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

Medical Progress Report
No. 2  October 2005


Older Drugs, Shorter Lives? An Examination of the Health Effects of the Veterans Health Administration Formulary

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

Executive Summary

This paper examines access to new drugs under the pharmacy benefits management system of the Veterans Health Administration (VHA). The VHA's National Formulary, implemented in 1997, discourages access to new drugs in an effort to control overall pharmaceutical costs. Some public figures have argued that this system should also apply to purchases under the new Medicare drug benefit, making the study of its effects on patient health particularly important.

Only 38% of the drugs approved in the 1990s, and 19% of the drugs approved by the FDA since 2000, are on the VA National Formulary. Only 22% (17) of the 77 priority-review drugs approved since 1997 are on the 2005 National Formulary.

The drugs used in the VA health system from 1999 to 2002 were older than the drugs used in the rest of the U.S. health-care system. For example, the percentages of VA and non-VA prescriptions for drugs less than five years old were 5.6% and 8.6%, respectively, and the percentages for drugs less than fifteen years old were 31.4% and 39.0%.

This paper estimates the impact of the use of new drugs on longevity, based on annual data on Medicaid drug use and mortality by state, disease, and year, for all fifty states during the period 1991-2001. These estimates imply that increased use of older drugs in the VA system, as a result of the Formulary, has reduced mean age at death of its patients by 0.17 years, or 2.04 months; the value of this reduction in longevity may be nearly $25,000 per person.

Moreover, demographic data published by the VA indicate that the life expectancy of veterans increased substantially before the National Formulary was introduced (during 1991-97) but did not increase, and may even have declined, after it was introduced (1997-2002).

There are many proposals in Congress to adopt a system similar to the VA National Formulary for purchases under the new Medicare drug benefit. These data suggest that such a proposal could reduce life span and survival rates among the Medicare population, raising serious questions about the wisdom of these proposals.

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

Introduction

The Medicare drug benefit (part D) will go into effect on January 1, 2006. Some people have proposed using the VA pharmacy benefit system, including the VA National Formulary*, as a model for the Medicare drug benefit. In this paper, I consider the wisdom of such a policy.

The VA National Formulary generated controversy when it was implemented because Congress "learned that the formulary prevents physicians from meeting the unique health-care needs of individual veterans and is overly restrictive" (Blumenthal and Herdman, 2000). Congress requested that the Institute of Medicine (IOM) review the experience with the National Formulary and formulary systems. The commission found that formularies and formulary systems (the many policies and procedures necessary to manage implementation of formularies) are an essential part of modern health-care systems and that the VHA therefore was justified in creating its National Formulary.[1] However, the IOM committee found almost no data relating the implementation and management of the National Formulary to the quality of the process and outcomes of veterans' care. To this end, this paper supplements and updates the commission's analysis. It reassesses the impact of the National Formulary system, paying particular attention to its impact on VA enrollees' access to new drugs and the relationship such access has to life expectancy and well-being.

To do this, I examine data on the fraction of drugs that are on the National Formulary, by period of FDA approval. I will also update calculations done by the commission on the extent to which priority-review drugs** approved since 1997 are on the National Formulary.

That a drug is not listed on the National Formulary does not necessarily mean that VA patients do not have access to the drug. A drug not listed on the National Formulary may be listed on one of twenty-three Veterans Integrated Service Networks (VISNs) or local formularies[2]; even if it is not, the patient may obtain access via a nonformulary exceptions process. Therefore, to assess the impact of the National Formulary system on the pattern of drug use, it is necessary to examine data on the drugs actually used by people in the VA system. The Medical Expenditure Panel Survey provides such data for the years 1996-2002 and allows us to compare VA drug use with non-VA drug use. Using these data, I will show that drugs used in the VA system are older than drugs used in the non-VA sector and that the gap has widened since the National Formulary was implemented.

I will then consider the effect of the VA Pharmacy Benefits Management (PBM) system on an important patient outcome: survival. There are two ways to do this. The first (indirect) way is to estimate the effect of using older drugs on the probability of survival, or life expectancy. I have estimated this effect in several previous papers, and I will present some new estimates here, based on longitudinal data by state, major disease category, and year, during the period 1990-2001. The second (direct) way is to compute estimates of the life expectancy of veterans from 1990 to 2002 (i.e., before and after the VA PBM system was implemented), and to compare them with data on the life expectancy of American men in general (the vast majority of veterans are men). I will compute these estimates from VetPop2001, the VA's official estimate and projection of the number and characteristics of veterans as of September 30, 2001.


* The VA National Formulary is a list of drugs, devices, and supplies that provides the basis for uniform national access to listed agents including drugs, devices, and supplies for all VHA facilities. It was implemented in 1997 by the Veterans Health Administration (VHA) to help control costs and improve the quality of drugs prescribed in the VHA’s health-care facilities, which include 172 hospitals, more than 600 ambulatory facilities, and 132 nursing homes.
** Priority-review drugs are drugs considered by the FDA to offer significant improvement compared with marketed products, in the treatment, diagnosis, or prevention of a disease.

Addition of New Drugs to the VA National Formulary

As indicated in the commission report[3], "under current policy, drugs newly approved by the FDA are considered for addition to the VA National Formulary only after a 1-year delay, except in special cases of important new 1P category drugs, that is, new chemical entities classified for priority review by the FDA."[4] In practice, that policy has meant adding new drugs for the treatment of HIV/AIDS with less than a year lag, whereas other 1P drugs have been added only after a year or more. Currently, these decisions are made by a consortium of the Medical Advisory Panel (MAP), VISN formulary leaders, and the VA PBM.[5]

The VHA policy of a one-year waiting period is a safety precaution that allows evidence of adverse drug effects to accumulate. It also provides time to compare the safety, efficacy, or cost-effectiveness of new drugs with existing therapeutic alternatives, or with drugs for similar indications. Such studies are usually not done during the FDA new drug-approval process. Data, especially in the peer-reviewed open literature, to inform a decision (on whether a new drug is an improvement over existing drug therapies) are generally not available until sometime after release, if at all. In fact, Sloan et al. (1997) noted a dearth of pharmacoeconomic or cost-effectiveness studies even beyond a year after market entry of new drugs. Waiting for a year does not guarantee that adequate comparative evaluations will be available.[6]

The commission reviewed the forty-two FDA 1P drugs approved in 1996, 1997, and 1998. Ten of the 1P drugs that were introduced before the implementation of the VA National Formulary were included in the initial version. Four drugs were subsequently approved and added, primarily for the treatment of HIV/AIDS. By July 1999, the 28 remaining 1P drugs had either been reviewed and not approved (5), had not been reviewed (21), or were pending (2). The reasons for disapproving additions included "no advantages over contract agents," "evidence regarding efficacy was inconclusive," and "safety/cost concerns." At the same time, the FDA Center for Drug Evaluation and Research 1998 Report to the Nation[7] proposed that 1P drugs "represent an advance in medical treatment" and described a number of the drugs that had been disapproved or not reviewed by the VHA as "notable 1998 new drug approvals." The MAP, VA PBM, and VISN formulary leaders must employ stringent evidentiary requirements for the addition of newly introduced drugs, since few are added to the National Formulary. As far as the committee could determine, however, there is no VHA policy or practice of identifying and reviewing new 1P drugs (for example, the twenty-one "not-reviewed" 1996, 1997, or 1998 1P drugs) or other new-to-market drugs in a systematic way.

VISN and local policies and practices, although variable, appear to be more permissive, so existing or newly introduced drugs are less likely to be added to the National Formulary than to the formularies of other organizations, or to VISN or local formularies. Listed drugs are also less likely to be deleted. One or more VISN or local formularies added 4 of the 5 disapproved 1P drugs and 4 of the 21 non-reviewed 1P drugs. In one case, 18 VISNs added clopidogrel (Plavix), a nationally non-reviewed 1P drug. A decision was then made at the national level not to add this drug to the National Formulary, but it remained on VISN formularies. Changes to these VHA formularies vary considerably from VISN to VISN.

To what extent are FDA-approved drugs listed on the VA National Formulary, so that all VHA patients are guaranteed access to them? To answer this question, I will calculate the percent of drugs approved by the FDA since 1950 on the 2005 VA National Formulary, by decade of FDA approval. I compiled a list of about 1,300 drugs approved, and their approval dates, from the Drugs@FDA Data Files.[8] I determined whether each of these drugs was on the VA National Formulary by examining data in the VA's National Drug File.[9] Figure 1 shows the percent of drugs on the 2005 VA National Formulary, by decade of FDA approval. The fractions of drugs approved in the 1950s, 1960s, 1970s, and 1980s on the VA National Formulary are almost identical: 52-53%. Only 38% of the drugs approved in the 1990s, however, and only 19% of the drugs approved since 2000, are on the VA National Formulary.

The Drugs@FDA Data Files don't indicate whether the drugs approved were priority-review or standard-review drugs. This information is available, though, for drugs approved since 1997 from New Drug Approval Reports, published by the FDA's Center for Drug Evaluation and Research.[10] The following table shows the number of new molecular entities approved by the FDA since 1997, by review status and formulary status.

Priority reviewStandard reviewTotal
Listed on 2005 NF171431
Not listed on 2005 NF6098158
Total77112189

Only 22% (17) of the 77 priority-review drugs approved since 1997 are on the 2005 National Formulary. This is lower than the percentage (33%) of 1P drugs approved in 1996, 1997, and 1998 that the IOM committee reported to be on the National Formulary.

Comparison of VA versus Non-VA Use of New Drugs

In what follows, I compare use of new drugs in the VA health system with their use in the rest of the U.S. health-care system. I use data from the Medical Expenditure Panel Survey (MEPS), which collects data on a nationally representative sample of families and individuals.[11]

MEPS data are currently available for the years 1996-2002. There is a Prescribed Medicines file for each year. This file contains records of all prescriptions obtained by households in the sample. Each record includes the National Drug Code of the drug and the amount paid for the prescription, by payer. There are twelve payers, and one of these is "Veterans."[12] I will define a "VA prescription" as a prescription for which the amount paid by veterans exceeded zero. In 1999, for example, there were 173,950 prescriptions; 5,083 (2.9%) of these were VA prescriptions.

I determined the year in which the FDA first approved the active ingredient of each prescription. I then defined the age of a prescription as the year in which the prescription occurred minus the year in which the FDA first approved the prescription's active ingredient. For example, the age of a 1999 prescription for a drug first approved in 1990 is nine years. I defined three variables indicating whether the age of the prescription was greater than 5, 10, and 15 years.[13] Finally, I calculated the mean values of these three variables, for both VA and non-VA prescriptions, using data for MEPS prescriptions during the years from 1999 to 2002.[14]

Figure 2 (page 4) shows the percent of 1999-2002 VA and non-VA prescriptions for drugs less than 5, 10, and 15 years old. All three measures indicate that the drugs used in the VA health system from 1999 to 2002 were older than the drugs used in the rest of the U.S. health-care system. For example, the percentages of VA and non-VA prescriptions for drugs less than five years old were 5.6% and 8.6%, respectively, and the percentages for drugs less than fifteen years old were 31.4% and 39.0%.

Since we have prescription data both pre- and post-implementation of the National Formulary, we can also assess whether the gap between VA and non-VA drug age widened over time.[15] From 1996 to 2002, new-drug use increased less quickly in the VA health system than in the rest of U.S. health care. The quantity of drugs less than ten years old increased by 1.4 percentage points per year in the non-VA sector, and by 0.6 percentage points per year in the VA sector. The proportion of drugs less than fifteen years old increased by 1.9 percentage points per year in the non-VA sector and had virtually no increase in the VA sector. These estimates are consistent with the hypothesis that implementation of the VA National Formulary beginning in 1997 reduced use of new drugs in the VA health-care system.

The Effect of Using Older Drugs on the Probability of Survival, or Life Expectancy

We have seen that only 16% of all drugs approved since 1997, and 22% of priority-review drugs, are listed on the 2005 VA National Formulary; that the drugs used in the VA health system from 1999 to 2002 were older than the drugs used in the rest of the U.S. health-care system; and that new-drug use increased more slowly from 1996 to 2002 in the VA health system than it did in the rest of U.S. health care. I will now consider the implications of these facts for a patient outcome that many people might consider the most important and that is undoubtedly the best measured: survival.

In what follows, I present new evidence on the impact of the use of new drugs on longevity, based on annual data on Medicaid drug use and mortality by state, disease, and year, for all fifty states, during the period 1991-2001.

A model based on these data[16] enables us to test the hypothesis that there have been above-average increases in mean age at death (in state-disease cells that have experienced above-average increases in the prescription of new drugs by Medicaid).[17] This analysis enables us to control for many potentially confounding variables, such as unobserved state-specific trends (e.g., state fiscal condition) that might affect mortality and be correlated with Medicaid drug use.[18]

I construct the mortality data from the 1991-2001 Multiple Cause of Death data files.[19] These contain records of every death in the U.S. (about 2 million per year), including data on where the death occurred, exact age at death, and cause of death.

State drug-use information is available for outpatient drugs purchased on or after January 1, 1991, by State Medicaid agencies.[20] In particular, we have quarterly data on the number of prescriptions, by National Drug Code (NDC) and state, for the period 1991-2004.[21]

The Centers for Medicare and Medicaid Services (CMS) data do not contain any information about the diseases for which the drugs were prescribed, but there is a good way to allocate the prescriptions by NDC by disease: by using data in the 1996-2001 Medical Expenditure Panel Survey Prescribed Medicines Files. These files contain about 1.5 million records of individual prescriptions. Each record contains both an NDC and a three-digit ICD9 diagnosis code. Hence, we can determine the relative frequency with which each NDC was used for different diseases. The MEPS diagnosis codes are quite detailed, so I aggregate them (and the mortality data) to broad disease groups, e.g., cardiovascular disease, cancer, and respiratory disease.

Note that there is a misalignment between the mortality data and the drug-use data: the mortality data pertain to all decedents, i.e., those who had been enrolled in Medicaid and those who hadn't, while the use data pertain only to the Medicaid program. It is reasonable to hypothesize, however, that changes in Medicaid drug use may be correlated, across states and diseases, and over time, with changes in non-Medicaid drug use (e.g., due to spillovers in prescribing). Changes in Medicaid drug use, which can be measured extremely precisely with the CMS data, might be considered a good indicator of changes in overall drug use.

By using data from another source, covering a more recent time period, I can test the hypothesis that the extent of use of new drugs in the Medicaid program is strongly correlated with the extent of use of new drugs in general. I have data from a private company, NDCHealth, on the number of prescriptions, by NDC, 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, diabetes drugs, osteoporosis/menopause drugs, and pain-management medications.[22] These data show that the extent of new-drug use in the Medicaid program strongly correlates with the extent of the use of new drugs in general. Controlling for disease-state, disease-year, and state-year effects, the data also indicate that longevity (mean age at death) increased more rapidly in state-disease cells experiencing higher increases in post-1990 drug use.

We can use these data to calculate how much of the increase in mean age at death from 1991 to 2001 is attributable to the increasing use of post-1990 drugs. From 1991 to 2001, mean age at death increased by 1.74 years, from 73.24 to 74.99 years, and the fraction of prescriptions that were for post-1990 drugs increased by 0.314. The increase in mean age at death attributable to increasing use of post-1990 drugs is estimated to be 0.79 years.[23] About 46%[24] of the total increase in mean age at death during the period 1991-2001 is attributable to the increasing use of post-1990 drugs. This is similar to the 40% share of longevity increase in fifty-two countries during 1986-2001 that I estimated to be attributable to new drug launches.[25]

The fraction of post-1990 drugs used in the VA health system during 1999-2002 (25.2%) was lower than the fraction of post-1990 drugs used in the non-VA sector (31.9%). The estimates imply that use of older drugs in the VA system reduced mean age at death of its patients by 0.17 years (= 2.53 * [31.9% - 25.2%]), or 2.04 months. Murphy and Topel (2003) argue that the value of a U.S. statistical life-year is not less than $150,000, which would imply that the per-patient value of this reduction in longevity is not less than $25,000.

Life Expectancy of Veterans, 1991–2002

Demographic data published by the VA enable us to compute the life expectancy of veterans before and after the National Formulary was implemented. Life-expectancy calculations are based on life tables. There are two types of life tables: cohort (or generation) life tables; and period (or current) life tables. The cohort life table presents the mortality experience of a particular birth cohort (e.g., all persons born in the year 1900) from the moment of birth through consecutive ages in successive calendar years. Based on age-specific death rates observed through consecutive calendar years, the cohort life table reflects the mortality experience of an actual cohort from birth until no lives remain in the group. To prepare just a single complete cohort life table requires data over many years. It is usually not feasible to construct cohort life tables entirely on the basis of observed data for real cohorts due to data unavailability or incompleteness (Shryock et al., 1971). For example, a life-table representation of the mortality experience of a cohort of persons born in 1970 would require the use of data projection techniques to estimate deaths into the future (Moriyama and Gustavus, 1972; Preston et al., 2001).

Unlike the cohort life table, the period life table does not represent the mortality experience of an actual birth cohort. Rather, the period life table presents what would happen to a hypothetical (or synthetic) cohort if it experienced throughout its entire life the mortality conditions of a particular period in time. Thus, for example, a period life table for 2002 assumes a hypothetical cohort subject throughout its lifetime to the age-specific death rates prevailing for the actual population in 2002. The period life table may thus be characterized as rendering a "snapshot" of current mortality experience and shows the long-range implications of a set of age-specific death rates that prevailed in a given year. Official government estimates of U.S. life expectancy are based on period life tables (Arias, 2004); my calculations of the life expectancy of veterans will also be based on period life tables.

Calculation of the life table is derived from the probability of death, which depends on the number of deaths and the midyear population for each age group observed during the calendar year of interest. The VA publishes historical data on and projections of the number of deaths and the number of living veterans, by age group and year, 1990-2030.[26] Data for 1991-2002 are shown in Table 2 (see Appendix, page 16). The top part of the table shows the number of veteran deaths during the year, by age group. The middle part shows the number of veterans alive at the beginning of the year, and the bottom part shows the mortality rate.[27]

Estimates of veterans' life expectancy during the period 1991-2002 are shown in Figure 3. Since the estimates are based on rough approximations, the average level of life expectancy should be viewed with caution. The mean value of life expectancy during the entire period is 6.6 years higher than the mean value of the life expectancy of all U.S. males at birth[28] (over 94% of veterans alive in 2002 were male).

Figure 3 indicates that veterans' life expectancy increased substantially before the National Formulary was introduced (during 1991-1997) but did not increase, and may even have declined, after it was introduced (1997-2002). Figure 4 juxtaposes the path of veterans' life expectancy with the path of life expectancy of all U.S. males at birth. The life expectancy at birth of all U.S. males increased after-as well as before-1997, although the rate of growth declined by about a third.

Conclusion

In this paper, I have examined access to new drugs under the Pharmacy Benefits Management system of the Veterans Health Administration. Since 1997, the VA National Formulary has played a key role in that system.

The fractions of drugs approved in the 1950s, 1960s, 1970s, and 1980s on the 2005 VA National Formulary are almost identical: 52-53%. Only 38% of the drugs approved in the 1990s, however, and 19% of the drugs approved since 2000, are on the VA National Formulary. Only 22% (17) of the 77 priority-review drugs approved since 1997 are on the 2005 National Formulary. This is lower than the percentage (33%) of priority-review drugs approved in 1996, 1997, and 1998 that the IOM committee reported to be on the National Formulary.

The drugs used in the VA health system from 1999 to 2002 were older than the drugs used in the rest of the U.S. health-care system. For example, the percentages of VA and non-VA prescriptions for drugs less than five years old were 5.6% and 8.6%, respectively, and the percentages for drugs less than fifteen years old were 31.4% and 39.0%.

The percent of drugs less than ten years old increased by 1.4 percentage points per year in the non-VA sector, and by 0.6 percentage points per year in the VA sector, during 1996-2002. The percent of drugs less than fifteen years old increased by 1.9 percentage points per year in the non-VA sector and had virtually no increase in the VA sector. These estimates are consistent with the hypothesis that implementation of the VA National Formulary beginning in 1997 reduced the use of new drugs in the VA health-care system.

I presented estimates of the impact of use of new drugs on longevity, based on annual data on Medicaid drug use and mortality by state, disease, and year, for all fifty states during the period 1991-2001. The estimates implied that the use of older drugs in the VA system reduced mean age at death of its patients by 0.17 years, or 2.04 months. Murphy and Topel (2003) argue that the value of a U.S. statistical life-year is not less than $150,000, which would imply that the per-patient value of this reduction in longevity is not less than $25,000.

I used demographic data published by the VA to compute the life expectancy of veterans before and after the National Formulary was implemented. Veterans' life expectancy increased substantially before the National Formulary was introduced (during 1991-1997) but did not increase, and may even have declined, after it was introduced (1997-2002). The life expectancy at birth of all U.S. males increased after-as well as before-1997, although the rate of growth declined by about a third.

There are many proposals in Congress to adopt a system similar to the VA National Formulary for purchases under the new Medicare drug benefit. These data suggest that this shift could reduce well-being, life span, and survival rates among the Medicare population, raising serious questions about the wisdom of these proposals.

 


Center for Medical Progress.

EMAIL THIS | PRINTER FRIENDLY

MPR 02 PDF (200 kb)

PRESS RELEASE

WHAT THE PRESS SAID:

Don't Scrimp on Quality, Daily Policy Digest, National Center for Policy Analysis, 10-25-05
Beware of side effects if eyeing VA Drug Plan, Stars and Stripes, 10-24-05
Rules & Red Tape, Mandates, Galen Institute, 10-24-05
Want Cheap Drugs? Don't Scrimp On Quality, Says Prof, Investor's Business Daily, 10-24-05

SUMMARY:
This report analyzes the implications of the VHA’s National Formulary, implemented in 1997 to reduce overall pharmaceutical costs by discouraging access to new drugs. Author Frank Lichtenberg of Columbia University and the National Bureau of Economic Research finds that this policy has had a dramatic effect on survival; though veterans’ life expectancy increased substantially before the National Formulary was introduced, it did not increase, and may have even declined, after it was introduced. There are many proposals in Congress to adopt a system similar to the VA National Formulary for purchases under the new Medicare drug benefit, but these data suggest that this change could reduce well-being, life span, and survival rates among the Medicare population, raising serious questions about the wisdom of these proposals.

TABLE OF CONTENTS:

EXECUTIVE SUMMARY

INTRODUCTION

Addition of New Drugs to the VA National Formulary

Figure 1: Percent of Drugs on 2005 VA National Formulary, by Decade of FDA Approval

Comparison of VA versus Non-VA Use of New Drugs

Figure 2: Percent of 1999-2002 VA and Non-VA Prescriptions for Drugs Less Than 5, 10, and 15 Years Old

The Effect of Using Older Drugs on the Probability of Survival, or Life Expectancy

Life Expectancy of Veterans, 1991–2002

Figure 3: Life Expectancy of Veterans, 1991–2002

CONCLUSION

Figure 4: Veterans’ Life Expectancy vs. Life Expectancy at Birth of All U.S. Males

ABOUT THE AUTHOR

REFERENCES

ENDNOTES

APPENDIX

Table 1: Priority Review Drugs Approved After 1997 Not Listed on 2005 National Formulary

Table 2: Demographic Data on Veterans, 1991–2002

 


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 © 2014 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