Content uploaded by Richard A Hirth
Author content
All content in this area was uploaded by Richard A Hirth on Oct 21, 2015
Content may be subject to copyright.
What Is the Price of Life and Why Doesn’t It Increase
at the Rate of Inflation?
A
N ARTICLE in the
February 2000 is-
sue of JAMA con-
cluded that annual
retinal screening for
many individuals with type 2 dia-
betes mellitus may not be war-
ranted on grounds of cost-effective-
ness. Vijan et al
1
reported that,
compared with biannual screen-
ing, annual retinopathy screening for
low-risk patients with diabetes costs
more than $100000 for each addi-
tional quality-adjusted life year
(QALY) gained. The results of a
study published in the March 2000
issue of the New England Journal of
Medicine concluded that extending
hospital stays beyond 4 days for pa-
tients with uncomplicated myocar-
dial infarctions was economically
unattractive, costing more than
$105000 per QALY gained.
2
These
studies demonstrate that com-
monly used interventions may not
be worthwhile investments of health
care resources. By contrast, a study
published in the June 2000 issue
of the Annals of Internal Medicine
concluded that, compared with no
treatment, sildenafil (Viagra) is a
cost-effective treatment for erectile
dysfunction, producing an incre-
mental QALY for the relatively low
cost of $11 000.
3
The latter study
raises questions about whether many
health care insurers were hasty in de-
ciding that they would not add silde-
nafil to the list of services covered
by their health plans.
4
In each of the 3 articles noted
above, the authors made promi-
nent mention of a threshold by
which cost-effectiveness experts
judge the affordability of health care
services: those interventions that
produce a QALY for $50000 or less
are a bargain, whereas those that
require $100000 or more are con-
sidered unaffordable. (A QALY is a
measure that allows comparisons be-
tween health care interventions that
save lives and those that improve
quality of life. For example, an in-
tervention that cures a patient of a
health condition with a quality of life
halfway between perfect health and
death yields 0.5 QALYs per year,
whereas an intervention that saves
a person’s life and returns him or her
to perfect health yields 1 QALY per
year.) The $50000 to $100000
QALY cutoff goes back to at least
1982.
5
A commonly used justifica-
tion for this threshold figure has
been that $50000 per QALY gained
was approximately the cost-
effectiveness ratio calculated for the
use of dialysis for patients with
chronic renal failure. Renal dialysis
is a federal entitlement to all US citi-
zens under Medicare. As the argu-
ment goes, if the US government
thinks that dialysis should be of-
fered to all who need it, then inter-
ventions with similar or better cost-
effectiveness should likewise be
offered to everyone.
The $50000 to $100000 per
QALY gained amount is often
cited in the literature as the cost-
effectiveness threshold, as re-
flected in the 3 economic analyses
described above. For example, in a
review of 228 cost-effectiveness
analyses of pharmaceutical inter-
ventions, Neumann et al
6
found that
the investigators in 34% of the stud-
ies refer to the $50000 per QALY
figure when discussing the implica-
tions of their findings. Most of the
others cite no threshold. Many phy-
sicians have expressed skepticism
about the role of cost-effectiveness
information in making health care
decisions.
7,8
Is it any wonder that
they are skeptical, given that the
$50000 to $100000 per QALY
threshold has persisted for 2 de-
cades without adjustment? After all,
if society was willing to spend
$50000 for a QALY 20 years ago,
shouldn’t it be willing to spend more
for one now?
In this article, we discuss one
factor that may contribute to phy-
sicians’ skepticism about using cost-
effectiveness analysis in health re-
source allocation decisions: that the
commonly cited threshold figure of
$50000 to $100000 per QALY is too
low. We then explain that a signifi-
cantly higher threshold is more con-
sistent with societal willingness to
pay for medical interventions. Fi-
nally, we argue that the accepted
threshold is dynamic and must
change over time. Changes in the op-
timal price of a QALY are depen-
dent not only on inflation but also
on complex interactions between so-
cial desires to control health care
costs and the rate of development of
new health care technologies. We
conclude that if society hopes to con-
strain health care cost growth, the
inflation-adjusted amount it should
spend to produce an additional
QALY will likely need to decrease
over time if technological innova-
tion continues at its current pace.
THE ARGUMENT FOR
A COST-EFFECTIVENESS
THRESHOLD
Before determining an appropriate
cost-effectiveness threshold (ie, the
amount of money that should be
spent to produce a QALY), it is
worth asking whether the idea of
having any threshold is worth-
while. The appeal of a precise thresh-
old is easy to understand. The exis-
COMMENTARY
(REPRINTED) ARCH INTERN MED/ VOL 163, JULY 28, 2003 WWW.ARCHINTERNMED.COM
1637
©2003 American Medical Association. All rights reserved.
at UCSF/Library, on June 29, 2006 www.archinternmed.comDownloaded from
tence and acceptance of such a
benchmark would greatly facilitate
decision making by reducing the
bottom line of any cost-effective-
ness study to whether or not the ra-
tio falls above or below the prede-
termined cutoff value.
Determining the appropriate
threshold is a complex task that is
dependent on multiple interrelated
factors. Decision-makers’ willing-
ness to pay for a QALY depends on
the context and perspective of the
decision itself. A single threshold
may not be appropriate for all indi-
viduals. Moreover, broad societal re-
source allocation decisions, such as
how much to spend on health care
vs education vs a tax cut, may yield
one threshold, while narrower de-
cisions, such as how to maximize
health benefits within a fixed bud-
get (eg, Medicaid), may yield an-
other. Given the potential for mul-
tiple contexts, it is unlikely that a
unique, broadly applicable, thresh-
old value per QALY gained exists.
In light of the reality that a
single acceptable threshold will not
emerge, the US Public Health Ser-
vice panel on cost-effectiveness con-
cluded that any threshold should be
recognized as a guide to health care
spending decisions, not as a deter-
minant of such decisions.
9
In order
to guide such decisions, it is useful
to be able to refer to some cost-
effectiveness threshold, so that de-
cision makers can place new cost-
effectiveness analyses into a broader
context. Thus, for the purposes of
this article, we will assume, along
with much of the cost-effectiveness
community, that such a bench-
mark value exists.
EVIDENCE THAT
THE $50000 TO $100000
THRESHOLD IS TOO LOW
If the desire for a cost-effectiveness
threshold exists, the next question
to ask is whether the long-cited
benchmark of $50000 to $100000
per QALY is consistent with soci-
etal preferences. Two types of evi-
dence inform this question. The first
type is qualitative evidence regard-
ing Americans’ level of comfort in
basing resource allocation deci-
sions on the existing thresholds. The
second type is quantitative evi-
dence based on the results of stud-
ies evaluating the statistical value of
life. These studies seek to infer peo-
ple’s willingness to pay for years of
life or QALYs from observations of
actual behavior (eg, wage differen-
tials for riskier occupations) or sur-
veys explicitly incorporating risk/
money tradeoffs.
Qualitatively, many members of
the medical profession are uncom-
fortable with the notion of denying
health care interventions that have
been proved to improve clinical out-
comes, even when economic evalu-
ations find their use not to be cost-
effective relative to conventional
thresholds.
7,10
This discomfort may
reflect a general distaste for thresh-
olds at any level (the “save a life at
any cost” mentality), which says little
about the merits of existing thresh-
olds.
11
Such distaste may arise from
distrust of the motives of the pro-
ponents of cost-effectiveness stud-
ies (eg, insurers or pharmaceutical
manufacturers) or from a belief that
cost-effectiveness methods are not
well defined or consistently ap-
plied. Alternatively, this discom-
fort may not reflect the acceptance
of the concept of a threshold, but a
perception that conventional thresh-
olds are too low. We suspect that all
these factors are contributing to cli-
nicians’ discomfort with cost-
effectiveness analyses.
The common practice of per-
forming annual Pap smear screen-
ing for women at low risk of cervi-
cal cancer, despite analyses that
indicate cost-effectiveness ratios of
greater than $700000 per QALY
gained, probably illustrates discom-
fort with any threshold.
12
How-
ever, public controversy surround-
ing screening mammography for
women aged 40 to 49 years may re-
flect a willingness to accept a thresh-
old, but one greater than the com-
monly cited $100000 per QALY
gained. In that case, a National In-
stitutes of Health Consensus Devel-
opment Panel issued a statement that
for this population current data do
not support universal screening,
which has a cost of approximately
$150000 per QALY gained.
11
The
negative reaction to this statement
included dissent from 2 of the Na-
tional Institutes of Health panel’s 12
members, a 17-1 vote by the Na-
tional Cancer Institute’s Advisory
Board to support screening in this
age group, a 98-0 vote by the US Sen-
ate in favor of a nonbinding resolu-
tion to support screening in this
age group, and considerable media
coverage.
11
This reaction was dis-
missed by some as an inappropri-
ate substitution of special interest
politics for scientific evidence.
13
Al-
though the politics of the particu-
lar disease undoubtedly influenced
the reaction, an unwillingness to
deem $150000 per QALY gained as
an unworthy investment may also
have played a role. Examples from
recent coverage decisions for diag-
nostic tests that improve detection
of disease at increased costs (eg,
colonoscopy for colon cancer and
the addition of digital mammogra-
phy to routine mammography to de-
tect breast cancer) also support the
argument that a substantially higher
threshold is warranted.
One important caveat to this
sort of qualitative evidence con-
cerns the role of the health insurer
in the provision of a medical inter-
vention. Since most of the costs of
medical care are not borne by pa-
tients or providers, insurance cov-
erage creates an incentive to de-
liver and utilize care that would not
provide sufficient expected ben-
efits to justify its costs, if those costs
were actually borne by patients or
providers. In fact, third-party pay-
ment is a large part of the rationale
for applying a cost-effectiveness
threshold in the first place. Just as
we may consider a Mercedes-Benz
to be a cost-ineffective alternative to
a Chevrolet when spending our own
money, how many of us would head
to the Chevrolet dealership if some-
one else offered to buy us any auto-
mobile of our choosing?
Other evidence that the $50000
to $100000 per QALY threshold is
too low is derived from the statisti-
cal value-of-life literature. We previ-
ously reviewed the estimates (gen-
erally, dollars per life) in this
literature and converted them to dol-
lars per QALY.
5
The reviewed stud-
ies included studies of actual behav-
ior (eg, willingness to pay for safety
improvements, the extent to which
wages in riskier occupations exceed
those in lower risk but otherwise
similar occupations) as well as con-
(REPRINTED) ARCH INTERN MED/ VOL 163, JULY 28, 2003 WWW.ARCHINTERNMED.COM
1638
©2003 American Medical Association. All rights reserved.
at UCSF/Library, on June 29, 2006 www.archinternmed.comDownloaded from
tingent valuation studies (surveys
about hypothetical tradeoffs be-
tween money and risk). Although this
literature has considerable variation
in methods and study populations
and consequently yields variable es-
timates of dollars per QALY gained,
the clear impression is that the value
per QALY gained implied by the
statistical value-of-life literature
exceeds conventional cost-effective-
ness thresholds. The median behav-
ioral or contingent valuation study in
our review implied a value per incre-
mental QALY of $265000 in 1997 US
dollars. Furthermore, 28 of the 35
value-of-life estimates we reviewed
implied a value per QALY in excess
of $100 000.
While we do not wish to over-
state our confidence in any specific
threshold being correct, both the
qualitative evidence and the empiri-
cal data from the statistical value-of-
life literature suggest that the $50000
and even $100000 thresholds are
likely conservative estimates of the
value of a QALY. Furthermore, it does
not appear from multiple clinical ex-
amples that the frequently cited
$50000 to $100000 thresholds are
generally enforced in practice.
THE THRESHOLD ONCE
CHOSEN MUST BE DYNAMIC
It should not surprise anyone that
the $50000 to $100000 threshold is
too low. The original $50000 thresh-
old was based on the reasoning that
if dialysis cost $50 000 per QALY,
then anything else that cost $50000
or less per QALY should be made
available to patients. However, by
this reasoning, the $50000 cutoff
should have been the floor of the
threshold, not the ceiling. If offering
dialysis to patients compelled soci-
ety to offer other interventions that
cost $50000 per QALY, who is to say
that society should not have been
compelled 2 decades ago to spend
more than $50000 for a QALY?
For the purposes of argument,
let us suppose that society agrees to
apply $265000 per QALY as the
benchmark for judging the cost-
effectiveness of medical interven-
tions. Coming to this decision leads
to 2 questions. First, can the United
States really afford to call (and pro-
vide) care costing nearly 10 times its
per capita income ($29676 in 2000)
cost-effective? Second, with enforce-
ment of a such a cutoff, will health
care costs be affordable over time?
The key to answering the first
question lies in the distinction
between the average cost and the
marginal cost of a QALY. The cost-
effectiveness threshold selected rep-
resents the cutoff for the accept-
able marginal cost of a QALY (the
most society would be willing to pay
to produce 1 additional QALY), while
the average cost of health care rep-
resents the amount of non–health
care consumption forgone in order
to receive health care on average.
Looked at in another way, 50% of the
population incurred less than 5% of
aggregate health care expendi-
tures. It is this relatively low aver-
age cost of health care that allows us
to potentially “afford” what may
seem to be an exorbitant “marginal
cost” of that last additional QALY.
That is, we would spend close to
$265000 per QALY for very few in-
dividuals in any given year.
Ultimately, the issue of afford-
ability depends on how adoption
(and enforcement) of a higher thresh-
old value for defining cost-effective-
ness would affect current health care
spending and alter spending over
time. Because currently cited stan-
dards are often ignored, one cannot
be certain how a more generous
benchmark would influence patient
and physician behavior.
IMPACT OF THRESHOLD
ON LONG-TERM COSTS
For the sake of illustration, sup-
pose the cost per QALY of all inter-
ventions for all potential users was
calculated, and the United States
adopted and enforced a standard of
$265000 per QALY gained. Rela-
tive to the current situation, where
the $50000 to $100000 standards
are not enforced the resultant short-
run change in aggregate health ex-
penditures that resulted from adopt-
ing the new standard could be either
positive or negative. Total health care
costs might be decreased through re-
duction in the use of services that
cost more than $265000 to pro-
duce a QALY (eg, the routine Pap
smear test). However, if the $100000
threshold was having a stronger in-
fluence on clinical behavior than was
previously suspected, the relax-
ation of the standard could encour-
age the use of additional services that
cost less than $265000 but more
than $100000 per QALY. This in-
creased use would affect the bot-
tom line of third-party payers. Of
course, if thresholds at any level
are irrelevant to practice patterns,
changing the standard would have
no effect on spending.
What about the long run? What
would happen to health care expen-
ditures if decision makers consis-
tently followed the $265000 per
QALY cutoff (or any cutoff for that
matter)? Our best guess is that health
care expenditures would continue to
increase faster than the general in-
flation rate. To illustrate: imagine
that a new treatment is developed for
a disease. Offering the treatment to
a population with this disease costs
$200000 and produces 3 addi-
tional QALYs in that population, at
an incremental cost-effectiveness ra-
tio of $67000 per QALY. Because
this treatment passes the bench-
mark, it would be offered to the
population. Now imagine that an-
other treatment is developed for this
disease that, when used in addition
to the first treatment, increases
health care costs another $200 000
but produces another QALY. It too
would be offered. Theoretically, a
large number of such incremen-
tally beneficial services could be de-
veloped, each raising per capita
health care costs but producing
enough QALYs to fall beneath the
$265000 threshold.
Furthermore, it is important to
remember that new interventions
can cost less than previous interven-
tions and still raise total health
care costs, because they increase
health care use, the “behavioral
response.”
14
For example, on a case-
by-case basis, laparoscopic chole-
cystectomy is a cost-saving alterna-
tive to traditional cholecystectomy:
the extra costs of laparoscopy are
more than made up for by de-
creased hospital stays.
15
However,
these per-case savings can be over-
whelmed by the increased number
of additional patients who are now
deemed to be suitable for cholecys-
tectomy (eg, those thought either
to be at too high of a risk for tra-
(REPRINTED) ARCH INTERN MED/ VOL 163, JULY 28, 2003 WWW.ARCHINTERNMED.COM
1639
©2003 American Medical Association. All rights reserved.
at UCSF/Library, on June 29, 2006 www.archinternmed.comDownloaded from
ditional cholecystectomy or to have
symptoms too mild to warrant open
surgery). This example illustrates
that maintaining any arbitrarily de-
fined cost-effectiveness threshold
over time is unlikely to constrain
health care inflation in a substan-
tial way. New interventions are de-
veloped every day that produce ad-
ditional QALYs for patients for less
than the current $50 000 to $100000
threshold. Of course, even more new
interventions would fall under the
proposed, higher threshold. Only a
rare minority of new interventions
lead to net cost savings from the bud-
getary perspective. Most new phar-
maceutical products that come to
market have a cost-effectiveness that
is substantially better than even the
current $50000 to $100000 thresh-
old for a QALY.
6
Offering these
medications to patients will ulti-
mately drive up health care costs.
HOW SHOULD A
COST-EFFECTIVENESS
THRESHOLD BE USED?
In theory, a societal cost-effective-
ness threshold can be determined if
(1) all health care interventions can
be ranked from most cost-effective
to least cost-effective and (2) a fixed
health care budget can be deter-
mined. The budget would be spent,
in order, from the most cost-
effective intervention to the least
cost-effective. Budget planners
would estimate the cost of offering
each intervention to the popula-
tion in question. Moving down the
list, budget planners would esti-
mate how many interventions could
be offered before the budget was ex-
hausted. At that point on the list, a
line would be drawn such that the
more cost-effective interventions,
above the line, would be funded
and the less cost-effective interven-
tions, below the line, would not. The
cost-effectiveness of the last inter-
vention funded would effectively act
as the cost-effectiveness threshold.
As Oregon discovered in its
Medicaid rationing plan, this ap-
proach, while theoretically appeal-
ing, is practically flawed.
16,17
To
begin with, there is no societal con-
sensus about how much money to
spend on health care. Thus, it is dif-
ficult to know where on the list to
draw the line between funded and
unfunded interventions. More im-
portantly, good cost-effectiveness
data are lacking for many health care
interventions, and the cost effective-
ness depends not only on the inter-
vention but also on the characteris-
tics of the patients to whom it is
applied. Thus, it is impossible to gen-
erate a thorough cost-effectiveness
list: hence the appeal of an informal
cost-effectiveness threshold (a num-
ber of dollars per QALY that can be
used to guide health care decision
making). Formulary committees can
appeal to such a cost-effectiveness
threshold when deciding the merits
of new medications. Physicians can
look to the threshold when making
diagnostic and treatment decisions.
Managed care medical directors, in-
surance company executives, and
government administrators can use
the threshold to inform their treat-
ment coverage decisions. In none of
these cases should cost-effective-
ness data, or the admittedly arbi-
trary cost-effectiveness threshold, de-
termine any final decision. Instead,
cost-effectiveness data need to be in-
corporated with other information
before important decisions can be
made.
Our concern is that people’s
willingness to use cost-effectiveness
information is limited, in part, by the
implausibility of the current $50 000
to $100000 per QALY threshold.
The threshold is counterintuitively
low. And anyone who has been read-
ing cost-effectiveness literature over
the years recognizes that the thresh-
old has not been reevaluated for
decades. Given the implausibility of
the threshold, decision makers
may minimize their use of any cost-
effectiveness data. Revising the
threshold now to be greater than
$100000 per QALY, and revisiting
the threshold over time, may in-
crease the credibility of cost-
effectiveness analyses in guiding
health care decisions.
CONCLUSIONS
Cost-effectiveness information de-
serves a role in health care decision
making. Efforts spurred on by the US
Public Health Service panel on cost-
effectiveness to standardize and raise
the quality of cost-effectiveness
methods are intended to enhance the
credibility of the cost-effectiveness
literature and to improve compara-
bility across studies. Nevertheless,
without some context about when
a cost-effectiveness threshold is
acceptable, decision makers will
have difficulty making use of cost-
effectiveness information. We pro-
pose that currently cited thresh-
olds are too low, especially the
commonly cited $50000 per QALY
gained cutoff. To support this
point, we note several commonly
used clinical examples with cost-
effectiveness ratios that exceed con-
ventional thresholds. In line with
these intuitions, we think that it is
more appropriate to raise the cost-
effectiveness threshold signifi-
cantly, approaching $200000 or
more per QALY, after appropriate in-
put from all vested stakeholders. Set-
ting a new threshold is best accom-
plished by consensus process, such
as that used by the US Public Health
Service when it established stan-
dards for the field of cost-effective-
ness in the mid 1990s. We think that
the next time such a consensus pro-
cess is convened, experts should se-
riously consider revising the cost-
effectiveness threshold.
Maybe more important than the
setting of the threshold, we believe
that the cost-effectiveness thresh-
old should be refined regularly to al-
low for expected changes in cost of
goods (inflation), per capita in-
come, burden of disease, innova-
tion in diagnosis and treatment, and
patient preferences. For cost-
effectiveness analyses to be cred-
ible allocation tools for key deci-
sion makers, they need to reflect
changing circumstances. While in-
flation and budgets are likely to in-
crease the threshold over time, we
believe that continued innovation-
producing interventions that meet
the benchmark (and resultant de-
mand for them) will require that the
cost-effectiveness threshold actu-
ally be reduced over time.
We acknowledge that there is
no simple way to determine the ap-
propriate price of a QALY and that
whatever the price of a QALY should
be now, it likely needs to change
over time. At the same time, we need
to recognize that health care inter-
ventions that seem to be a bargain
(REPRINTED) ARCH INTERN MED/ VOL 163, JULY 28, 2003 WWW.ARCHINTERNMED.COM
1640
©2003 American Medical Association. All rights reserved.
at UCSF/Library, on June 29, 2006 www.archinternmed.comDownloaded from
may not always be affordable from
a population perspective owing to
the behavioral response. In the long
run, as more and more health care
interventions continue to be devel-
oped and used, what appears to be
an affordable QALY today may not
be an affordable QALY tomorrow.
Much in the way that a family can
bankrupt itself if it purchases too
many goods, even when those goods
are available at bargain prices, those
responsible for health care spend-
ing must understand the aggregate
budgetary impact of providing re-
imbursement for interventions that
produce one extra QALY for what
appears to be an affordable price. For
now, the price of a QALY should be
significantly greater than $50 000 to
$100000. But the price needs to be
revisited on a regular basis in order
to forge some consensus about what
society can continue to afford.
Peter A. Ubel, MD
University of Michigan
Health Systems
300 N Ingalls, Room 7C27
Ann Arbor, MI 48109-0429
(e-mail: paubel@umich.edu)
Richard A. Hirth, PhD
Michael E. Chernew, PhD
A. Mark Fendrick, MD
Ann Arbor
Dr Ubel is a recipient of a career de-
velopment award in health services re-
search from the Department of Veter-
ans Affairs and of a Presidential Early
Career Award for Scientists and En-
gineers (PECASE). This work was also
supported by grants RO1 HD40789-
01, RO1 HD38963-02, and R01-
CA78052-01 from the National Insti-
tutes of Health, Bethesda, Md.
The authors have no relevant fi-
nancial interest in this article.
We gratefully acknowledge Heidi
Rininger for help with manuscript
preparation and Allan Detsky, PhD,
and Rodrigo Cavalcanti, PhD, for com-
ments on an earlier draft.
REFERENCES
1. Vijan S, Hofer TP, Hayward RA. Cost-utility analy-
sis of screening intervals for diabetic retinopathy
in patients with type 2 diabetes mellitus. JAMA.
2000;283:889-896.
2. Newby KL, Eisenstein EL, Califf RM, et al. Cost ef-
fectiveness of early discharge after uncompli-
cated acute myocardial infarction. N Engl J Med.
2000;342:749-755.
3. Smith KJ, Roberts MS. The cost-effectiveness
of sildenafil. Ann Intern Med. 2000;132:933-
938.
4. Lerner M. Health insurers balk at paying $10 apiece
for Viagra pills: Minnesota’s three largest health
plans say they need more time and details before
they decide on whether—or to what extent—to
pay. Star Tribune. June 3, 1998:1A.
5. Hirth RA, Chernew ME, Miller E, Fendrick AM,
Weissert WG. Willingness to pay for a quality- ad-
justed life year: in search of a standard. Med De-
cis Making. 2000;20:332-342.
6. Neumann PJ, Sandberg EA, Bell CM, Stone PW,
Chapman RH. Are pharmaceuticals cost-
effective? a review of the evidence. Health Aff
(Millwood). 2000;19:92-109.
7. Ubel PA, Baron J, Nash B, Asch DA. Are prefer-
ences for equity over efficiency in health care al-
location “all or nothing”? Med Care. 2000;38:366-
373.
8. Ubel PA, Jepson C, Baron J, Hershey JC, Asch DA.
The influence of cost-effectiveness information on
physicians’ cancer screening recommendations.
Soc Sci Med. 2003;56:1727-1736.
9. Gold MR, Siegel J, Russell LB, Weinstein M, eds.
Cost-Effectiveness in Health and Medicine. New
York, NY: Oxford University Press; 1996.
10.
Hershey J, Asch DA, Jepson C, Baron J, Ubel PA.
Incremental and average cost-effectiveness ra-
tios: will physicians make a distinction? Risk Anal.
2003;23:81-89.
11.
Ritov II, Baron J. Protected values and omission
bias. Organ Behav Hum Decis Process. 1999;79:
79-94.
12.
Eddy DM. Screening for cervical cancer. Ann In-
tern Med. 1990;113:214-226.
13.
Fletcher S. Whither scientific deliberation in health
policy recommendations? Alice in the Wonder-
land of breast-cancer screening. N Engl J Med.
1997;336:1180-1183.
14.
Fendrick AM, Chernew ME, Hirth RA, Menon D.
Understanding the behavioral response to medi-
cal innovation. Am J Manag Care. 1996;2:793-
799.
15.
Chernew ME, Fendrick AM, Hirth RA. Managed
care and medical technology: implications for
health care cost growth. Health Aff (Millwood).
1997;16:196-206.
16.
Eddy DM. Oregon’s methods: did cost-
effectiveness fail? JAMA. 1991;266:2135-2141.
17.
Hadorn DC. Setting health care priorities in Or-
egon: cost-effectiveness meets the rule of res-
cue. JAMA. 1991;265:2218-2225.
(REPRINTED) ARCH INTERN MED/ VOL 163, JULY 28, 2003 WWW.ARCHINTERNMED.COM
1641
©2003 American Medical Association. All rights reserved.
at UCSF/Library, on June 29, 2006 www.archinternmed.comDownloaded from
