use increases the Medicaid drug vintage coefficient in the productivity equation by 26%; it reduces the Medicaid drug vintage coefficient in the life expectancy at birth and at age 65 equations by 25% and 17%, respectively, but they remain highly significant. Thus at least one attempt to control for the adoption of nonmedical innovations does not have a substantial impact on our estimates.

Now let’s consider the estimates of the per-capita medical expenditure equations. The coefficient on Medicaid drug vintage in the drug expenditure equation is .035 and is highly significant. This suggests that a one-year increase in Medicaid drug vintage causes drug expenditure to increase by 3.5%. This is quite consistent with Lichtenberg’s (2006) estimate of the slope of the vintage-price profile based on cross-sectional microdata from the 2002 Medical Expenditure Panel Survey; he found that a one-year increase in vintage was associated with a 3.0% increase in the price of a prescription. Increases in educational attainment and the incidence of AIDS also increase drug expenditure. But states whose Medicare drug vintage is growing rapidly have lower growth in per-capita drug expenditure.

The coefficients on the Medicaid drug vintage coefficient in the other expenditure equations (cols. 5–8) indicate that use of newer drugs is associated with increased utilization of home health care and nursing-home care and lower expenditure on physicians. The coefficients on both the Medicaid and Medicare drug coefficients in the total expenditure equation (col. 9) are insignificantly different from zero. This indicates that states in which drug vintage has increased the most have not had above-average increases in overall medical expenditure. While use of newer drugs has increased some types of medical expenditure, it has reduced other types, and the expenditure reductions approximately offset the expenditure increases. This suggests that pharmaceutical-embodied technological change, like equipment-embodied technical change, is non-neutral (Kopp and Smith, 1985; Bartel and Lichtenberg, 1987; Baltagi and Rich, 2005).

The other coefficients in column 9 suggest that increases in income, education, smoking, and the incidence of AIDS tend to increase per-capita medical expenditure and that expanded health insurance coverage reduces it.


V. Implications

Now we will use our estimates to assess the effects of the various factors on changes in U.S. life expectancy and on interstate differentials in life expectancy. The contribution of each factor to the 1991–2004 change in life expectancy is the coefficient of that factor in column 1 or 2 of Table 7 times the 1991–2004 change in the mean of that factor in the last row of Table 4. As shown in the middle column of Table 8, life expectancy at birth increased by 2.33 years from 1991 to 2004. The estimates indicate that the growth in obesity and the growth in income both reduced the growth in life expectancy. If obesity and income had not increased, life expectancy at birth would have increased by 3.88 years. The increases in Medicaid and Medicare drug vintage account for 2.43 years (63%) of the “potential increase” in life expectancy. The declines in AIDS incidence and smoking account for 0.23 and 0.12 year (6% and 3%), respectively, of the potential increase in life expectancy. About 1.1 years (28%) of the potential increase in life expectancy at birth is unexplained.

As shown in the last column of Table 8, life expectancy at age 65 increased by 1.29 years from 1991 to 2004. If obesity and income had not increased, life expectancy at age 65 would have increased by 2.15 years. The increases in Medicaid and Medicare drug vintage account for 1.19 years (55%) of the potential increase in life expectancy at age 65. The declines in AIDS incidence and smoking account for 0.07 and 0.12 year (3% and 5%), respectively, of the potential increase in life expectancy. About 0.8 year (36%) of the potential increase in life expectancy at age 65 is unexplained.[21]

Although use of newer drugs does not appear to have increased annual medical expenditure, it probably has increased lifetime medical expenditure. The increase in the latter may be approximately equal to total medical expenditure during the 2.43 additional years of life attributable to increasing drug vintage. As shown in Figure 10, in 1996 mean medical expenditure of people aged 75–84 was $6,153—56% more than the mean medical expenditure of all Americans. This implies that the increase in lifetime medical cost per life-year gained from using newer drugs has been about $6,153. Medical interventions that cost this amount are generally considered to be highly cost-effective.

Differences in drug vintage explain some of the interstate variation in life expectancy, but the fraction of cross-sectional variance explained is smaller than the fraction of aggregate time-series variance (growth) explained. For example, as shown in Table 5, the mean value of New Jersey’s Medicaid fixed-weighted index of drug vintage is almost three years higher than the value of Tennessee’s index. (These states used the newest and oldest drugs, respectively.) Our estimates imply that this difference would result in about a six-month difference in life expectancy at birth. This is about 20% of the mean actual life-expectancy differential (2.3 years) between the two states.


VI. Summary and Conclusions

The rate of increase in longevity has varied considerably across states since 1991. This paper has examined the effect of medical innovation, behavioral risk factors (obesity, smoking, and AIDS incidence), and other variables (education, income, and health insurance coverage) on longevity using longitudinal state-level data. This approach controls for the effects of unobserved factors that vary across states but are relatively stable over time (e.g., climate and environmental quality) and unobserved factors that change over time but are invariant across states (e.g., changes in federal government policies). We also analyzed interstate variation in productivity (output per employee) growth and in the growth of per-capita medical expenditure (total and by type, e.g., expenditure on physicians, prescription drugs, and hospital care).

We found that states in which the vintage of both self- and provider-administered drugs grew faster than average had above-average increases in life expectancy, whether or not we adjusted for state-specific changes in the distribution of disease. However, since we were unable to control for the vintage of non-pharmaceutical medical services—and the latter may be correlated with drug vintage—the drug vintage coefficients that we estimated may to some extent reflect the effect of other medical innovation as well as the effect of drug innovation.

Life expectancy grew more slowly in states with larger increases (or slower declines) in AIDS, obesity, and smoking rates. Consistent with a number of recent studies, states with high income growth had smaller longevity increases, ceteris paribus.

States with larger increases in Medicaid drug vintage had faster productivity growth, conditional on income growth and the other factors. The increase in Medicaid drug vintage is estimated to have increased output per employee by about 1% per year. Much of this may be attributable to increased hours worked per employee. In principle, the coefficient on Medicaid drug vintage in the productivity regression may be overestimated: it may be capturing the productivity effect of other, unmeasured innovations. But controlling for a potentially important nonmedical innovation—computer use in the home—did not have a substantial impact on our estimates.

Increases in income, education, smoking, and the incidence of AIDS tend to increase per-capita medical expenditure; expanded health insurance coverage reduces it.

States in which drug vintage has increased the most have not had above-average increases in overall medical expenditure. While use of newer drugs has increased some types of medical expenditure, it has reduced other types, and the expenditure reductions approximately offset the expenditure increases. This suggests that pharmaceutical-embodied technological change, like equipment-embodied technical change, is non-neutral. Although use of newer drugs does not appear to have increased annual medical expenditure, it probably has increased lifetime medical expenditure. But the increase in lifetime medical cost per life-year gained from using newer drugs has been quite low.

The estimates indicate that the growth in obesity and the growth in income both reduced the growth in life expectancy. If obesity and income had not increased, life expectancy at birth would have increased by 3.88 years, not just 2.33 years. The increases in Medicaid and Medicare drug vintage account for 2.43 years (63%) of the “potential increase” in life expectancy. The declines in AIDS incidence and smoking account for 0.23 and 0.12 year (6% and 3%), respectively, of the potential increase in life expectancy. About 1.1 years (28%) of the potential increase in life expectancy at birth is unexplained. Differences in drug vintage explain some of the interstate variation in life expectancy, but the fraction of cross-sectional variance explained is smaller than the fraction of aggregate time-series variance (growth) explained.

Appendix: Correlation across States between Changes in the Vintage of Medicaid and Non-Medicaid Prescriptions

This appendix describes a test of the hypothesis that the extent of utilization of new drugs in the Medicaid program is strongly correlated with the extent of utilization of new drugs in general. We had access to data from a private company, NDCHealth, on the number of prescriptions, by NDC code, state (and five U.S. territories), month (January 2001–December 2003), and payer (Medicaid, other third party, and cash), for six important therapeutic classes of drugs: antidepressants, antihypertensives, cholesterol-lowering drugs, diabetic drugs, osteoporosis/menopause drugs, and pain management medications. Here are some summary statistics:

These data were used to estimate the following equation:[22]

Two alternative measures of vintage were used: the mean FDA approval year; and the share of prescriptions containing active ingredients approved after 1980. Estimates of eq. (1) are shown in Table 1. In all four equations, the estimate of is positive and highly statistically significant (p-value <.0001). This indicates that the extent of utilization of new drugs in the Medicaid program is strongly correlated with the extent of utilization of new drugs in general. The vintage of non-Medicaid (and all) prescriptions tended to increase more in states with larger increases in the vintage of Medicaid prescriptions.

References

Arias, E. (2006). United States Life Tables, 2003. National Vital Statistics Reports 54, no. 14. Hyattsville, Md.: National Center for Health Statistics, http://www.cdc.gov/nchs/data/nvsr/nvsr54/nvsr54_14.pdf.

Bahk, Byong-Hyong, and Michael Gort (1993). “Decomposing Learning by Doing in New Plants,” Journal of Political Economy 101: 561–83.

Baltagi, Badi H., and Daniel P. Rich (2005). “Skill-Biased Technical Change in U.S. Manufacturing: A General Index Approach,” Journal of Econometrics 126, no. 2 (June): 549–70.

Bartel, Ann P., and Frank R. Lichtenberg (1987). “The Comparative Advantage of Educated Workers in Implementing New Technology,” Review of Economics and Statistics 69, no. 1 (February): 1–11.

Bils, Mark (2004). “Measuring the Growth from Better and Better Goods,” NBER Working Paper no. 10606 (July), http://www.nber.org/papers/w10606.

Boucekkine, Raouf, David de la Croix, and Omar Licandro. “Vintage Capital,” Department of Economics, European University Institute, Eco. No. 2006/08, http://cadmus.iue.it/dspace/bitstream/1814/4346/1/ECO2006-8.pdf.

Bresnahan, Timothy F., and Robert J. Gordon (1996). The Economics of New Goods (Chicago: University of Chicago Press).

Grossman, Gene M., and Elhanan Helpman (1993). Innovation and Growth in the Global Economy (Cambridge, Mass.: MIT Press).

Hulten, Charles R. (1992). “Growth Accounting When Technical Change Is Embodied in Capital,” The American Economic Review 82, no. 4 (September): 964–80.

Jones, Charles (1998). Introduction to Economic Growth (New York: Norton).

Kopp, Raymond J., and V. Kerry Smith (1985). “The Measurement of Non-Neutral Technological Change,” International Economic Review 26, no. 1 (February): 135–59.

Lai, D. J., P. M. Tarwatera, and R. J. Hardy (2006). “Measuring the Impact of HIV/AIDS, Heart Disease and Malignant Neoplasms on Life Expectancy in the USA from 1987 to 2000,” Public Health 120: 486–92.

Lichtenberg, Frank (2005a). “Pharmaceutical Knowledge-Capital Accumulation and Longevity,” in Measuring Capital in the New Economy, ed. Carol Corrado, John Haltiwanger, and Dan Sichel (Chicago: University of Chicago Press), 237–69.

——— (2005b). “The Impact of New Drug Launches on Longevity: Evidence from Longitudinal Disease-Level Data from 52 Countries, 1982–2001,” International Journal of Health Care Finance and Economics 5: 47–73.

——— (2005c). “Availability of New Drugs and Americans’ Ability to Work,” Journal of Occupational and Environmental Medicine 47, no. 4 (April): 373–80.

——— (2005d). “The Effect of Access Restrictions on the Vintage of Drugs Used by Medicaid Enrollees,” American Journal of Managed Care 11 (special issue): SP7–SP13.

——— (2006). “The Effect of Using Newer Drugs on Admissions of Elderly Americans to Hospitals and Nursing Homes: State-Level Evidence from 1997–2003,” Pharmacoeconomics 24, Suppl. 3, 2006, 5-25.

Miller, Richard D., and H. E. Frech (1999). The Productivity of Health Care and Pharmaceuticals: An International Comparison (Washington, D.C.: American Enterprise Institute).

Murphy, Kevin M., and Robert H. Topel (2005). “The Value of Health and Longevity,” NBER Working Paper no. W11405 (June).

Murray, C. J. L., et al. (2006). “Eight Americas: Investigating Mortality Disparities across Races, Counties, and Race Counties in the United States,” PLoS Med 3, no. 9: e260.

Nordhaus, William D. (2002). “The Health of Nations: The Contribution of Improved Health to Living Standards,” NBER Working Paper no. W8818 (March), at SSRN: http://ssrn.com/abstract=302579.

Ruhm, C. J. (2000). “Are Recessions Good for Your Health?” Quarterly Journal of Economics 115, no. 2: 617–50.

Ruhm, C. J., and W. E. Black (2002). “Does Drinking Really Decrease in Bad Times?” Journal of Health Economics 21, no. 4: 659–78.

Ruhm, C. J. (2003). “Good Times Make You Sick,” Journal of Health Economics 22, no. 4: 637–58.

——— (2004). “Healthy Living in Hard Times,” Journal of Health Economics 24, no. 2: 341–63.

——— (2006). “A Healthy Economy Can Break Your Heart,” NBER Working Paper no. w12102 (March).

——— (forthcoming). “Mortality Increases during Economic Upturns,” International Journal of Epidemiology.

Sakellaris, Plutarchos, and Dan Wilson (2001). “The Production-Side Approach to Estimating Embodied Technological Change,” Finance and Economics Discussion Series 2001-20, Board of Governors of the Federal Reserve System.

——— (2004). “Quantifying Embodied Technological Change,” Review of Economic Dynamics 7, no. 1: 1–26.

Skinner, Jonathan, and Douglas Staiger (2005). “Technology Adoption from Hybrid Corn to Beta Blockers,” National Bureau of Economic Research, Working Paper no. 11251, (March), http://www.nber.org/papers/w11251.

Solow, R. (1960). “Investment and Technological Progress,” in Mathematical Methods in Social Sciences 1959, ed. K. Arrow, S. Karlin, and P. Suppes (Stanford, Calif.: Stanford University Press), 89.104.

U.S. Dept. of Veterans Affairs (2007). Pharmacy Benefits Management Strategic Healthcare Group, National Drug File, http://www.pbm.va.gov/NationalFormulary.aspx.

Wennberg, John (2006). The Care of Patients with Severe Chronic Illness: A Report on the Medicare Program, The Dartmouth Atlas of Health Care 2006 (Hanover, N.H.: Dartmouth Medical School Center for the Evaluative Clinical Sciences), http://www.dartmouthatlas.org/atlases/2006_Chronic_Care_Atlas.pdf.

 

Endnotes

  1. Because of limitations on available data, this paper will analyze changes in longevity during the period 1991–2004.
  2. Solow (1960, p. 91) argued that “many if not most innovations need to be embodied in new kinds of durable equipment before they can be made effective. Improvements in technology affect output only to the extent that they are carried into practice either by net capital formation or by the replacement of old-fashioned equipment by the latest models.” We hypothesize that innovations may be embodied in nondurable goods (e.g., drugs) and services as well as in durable equipment.
  3. Source: CMS, Medicare Part B Physician/Supplier Data by BETOS, calendar year 2004,
    http://www.cms.hhs.gov/MedicareFeeforSvcPartsAB/Downloads/BETOS04.pdf.
  4. Our econometric model will control (via state fixed effects) for the effects of permanent, or relatively stable, differences between states in the relative incidence of various diseases.
  5. http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Search&db=PubMed&term=medical+practice+
    variation&tool=QuerySuggestion
    .
  6. Lichtenberg (2006) presents a theoretical argument that the vintage of drugs is also likely to depend on the extent of prescription drug coverage, and empirical evidence that supports this argument.
  7. Arizona is excluded from the sample because it does not participate in the Medicaid Drug Rebate Program.
  8. Murray et al. (2006) also computed state and local estimates of life expectancy.
  9. Vintage 2004, http://wonder.cdc.gov/Bridged-Race-v2004.html. We computed life expectancy using the following age classification: under 1 year, 1–4 years, 5–9 years, 10–14 years, 15–19 years, 20–24 years, 25–29 years, 30–34 years, 35–39 years, 40–44 years, 45–49 years, 50–54 years, 55–59 years, 60–64 years, 65–69 years, 70–74 years, 75–79 years, 80–84 years, 85 years and over.
  10. http://www.bea.gov/regional/gsp.
  11. http://www.bea.gov/regional/spi.
  12. http://www.cms.hhs.gov/NationalHealthExpendData/05_NationalHealthAccountsStateHealthAccounts.asp.
  13. http://www.cms.hhs.gov/MedicaidDrugRebateProgram/SDUD/list.asp.
  14. http://www.cms.hhs.gov/MedicaidDrugRebateProgram/09_DrugProdData.asp.
  15. http://www.multum.com/Lexicon.htm.
  16. http://www.fda.gov/cder/drugsatfda/datafiles/default.htm.
  17. CMS, “2004 Limitations for the Physician/Supplier Procedure Summary Master File.”
  18. http://www.cdc.gov/BRFSS/index.htm.
  19. http://www.cdc.gov/hiv/software/apids.htm.
  20. By 2001, life expectancy of AIDS patients at time of diagnosis is estimated to have increased to about 26 years.
  21. The unexplained component is reflected in the year fixed effects of eq. (1).
  22. This equation was estimated by weighted least squares, weighting by the total number of prescriptions, or the number of non-Medicaid prescriptions, in state i in month t.

 


Center for Medical Progress.

EMAIL THIS | PRINTER FRIENDLY

MPR 04 PDF

OP-ED
Yes, New Drugs Save Lives, Frank R. Lichtenberg, Washington Post, July 11, 2007
TELEVISION
"ABC News Now", 07-11-07
RADIO
Bloomberg News Radio, 07-11-07
BBC News Radio, 07-11-07
IN THE PRESS
Hot fun, healthy futures make Expo a must-go, Indianapolis Star, 07-12-07
How To Lead A Long And Miserable Life, New York Free Press, 07-11-07
Increased Life Expectancy Thanks to "Big Pharma"?, Reason Online, 07-11-07
New findings about longevity and screening for genetic disorders, Scripps Howard News Service, 07-11-07
Hawaii tops nation for life expectancy, study finds, Honolulu Advertiser, 07-11-07
Are Survivors' Costs a Pro-Life Issue?, ConcurringOpinions.com, 7-11-07
Long Live New Yorkers!, BusinessWeek, July 11, 2007
New York Leads U.S. States in Longevity Improvement, Study Says, Bloomberg News, July 11, 2007
Report: New Yorkers Living Longer Through Chemistry, New York Sun, July 11, 2007

TABLE OF CONTENTS:

EXECUTIVE SUMMARY

SUMMARY OF FINDINGS

ABOUT THE AUTHOR

Abstract

Introduction

Figure 1. U.S. Life Expectancy at Birth, 1900-2003

Figure 2. Increase in Life Expectancy at Birth 1991-2004, by State

Figure 3. Conceptual Framework

I. Innovation: Literature Review and Measurement Issues

Figure 4. Annual Incidence Rates (cases per 100,000) of Prostate and Colon & Rectum Cancer, Males, 2002, by State

II. Econometric Model

Figure 5. Age-Income and Age-Education Profiles

III. Data Sources and Descriptive Statistics

Figure 6. Comparison of Population-Weighted Mean of My State-level Estimates of Life Expectancy at Birth to NCHS National Estimate

Figure 7. Reported Aggregate Medicare Drug Treatment Service Counts and Allowed Charges, 1991-2004

Figure 8. Increase in Fixed-Weighted Drug Vintage Index 1991-2004, by State

IV. Empirical Results

Figure 9. Percentage of People Using Computer at Home, 1994-2003

V. Implications

Figure 10. Mean Medical Expenditure per Person in 1996, by Age

VI. Summary and Conclusions

Table 1. Distribution and Vintage of Medicaid Prescriptions in 1991 and 2004, by Major Therapeutic Class

Table 2. Top 25 Active Ingredients Contained in 2004 Medicaid Prescriptions, Ranked by Number of Prescriptions

Table 3. Top 25 Active Ingredients Contained in 2004 Medicare Drug Treatments, Ranked by Total Services Count

Table 4. Population-Weighted Sample Means, by Year

Table 5. Sample Means, by State (Average Values during 1991-2004)

Table 6. WLS Estimates of Equation 1 Based on the Standard Index of Medicaid Drug Vintage

Table 7. WLS Estimates of Equation 1 Based on the Fixed-Weighted Index of Medicaid Drug Vintage

Table 8. Estimated Effects of Various Factors on Changes in U.S. Life Expectancy

Appendix: Correlation across States between Changes in the Vintage of Medicaid and Non-Medicaid Prescriptions

Appendix Table 1. The Relationship between the Vintage of Medicaid Rxs and the Vintage of Other (or All) Rxs

REFERENCES

ENDNOTES

 


Home | About MI | Scholars | Publications | Books | Links | Contact MI
City Journal | CAU | CCI | CEPE | CLP | CMP | CRD | ECNY
Thank you for visiting us.
To receive a General Information Packet, please email support@manhattan-institute.org
and include your name and address in your e-mail message.
Copyright © 2009 Manhattan Institute for Policy Research, Inc. All rights reserved.
52 Vanderbilt Avenue, New York, N.Y. 10017
phone (212) 599-7000 / fax (212) 599-3494