continued from page 3...

POLICY ANALYSIS
cost-utility analysis in the same way one
considers a change in blood pressure,
A1C or LDLs as an effectiveness measure
Patient Preference Methods - A Patient
in a cost-effectiveness analysis. The con-
nection of a benefit in QALY to the long
Centered Evaluation Paradigm
term benefit in disease outcomes has John FP Bridges PhD, Assistant Professor, Johns Hopkins Bloomberg School of Public
largely not been made for most US payers. Health, Baltimore, MD, USA; Ebere Onukwugha PhD, Assistant Professor, University of
There are pockets of change from the Maryland, School of Pharmacy, Baltimore, MD, USA; F. Reed Johnson PhD, Senior Fellow,
perspective of the employer, where a con-
RTI Health Solutions, RTP, NC, USA; A. Brett Hauber PhD, Senior Economist, RTI Health
nection between the QALY and employee
productivity is clearer, and perhaps in the
Solutions, RTP, NC, USA
integrated health care system where all
he increased focus on patient reported out- • can be used to evaluate interventions that cur-
costs are accounted for. This would
appear to be highly relevant to a single T comes (PROs) in medical research in recent
years has heightened the awareness of the patient's
rently do not currently exist (i.e. hypothetical
treatments) allowing evaluation to commence
payer such as Medicare; however the drug
benefit for the elderly is delivered through perspective of health outcomes. It is unclear, how- before a program/intervention is implemented or
the fragmented private payer. ever, if the use of PROs alone will lead us toward even designed/developed;
patient-centered care or if we need to rethink some
In conclusion, I applaud the initiative to of the foundations of our research methods in order • address issues of patient choice, and hence can
bring researchers, decision makers, and to better understand the patient's point of view. The be used to understand diseases like obesity, dia-
observers together to contribute to an aim of this paper is to draw awareness to an alter- betes, and coronary-artery disease where long
ongoing dialogue of how to resolve current native class of methods that are rooted in econom- term prognosis depends directly on patient
barriers to the acceptance of the QALY as ic theory known as patient preference methods that, lifestyle choice [4-5];
a measure of patient benefit, the assign- although being used in many other sectors of the
ment of a cost to that benefit, and the economy (both public and private alike), have been • focus on issues of patient adherence [6]; and
application of this evidence to improved somewhat neglected in health care. While patient
decisions in the future. preference methods are related to PROs, health • focus on process-related aspects of health care
related quality of life and the expected-utility meth- in addition to health outcomes [1, 7].
For further details on this ISPOR invitation- ods used to motivate quality adjusted life years
al (funded by a grant from US Agency for (QALYs), they are also a distinct class of methods Patient preference and HR-QoL
Health Care Research and Quality) work- that are increasingly being applied in medicine [1]. While both patient preference methods and the tradi-
shop, see: http://www.ispor.org/meetings/ Despite the growing popularity of preference meth- tional patient reported HR-QoL methods aim to make
MeetingsInvited.aspx. ods such as conjoint analysis and willingness to outcomes research more patient-centered, it is
pay (WTP) in outcomes research, there remains important to distinguish between the two methods.
For QALY issues, see: https://www.ispor.org/ some confusion about them among outcomes
news/articles/Oct07/WTQ.asp. IC researchers, not only in terms of best practice, but • Patient-reported HRQoL methods are concerned
in terms of how they differ from the current evalua- with measuring the patient's status along several
tion paradigm in our field. This paper attempts to aggregate domains (for example, mental, physi-
address the second of these questions by highlight- cal, functional, social, and emotional domains).
WEB CONNECTIONS ing why patient preference methods constitute a Outcomes in a given domain are focused on
more patient-centered, theory-based, and flexible attempting to identify a given state of being (albeit
Not sure exactly what it is that you need group of methods than the current status quo. with bounds set a priori by the researcher) and
but you know it has to do with Outcomes are less focused on attempts to identify or value
Research? Benefits of Patient Preference how the respondent feels about being in that state
Visit www.HealthEconomics.Com, the Method of health; while
resource for outcomes research profession- While it has been well documented in the health
als. The website was founded in 1994 and • Patient preferences are concerned with measur-
economics literature that patient preference meth-
continues to evolve based on the outcomes
ods are far more grounded in economic theory [1- ing the patient's value for a specific component,
research world. It is designed to be that one
3] and are far more patient centered [4] than the or attribute, either in absolute terms or in relation
resource for everyone within our discipline.
current health related quality of life (HR-QoL) and to another attribute. The relative importance is
Not only does this website have forums for
questions/answers but also listings of con- QALY methods used in outcomes research, it is identified by choices that inevitably require trad-
sulting companies, employment opportuni- really the flexibility of the methods that should cap- ing off one or more desirable outcomes (includ-
ties and of course available databases. ture the interest of most outcomes researchers. ing price/co-payment) in a given area (or domain)
Patient preference methods are flexible in that they: in order to obtain a more desirable composite
Do you know of any websites that you
outcome.
would like to share with the ISPOR
community? If so, contact Bonnie M. • can be designed to focus either on the total value
of medical interventions (including both process Patient reported HR-QoL measures provide real-
Korenblat Donato PhD, at bonnie.
and outcome features), or on the marginal effect time information along general or disease-specific
[email protected].
that modifying a single factor has on the value domains as reported by the patient. This type of
(e.g. marginal innovation); information is useful for qualifying a person's health
4 December 15, 2007 ISPOR CONNECTIONS
status by, for example, adjusting survival measures to account for decreased contexts reveal their implicit preferences [2]. Revealed preference data indi-
quality of life. Unlike patient preference methods, PROs capture patient cate how patients choose among available alternatives under actual clinical
reports of outcomes in individual domains and thus do not provide informa- conditions. While such data can tell us what patients do, they often provide
tion about patient preferences across domains [8]. Information about patient little information on why they do it. SP methods give researchers experimen-
preferences across domains facilitates efforts to characterize and better tal control over choice alternatives, and ensure statistical variability neces-
understand patients' needs and wants. Consider a patient who receives a sary to estimate decision weights for individual features.
steroid treatment for her skin problem. A HR-QoL would capture the patient's
ratings of symptom relief (desired outcome, e.g. relief of itching) and adverse In medicine, while revealed-preference methods have the advantage of incor-
reactions (undesirable outcome, e.g. weight gain) by self-report. Clinicians porating the clinical, emotional, and financial consequences of actual health
or researchers would understand the status or improvement of status with the care decisions, revealed-preference data often cannot answer decision-mak-
steroid treatment. However, the preference of the person regarding the rela- ers' questions. Revealed-preference data may lack variability in these factors
tive importance of the advantage (i.e. symptom relief) versus the discomfort or factors of interest may be confounded with other variables. For example,
(i.e. adverse reaction), or vice versa, is not captured. This relative importance suppose there are only two available treatments. Treatment A has better effi-
is exactly what patient preference methods seek to quantify. Another impor- cacy, a worse side-effect profile, and higher cost. Treatment B has poorer
tant difference between the two approaches relates to the scientific nature of efficacy, a better side-effect profile, and lower cost. If patients choose
the methods - the HR-QoL approach aims to provide a statistically valid meas- Treatment B instead of Treatment A, we cannot say whether they did so
ure that provides general results, while preference methods focus on an eco- because the better side-effect profile was more important than improved effi-
nomic sense of validity (an outcome is valid if it is valued by patients) and can cacy, lower cost was more important than improved efficacy, or some com-
be used to estimate very specific results that allow detailed sub-group analy- bination of effects. Stated-preference methods give researchers experimen-
sis - potentially conducted even at the individual level. tal control over the choice context, which makes it possible to estimate the
relative importance of each factor in the experimental design.
Methods to Measure Patient Preferences
Preferences can be understood and measured in two ways. Stated-prefer- The contingent-valuation or willingness-to-pay (WTP) method is one exam-
ence methods use surveys/questions to elicit subjects' preferences for hypo- ple of a stated-preference method. In its simplest form, subjects are offered
thetical options in an experimental framework. Alternatively, revealed-prefer- a hypothetical treatment with specified features at a specified cost and asked
ence methods are based on observed data relating to individuals' actual whether they would be willing to pay the cost or not. Conjoint analysis (also
behavior. In the latter, individuals never directly state their preferences called discrete-choice experiments or stated-choice method), employs a
(except maybe at an auction), but observed choices made in real decision sequence of evaluations of hypothetical choice sets. In each evaluation, >

< advertisement >

December 15, 2007 ISPOR CONNECTIONS 5

respondents are asked to choose their preferred alternative from a set of two to value the QALY using method more grounded in economic theory such as
or more alternatives in which each alternative is defined by attributes of the WTP, conjoint analysis and other preference related methods. This growing
alternative. Conjoint analysis has a history in market research, environmen- body of research, in some respects, can be thought of as the missing link
tal economics, and transport economics and has found increasing use in between tradition outcomes research methods and modern applications of
health care settings [1, 3, 9]. patient preference methods.

What Do We Learn from Preference Methods? Challenges with Using the Methods
Unlike other PRO methods, stated-preference methods have a clear concep- Despite their wide-ranging applications, challenges exist in the application of
tual basis in neoclassical consumer theory [1, 9]. Although the term utility patient-preference methods, particularly with stated preference methods. As
in health contexts usually refers to von Neumann-Morgenstern expected (car- with any approach that focuses on self-reported results, there are concerns
dinal) utility used to construct quality-adjusted life years, most applied as to whether patients are providing responses that are consistent with their
research in economics relies on neoclassical (ordinal) utility theory based on true preferences [1, 3]. The researcher must create an environment in which
the concept of preferences (i.e. an ordering of possibilities). This theory pos- responders consider choice alternatives and constraints seriously, despite
tulates that an individual has complete preferences (any two options can be not having to experience the consequences of those choices. This bias also
compared to one another), reflexive (a given option can be compared to can be characterized as a lack of acceptability [13]. A related issue has to
itself), and transitive (if A is preferred to B and B is preferred to C, then one do with the extent to which responses are internally consistent [13]. Internal
can conclude that A is preferred to C). These properties are testable. Using consistency can be tested in a number of ways including the addition of dom-
the concept of utility to interpret preferences (it is important to recall that in inant alternatives, use of multiple techniques to value a pair of alternatives,
neoclassical economics utility is derived from preferences, not the reverse), measures of correlation between answers to similar questions, and internal
provides additional mathematical assumptions of continuity (that very small checks for transitive preferences [3].
changes in the consumption basket should not lead to very large changes in
utility), monotonicity (to allow linear modeling of factors), and convexity (to In addition to the challenges of the experimental setting, there are concerns
avoid multiple equilibrium). as to whether patients whose decision making capacity is otherwise incapac-
itated are able to provide, either directly or through a proxy, responses that
WTP provides a monetary measure of what a person is willing to give up in are consistent with their true preferences. These concerns arise, for exam-
order to obtain something else or achieve a specific outcome. Therefore, ple, in the elicitation of preferences among patients who suffer from
WTP is a measure of implicit value rather than a measure of cost. While econ- Alzheimer's disease, dementia, or stroke such that the patient may be unwill-
omists often use WTP to measure value, many people are uncomfortable ing or unable to participate in a study. In addition there are other challenges
about valuing health benefits in monetary units. The same neoclassical util- related to the collection and analysis of the data that might impact the relia-
ity framework that allows us to value dissimilar outcomes in comparable bility of the inferences drawn from the study.
monetary units can be used to value dissimilar outcomes in other common
units. For example, time-tradeoff surveys use stated-preference data to iden- As currently practiced, researchers using conjoint methods apply a set of
tify time-equivalent health states. Conjoint methods can be used to value rules in determining the experimental design - the final set of choice scenar-
both process and outcome attributes of drug therapy [9]. This flexibility of ios that are presented to the patient. The implicit assumption is that the addi-
the method is due to its roots in utility theory and the recognition that patients tion of deleted scenarios to the choice set will not alter the results regarding
care not just about their physical, emotional and mental state while on a given which attributes are important to the decision making process. This assump-
therapy but also about the mode, location, and timing of administration. tion may not hold; there may be omitted scenarios that, if presented, would
Recently, several studies have used conjoint methods to estimate maximum result in a different set of attributes that are identified as important. However,
acceptable risk instead of maximum WTP [10, 11]. In many situations, it continuing advances in experimental design methodology ensure that the
may be sufficient, and even helpful, simply to quantify the relative importance likelihood of a bias is minimized.
weights for efficacy, mild-to-moderate side effects, serious adverse event
risks, convenience factors, and cost without scaling the preference parame- Other challenges can arise once the quantitative data have been assembled
ters in money, time, or risk units [12]. and are ready for analysis. Misspecification of the model used in the analy-
sis might result in misleading or inaccurate assessments of attribute levels
Applications of the Methods important for decision making as well as correlates of WTP. Misspecification
Recent examples of stated-preference methods indicate the wide range of can arise in the form of an omitted conditioning variable; inappropriate func-
applications including outcomes (e.g. illness, death) and process issues tional form (e.g. assuming additive, instead of multiplicative or nonlinear
related to the provision of care (treatment options, treatment locations; edu- effects); violation of the independence of irrelevant alternatives assumption;
cational programs) [3]. Examples include studies to estimate predictors of failure to adjust for endogenous covariates; or failure to adjust for correlated
choice over anemia treatment scenarios among cancer patients [13], deter- data [1]. The increasing use of advanced multivariable models like mixed
minants of non-adherence to bipolar disorder treatments [6], relative impor- logit and hierarchical Bayes regression models lead to more reliable inference
tance of aspects of insulin therapy [12], WTP for a diabetes reduction pro- using stated choice data. Along with the opportunity comes the challenge of
gram among high-risk individuals [5], preferences for HIV treatment options specifying, testing, and drawing correct inference from the model results.
[9], patient preferences over treatment location involving intravenous antibi-
otic therapy [14], parents' WTP to avoid a cold [5], WTP to reduce the risk of Conclusion
death from severe acute respiratory syndrome [16]. It is clear that patient preference methods present an alternative method for
characterizing patient needs and wants. Unlike patient-reported HRQoL
While many think that QALY measure attempt to value the health care states, methods, the focus is on understanding the relative importance of attributes
they fail to adequately embrace preference theory, focusing rather on expect- via revealed or stated preferences. Preference methods are flexible and
ed-utility theory. This said, there has been a movement in outcomes research adaptable to practically any health-related question and are thus uniquely

6 December 15, 2007 ISPOR CONNECTIONS

suited to quantifying the effect of treatment features on adherence, the trade- REFERENCES
offs between health outcomes and other treatment features, the risk-benefit 1 Bridges JF. Stated preference methods in health care evaluation: an emerging methodological par-
adigm in health economics. Appl Health Econ Health Policy 2003;2:213-24.
tradeoffs, and/or monetary valuations related to treatment options. Patient
2 Ryan M, Farrar S. Using conjoint analysis to elicit preferences for health care. BMJ
preference methods offer a scientifically rigorous alternative to traditional 2000;320:1530-3.
patient-centered outcomes research methods and are worth a closer look. 3 Ryan MD, Scott A. Eliciting public preferences for healthcare: a systematic review of techniques.
Health Technol Assess 2001;5:1-186.
Postscript 4 Bridges JF, Jones C. Patient-based health technology assessment: A vision of the future. Int J
Technol Assess Health Care 2007;23:30-5.
Future work of the ISPOR Patient Preference Methods Special Interest Group
5 Johnson FR, Manjunath R, et al. High-risk individuals' willingness to pay for diabetes risk-reduc-
(SIG) aims to increase the appropriate application of contingent valuation,
tion programs. Diabetes Care 2006;29:1351-6.
conjoint analysis, stated choice/preference, discrete-choice experiment and 6 Johnson FR, Ozdemir S, et al. Factors that affect adherence to bipolar disorder treatments: a stat-
similar methods for assessing patient preferences and values in outcomes ed-preference approach. Med Care 2007;45:545-52.
research and encourage the use of this information by decision-makers. In 7 Donaldson C, Hundley V, et al. Willingness to pay: a method for measuring preferences for mater-
the coming year we will focus on developing recommendations for good nity care? Birth 1998;25:32-9.

research practices for patient preferences methods, initially focusing on 8 Johnson FR, Hauber AB, Osoba D, et al. Are chemotherapy patients' HRQoL importance weights
consistent with linear scoring rules? A stated-choice approach. Qual Life Res 2006;15:285-98.
conjoint analysis methods. Interested ISPOR members who would like
9 Phillips KA., Maddala T, et al. Measuring preferences for health care interventions using conjoint
to contribute to these activities should refer to our web page at: analysis: an application to HIV testing. Health Serv Res 2002;37:1681-705.
(www.ispor.org/sigs/PRO_PPMl.asp) or contact the group's chair, John F. P. 10 Johnson FR, Ozdemir S, Hauber AB, Kauf T. Women's willingness to accept risk for perceived
Bridges at: [email protected]. vasomotor symptom relief. Journal of Women's Health 2007 in press.
11 Johnson FR, Ozdemir S, Mansfield CA, et al. Crohn's disease patients' benefit-risk preferences:
serious adverse event risks versus treatment efficacy. Gastroenterology 2007.
Acknowledgements
12 Hauber AB, Johnson FR, et al. Risking health to avoid injections: preferences of Canadians with
This article was published on behalf of ISPOR's Patient Reported Outcomes type 2 diabetes. Diabetes Care 2005;28:2243-5.
Special Interest Group, Patient Preference Methods Working Group. The 13 Ossa DF, Briggs A, et al. Recombinant erythropoietin for chemotherapy-related anaemia: eco-
authors acknowledge the input of past and present members of the working nomic value and health-related quality-of-life assessment using direct utility elicitation and discrete
group. For further information on the activities and output of the group, choice experiment methods. Pharmacoeconomics 2007;25:223-37.

please visit our webpage: http://www.ispor.org/sigs/PRO_PPMl.asp. IC 14 Marra CA, Frighetto L, et al. Willingness to pay to assess patient preferences for therapy in a
Canadian setting. BMC Health Serv Res 2005;5:43.
15 Liu JT, Hammitt JK, et al. Mother's willingness to pay for her own and her child's health: a con-
tingent valuation study in Taiwan. Health Econ 2000;9:319-26.
16 Liu JT, Hammitt JK, et al. Valuation of the risk of SARS in Taiwan. Health Econ 2005;14:83-91.

< advertisement >

December 15, 2007 ISPOR CONNECTIONS 7