TOWARDS COST-VALUE ANALYSIS IN HEALTH CARE?
Health Care Analysis 1999,7,167-175
Erik Nord, National Institute of Public Health, P.O. Box 4404
Torshov, N-0403 Oslo, Norway.
phone: (47) 22 04 23 42, fax: (47) 22 04 25 95, e-mail: erik.nord@folkehelsa.no
Key words: health economics, QALY, equity, fairness, cost-value
analysis, SAVE
Abstract
By describing societal value judgements in health care in numerical
terms one may in theory increase the precision of guidelines for priority
setting and allow decision makers to judge more accurately the degree to which
different health care programs provide societal value for money. However,
valuing health programs in terms of QALYs disregards salient societal concerns
for fairness in resource allocation. A different kind of numerical valuation of
medical interventions, that incorporates concerns for fairness, is described. The
usefulness to decision makers of such numerical information remains to be tested
(see also paper on transforming utilities (click).
Introduction
When
we spend money, we want to get good value for it. This is true when we do
private shopping . It is equally true when we are asked as citizens how we
think scarce resources in a national health service should be distributed
across patient groups.
Judging
value in public health care is difficult, and better ways of doing so are being
explored continuously. The purpose of this article is to present a recent
development in numerical modelling of societal valuations of medical
interventions, in which salient concerns for fairness across patient groups are
taken into account in addition to concerns for efficiency. Such modelling opens
up for a more comprehensive and valid type of economic evaluation – called
cost-value analysis - than one has seen hitherto in health economics. Hopefully
this development will increase the relevance of economic analysis in informing
resource allocation decisions in health care.
Guidelines for priority setting
In
national health services, some elected or selected individuals or groups of
individuals are given the rather ungrateful task of judging the value of
different health care programs relative to their costs and of prioritising
between the programs in accordance with these judgements in budget decisions. To
make their task easier, and ultimately to improve the performance of these
analysts and decision makers, many countries have developed guidelines for
priority setting in health care. The guidelines are based partly on ethical
reflection in academics and policy makers (e.g. Daniels, 1985; Menzel, 1990;
The Norwegian Commission on Priorities in Health Care, 1987), partly on
measurements of values, attitudes and preferences in samples of the general
population (e.g. Charny et al, 1989; The Oregon Health Services Commission,
1990; Campbell and Gillett, 1993; Olsen, 1994; Nord, Richardson et al, 1995;
Ubel et al, 1996; Pinto Prades, 1997; Dolan and Cookson, 1998).
A
review of existing materials of the above kinds in industrialised countries
like Australia, England, Holland, New Zealand, Norway, Spain, Sweden and the US
(Nord, 1996) suggests that ethicists’ and policy makers’ reflections, and
results from public preference measurements, converge on the following points:
A.
Society demands that medical interventions satisfy a minimum
requirement
of effectiveness in terms of value to the patients concerned.
B.
Society’s appreciation (valuation) of medical interventions increases strongly
with increasing severity of the patient’s condition.
C.
Life saving or life extending procedures are particularly highly valued, and
significantly more highly than interventions even for patients with severe
chronic conditions.
D.
When the minimum requirement of effectiveness is satisfied (point A above),
society worries less about differences in the size of the health benefits
provided by treatment programs for different patient groups, the underlying
attitude being that people are entitled to realising their potential for
health, whether that be large or moderate given the state of art in different
areas of medicine.
E.
As a special case of point D, society in most cases does not wish to
discriminate between people with different potentials for health in decisions
about life saving or life extension. For instance, society regards the
prevention of premature death in people with chronic disease as equally worthy
of funding as the prevention of premature death in otherwise healthy people. (Life
extending interventions for people in vegetative states or states of very low
subjectively perceived quality of life is a different matter.)
Points
A-E constitute a set of guidelines for resource allocation in a national health
service. Presumably decisions makers are able to do a better job when such
guidelines are available than when they are completely in the dark regarding
the criteria that society at large would like to see applied in resource
allocation decisions.
Verbal
guidelines are, however, imprecise. Consider for instance two states of illness
X and Y, where X is more severe than Y. Guideline B says that a treatment for X
is to be regarded as more valuable than a treatment for Y (assuming equal
effectiveness). But it does not say how much more valuable X is. Is it for
instance so much more valuable that it would justify that a program for illness
X were given priority over a program for illness Y even if, for the same amount
of money, only half as many people could be treated in program X compared to
program Y?
Similarly,
guideline C gives decision makers only a very rough idea about the trade-offs
that society would like to make between life saving and life improving
programs. Guideline A says little about what is a minimum requirement of
effectiveness, and, particularly given this unclarity, guideline D does not
make it clear how little differences in effectiveness should count.
Cost-utility analysis
In
theory it is possible to supplement verbal guidelines with numbers that may
indicate more precisely to decision makers the trade-offs that society at large
would want to make between health care programs that affect (a) different
numbers of people at (b) different levels of severity of illness and with (c)
different potentials for improvement in health. To provide such numbers as a
potential aid to decision making has been the ambition of health economists for
almost three decades (Culyer et al, 1971).
Until
recently the numbers provided by health economists were based on the assumption
that it is rational for a society to aim at maximising the sum of individual
health benefits produced by its health care system within given budget
constraints. Health benefits are calculated as a product of health related
gains in quality of life and the number of years that patients get to enjoy
health benefits. Health related quality of life has been measured in terms of
’utility’ on a scale from zero (dead) to unity (healthy), and the product of
quality of life gains and years of benefit has been expressed in terms of
gained Quality Adjusted Life Years (QALYs), see for instance Torrance (1986). The
recommendation of health economists has been for decision makers to allocate
scarce resources in health care such as to maximise the number of QALYs gained.
This would be achieved by giving priority to medical interventions that have a
relatively low cost per QALY gained, or in other words a favourable
’cost-utility ratio’.
It
is increasingly clear that the basic premiss of this recommendation - that the
goal of a national health service is to maximise health benefits - is
incorrect. There is strong evidence that the societal structure of concern in
health care is more like in points A-E above, in which the idea of health
benefit maximisation has no place. A numerical supplement to verbal guidelines
in health care priority setting must therefore be quite different from the
numbers that hitherto have been on offer in health economics.
Cost-value analysis
A
different set of numbers, that purports to be consistent with points A-E above,
was suggested recently by a team of researchers from
The
numbers in the table are based on a series of empirical studies in later years
of the trade-offs that the general public would want to make between health
care programs that affect (a) different numbers of people at (b) different
levels of severity of illness and with (c) different potentials for improvement
in health. I emphasise that the method of synthesis was informal, the ambition
being only to indicate somewhat roughly what seems to be a widespread societal
structure of concern.
The
concerns for severity and life saving expressed in points B and C in the verbal
guidelines above come through in the upper diagonal in the table: One step up
on the scale is valued more highly (and much more so) the lower the start
point. The concerns for effectiveness and realisation of potential expressed in
points A and D come through in each horisontal line: A movement from any given
start point scores better the higher the end point, but marginal value
decreases significantly with increasing treatment effect. For instance, a
person with a potential to go from level 7 to level 4 will score almost as much
as a person with a potential to go all the way from level 7 to level 1. The
concern for non-discrimination in matters of life saving or life extension is
expressed in the bottom line of the table , according to which the avoidance of
death scores 1 no matter what the resulting state is, although the bottom left
hand cell is left void to indicate that at some level of severity the value of
life extension will be questioned.
The
first seven lines plus the lower right hand cell of table 1 correspond to a set
of health state values as shown in table 2. If we accept that the eight-point
scale approximates an equal-interval one in terms of individual utility, the
table shows decreasing marginal societal value of utility gains. The numbers
correspond to a curve for societal health state values as a function of
individual utilities that is convex to the y-axis.
Table
2 is not applicable to interventions that extend life at levels below full
health, as these in the main should be assigned full value, see above.
The
numbers indicated above could in principle be multiplied by the number of years
that people get to enjoy health improvements, as is done in QALY calculations. However,
this is not recommendable. Multiplying with life years presupposes that
societal value is proportional to the size of the health benefit. We know that
it isn’t, cfr. point D in the guidelines above. For instance, Dolan and Cookson
(1998) report that subjects would not discriminate between two programs that –
all else equal - gave two different groups of patients 8 and 20 years of
benefits respectively. The exact relationship between duration of benefits and
societal value is not yet known. At the moment, use of numbers of the kind
indicated above should therefore be restricted to comparisons of programs that are
fairly equal in terms of duration of benefits, or comparisons of programs in
which benefits are so durable that differences in duration between the programs
become a minor concern.
Tables
1 and 2 refer to health problems in terms of reduced mobility. This is because
so much of the existing societal preference data pertain to this particular
dimension. To apply the numbers to other kinds of health problems, one needs to
know where they belong on the severity scale of table 1. This may be judged by
judging the effect on quality of life of those other problems compared to the
effects on quality of life of the various mobility problems indicated in the
table. Studies of quality of life in patients with disabilities and chronic
illnesses, combined with the experience and expertise of health professionals,
may facilitate such judgements.
Note
finally that the approach described above purports to encapsulate concerns for
several different aspects of response to health care - in this case initial
severity, potential for health and the actual health gain - in one single set
of numbers. Some may prefer to make the nature and the extent of the
severity-effectiveness trade-off explicit by adopting a decomposed approach, in
which separate equity weights are introduced for distributive concerns, see for
instance Williams (1988;1997) and Dolan (1998). Interested readers are referred
to Nord et al (1999) for a further exploration into this alternative. While
technically different, the objective is the same: To obtain estimates of value
that incorporate concerns for fairness and thus allow a more comprehensive and
valid cost-value analysis of health care.
Example
An
example of cost-value analysis is a follows: Assume that intervention A takes
one type of patient from level 6 to level 4 at a cost of 10,000 USD, while
intervention B takes another type of patient from level 4 to level 1 at a cost
of 5,000 USD. In terms of individual utility gain, B is more valuable (more
efficient) than A (given that the steps on the severity scale are roughly
equidistant in terms of individual utility). But the societal value of
intervention A is 0.27, compared to 0.08 for intervention B. The societal value
per 10,000 USD spent is 0.27 and 0.16 for A and B respectively, indicating that
allocating resources to area A rather than B gives more value for money when
not only concerns for efficiency, but also concerns for equity are taken into
account. This may be useful information to decision makers.
Rationale and validity
Generally
speaking, the rationale for describing societal value judgements in numerical
terms is to increase the precision of guidelines for priority setting and to
allow decision makers to judge more accurately the degree to which different
health care programs provide societal value for money.
The
kind of numbers indicated in this paper purport to do this job better than the
utilities of conventional health economics. The reason is simply that societal
value judgements are governed quite strongly by concerns for equity (in addition
to concerns for the sum of individual utility gains), and that the most salient
equity concerns have turned out to be amenable to quantification at the same
level of measurement as individual utilities are.
However,
even if societal value numbers conceptually are more relevant than individual
utility scores in informing resource allocation decisions, the appropriate way
of establishing such numbers remains to be determined.
One
common criticism of societal value numbers is that people’s responses to numerical
preference questions in mailed questionnaires are unreflective and unreliable. This
is true. But there is nothing that prevents researchers from collecting
preference data in more high quality ways, for instance in focus groups that
discuss ethical issues carefully before each participant gives his or her
responses to specific quantitative questions. This is already happening (Nord,
1995; Murray and Lopez, 1996; Dolan, 1998), and may very well become the
default approach in the future.
Another
issue is: From whom should societal preferences for resource allocation be
elicited? Should it for instance be from the general public, or perhaps from a
selection of highly reflective members of society?
This
is really a political question, to which there is no simple value-free answer. In
principle, analysts and health planners might be interested in having access to
value tables representing a number of societal subgroups. The important thing
to bear in mind is that any value table primarily expresses the values ot the
specific group of people whose preferences where elicited in the first place. Potential
users of the table will have to decide for themselves what status they wish to
assign to the thoughts and values of that particular group.
Having
said this, there is much to be said for the following approach: Judgements as
to where on a scale of severity (as in table 1) different kinds of health
problems belong are judgements of health related quality of life. In such
judgements, people with personal experience with the problems in question can
claim expertise, and they are therefore arguably the right people to ask. On
the other hand, judgements regarding the relative societal values of different
improvements on the severity scale are judgements of distributive fairness. In
such judgements, it is not so clear that some are so much greater experts than
others that judgements should be delegated to them. On this issue, the general
public seems a more appropriate source of values (Nord et al, 1999).
Concluding remarks
In
this paper I have claimed that there is a theoretical rationale for modelling
society’s valuation of medical interventions in numerical terms, and that this
modelling can be done in a meaningful and reasonably valid way.
It
does not follow that such modelling in practice really will be felt as helpful
by analysts, health planners and policy makers. Numerical approaches using
indexes of value are inherently reductionist. They are also linguistically
alienating to many people (not everybody likes numbers).
It
would not be fair to judge the usefulness of numerical approaches on the basis
of experiences with the specific numerical models that have been on offer
hitherto in health economics, since these have been clearly off target due to
their focus on health benefit maximisation. It is a task for empirical research
to study whether an improved model as described in this paper will prove more
helpful.
REFERENCES
Campbell
A, Gillett G. Justice and the right to health care. In Ethical Issues in
Defining Core Services. Wellington: The National Advisory Committee on Core
Health and Disability Support Services, 1993.
Charny
MC, Lewis PA, Farrow SC. Choosing who shall not be treated in the NHS. Social
Science & Medicine, 1989;28:1331-8.
Culyer
AJ, Lavers RJ, Williams A. Social indicators: Health. Social Trends
1971,2,31-42.
Daniels
N. Just Health Care. Cambridge, MA: Harvard University Press, 1985.
Dolan
P. The measurement of individual utility and social welfare. Journal of Health
Economics 1998, 17, 39-52.
Dolan
P, Cookson R. Measuring preferences over the distribution of health benefits. Mimeo.
University of York: Centre for Health Economics. 1998.
Menzel
P. Strong Medicine. New York: Oxford University Press 1990.
Murray
C, Lopez A. The Global Burden of Disease. Harvard University Press 1996.
Nord
E. The trade-off between severity of illness and treatment effect in cost-value
analysis of health care. Health Policy,1993,24,227-238.
Nord
E. The person trade-off approach to valuing health care programs. Medical
Decision Making, 1995,15,201-208.
Nord
E. Health status index models for use in resource allocation decisions. A
critical review in the light of observed preferences for social choice. International
Journal of Technology Assessment in Health Care 1996, 12,31-44.
Nord
E, Richardson J, Street A, Singer P, Kuhse H. Maximising health benefits versus
egalitarianism: An Australian survey of health issues. Social Science &
Medicine,1995,41,1429-1437.
Nord
E, Pinto JL, Richardson J, Menzel P, Ubel P. Incorporating societal concerns
for fairness in numerical valuations of health programs. Health Economics
1999,8,25-39.
Norwegian
Commission for Prioritising in Health Care. Retningslinjer for prioritering
innen helsevesenet. (Guidelines for prioritising in health care.) NOU 1987:23.
Oslo: Universitetsforlaget, 1987.
Olsen
JA. Persons vs years: Two ways of eliciting implicit weights. Health Economics
1994,3,39-46.
Oregon
Health Services Commission. Prioritization of health services. A report to the
Governor and Legislature. Salem, Oregon 1991.
Pinto
Prades, José-Luis. Is the person trade-off a valid method for allocating health
care resources? Health Economics 1997,6, 71-81.
Sintonen
H. An approach to measuring and valuing health states. Soc Sci & Med
1981;15c:55-65.
Torrance
GW. Measurement of health state utilities for economic appraisal. Journal of
Health Economics 1986, 5, 1-30.
Ubel
PA, Kamlet M, Scanlon D, Loewenstein G. Individual utilities are inconsistent
with rationing choices: A partial explanation of why Oregon’s
cost-effectiveness list failed. Medica Decision Making 1996,16,108-119.
Williams
A. Ethics and efficiency in the provision of health care. In Bell JM, Mendus S
(eds). Philosophy and Medical Welfare. Cambridge University Press 1988.
Williams
A. Intergenerational equity: An exploration of the ‘fair innings’ argument. Health
Economics 1997,6,117-132.
Table 1. Societal values for
health improvements.
To level:
|
From level |
7 |
6 |
5 |
4 |
3 |
2 |
1 |
|
1. None |
|
|
|
|
|
|
|
|
2. Slight problem |
|
|
|
|
|
|
0.001 |
|
3. Moderate |
|
|
|
|
|
0.009 |
0.01 |
|
4. Considerable |
|
|
|
|
0.07 |
0.079 |
0.08 |
|
5. Severe |
|
|
|
0.12 |
0.19 |
0.199 |
0.20 |
|
6. Very severe |
|
|
0.15 |
0.27 |
0.34 |
0.349 |
0.35 |
|
7. Completely disabled |
|
0.25 |
0.40 |
0.52 |
0.59 |
0.599 |
0.60 |
|
8. Dead |
|
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
1.00 |
Examples at levels 2-7:
2. Can move about anywhere,
but has difficulties with walking more than 2 kms.
3. Can move about without
difficulty at home, but has difficulties in stairs and outdoors.
4. Moves about without
difficulty at home. Needs assistance in stairs and outdoors.
5. Can sit. Needs help to
move about – both at home and outdoors.
6. To some degree
bedridden. Can sit in a chair part of the day if helped up by others.
7. Permanently bedridden.
Table 2. Health state
values encapsulating concerns for severity and realisation of potential.
|
Problem level |
Value |
|
1. Healthy |
1.00 |
|
2. Slight problem |
0.999 |
|
3. Moderate |
0.99 |
|
4. Considerable |
0.92 |
|
5. Severe |
0.80 |
|
6. Very severe |
0.65 |
|
7. Completely disabled |
0.40 |
|
8. Dead |
0.00 |