Third Plenary Session
ISPOR 14th Annual International Meeting
Effectiveness in CEA: QALYS
Life expectancy multiplied by health-related quality of life:
Quality-adjusted life years
Calculating QALYS:
If, HRQL= 0.7
And,A treatment gives 10 extra years of
life (@ 0.7 per year)
Then…. People receiving the treatment gain
an average of 7 QALYs each
A QALY is a QALY is a QALY
#People HRQL LE = QALYsSaves 100 x 0.8 x 50 = 4000 Lives Improves 10,000 x 0.1 x 4 = 4000 HRQL
The cost-effectiveness of one thing compared to
another…
Cost treatment 1 – Cost treatment 2 Effectiveness treatment 1 – Effectiveness treatment 2
= COST per QALY
For example…
Cost Life Expectancy HRQL QALYS
Group A $80,000 2 Years X .6 = 1.2 Group B $ 8,000 1 Year X .8 = 0.8 Cost-effectiveness:
$80,000 - $8,000 = $72,000 = $180,000/QALY 1.2 – 0.8 0.4
The QALY is a widely used measure of health gain.
Long-standing criticism of the theoretical basis, and practical application of, QALYs.
Plenary session, 10th Annual International Meeting (Kahneman), May 2005.
Issues panel, 11th Annual International Meeting (Fryback, Kahneman, McGuire), May 2006.
Two-day invitational consensus development workshop, November 2007.
Funding from AHRQ and NCI. 25 participants. Discussion of:
- the basics of QALYs;
- the main challenges surrounding QALYs;
- retaining and enhancing QALYs;
- the use of QALYs in decision-making;
- towards a consensus on the QALY
Milton C. Weinstein PhDHarvard School of Public Health, Boston MA, USA
George Torrance PhDMcMaster University, Hamilton, ON, Canada
Alistair McGuire PhDLondon School of Economics, London, UK
Method for valuing health effectiveness in cost-effectiveness analysis for resource allocation decisions
Values health based on time spent in health states
Endorsed by US Panel and NICE for reference case
Represent individual patient preferences
Reflect equity, fairness, or political goals
Grounded in decision science (based on expected utility theory)
Individuals move through health states over time.
Each health state has a “value” Health = value-weighted time (QALYs)
Perfect health = 1.0, dead = 0.0
Interval scale properties e.g., 0.2 0.4 = 0.6 0.8
States worse than dead have negative value
Value = preference (desirability) Valued by whom?
individuals experiencing a health state or illness
individual who may or may not experience that health state in the future
individuals considering the health of a community
(Values are for health states, not for changes in health states)
What is being valued? Whom do we ask? What do we ask? How are health outcomes defined?
What is being valued? Whom do we ask? How are health outcomes defined?
Conventional QALYs allow for different answers to these questions
The answer depends on the question being asked
Societal resource allocation: priority setting across proposed programs or interventions
Societal (programmatic) audit: evaluation of ongoing activities/programs
Personal clinical decisions or decisions about insurance coverage
Personal clinical/insurance choice
desirability of health outcomes to the individual
ex ante perspective
Societal/program audit
current health of affected population members, as valued by themselves
ex post perspective (i.e. patient preferences/experience utility)
Societal resource allocation
individual health (aggregated) or community health
ex ante or ex post
Personal clinical/insurance
the individual, +/- informed by patients/disabled people
Societal/program audit
members of the affected population
Societal/individual health (aggregated)
or Societal/community health
representative sample of population○ including patients/disabled people○ informed by patients/disabled people
HUI-2HUI-3SF-6DQWBEQ-5D15DAQOL
patient health statecommunity survey value scoreUS Panel and NICE reference case method
different instruments give different resultsvalue scores may be population-specific
Health statesvaluation independent of duration or
sequence
Health statesconventional QALY approachvaluation independent of duration or sequence
Health paths (profiles)theoretically superiorpractical problem: large number of paths
Health changescan incorporate equity or fairnesspractical problem: large number of changesorder of changes matters
QALYs should be discounted at the same rate as costs (US Panel, NICE)
Erik Nord PhDNorwegian Institute of Public Health, Oslo, Norway
Norman Daniels PhDDepartment of Population and International Health, Harvard School
of Public Health, Boston, MA, USA
Mark Kamlet PhDHeinz School of Public Policy and Management, Carnegie Mellon
University, Pittsburgh, PA,USA
The Conventional QALY
Defined as expressing the personal utility of health outcomes as judged ex ante, “on average,” by the general public, from behind a veil of ignorance, about future health, based on self interest.
Issues: (i) Substantial, empirical inter-method
variation in ex ante assessmentSG yields higher values than TTO and greater in
turn than rating scaleWhich one is “right”?
(ii) Empirical unwillingness to trade-off lifetime Means that less is invested in preventing the
outcome “confined to a wheel chair”Use of “experience” utility disfavors prevention,
use of ex-ante utility doesn’t capture adaptation/foregone opportunities
iii. Concerns for fairness
No consideration for pretreatment health state
○ At odds with ethical theory/public opinion that suggests that in setting priorities societies often emphasize how bad off individuals would be without intervention
○ i.e. concern for “severity”
iii.Concerns for fairness
Conventional QALY model implies that the value of an intervention is proportional to the beneficiary’s capacity to benefitAt odds with theory/public opinion that it should not
be held again people that they have conditions for which there are no complete cures or whose remaining lifetime is shorter
Similarly, life years gained for those at full health valued more than life years gained for those at less than full healthConflicts with equal right to protection of life by all
iv. Subtraction doesn’t “add up”
Standard QALYs measure differences in health states, not gains in healthEx ante preference elicitation on health states and
subsequent subtraction of health state values from one another
Decreases data requirements, i.e. the number of possible changes is much highter than the number of possible states
Nonetheless, this is a proxy approach, yet to be validated in the health economics literature
Incorporating concerns
for fairness Count as “1” all gained life years if good enough to be desired by affected personsLeads to inconsistencies with individual preferences
Place less weight on the duration of health benefits in comparisons of programs for patients with different life expectancies
Add explicit equity weightsOverload the model?
Different “priority classes” for QALYS with different ratio cut-offs
Treat “prevention” differently than “treatment”
Joseph Lipscomb PhDDepartment of Health Policy and Management, Rollins School of
Public Health, Emory University, Atlanta, GA, USA
Dennis Fryback PhDUniversity of York, York, UK
Marthe Gold MD, MPHUniversity of Wisconsin, Madison, WI, USA
Dennis Revicki PhDCity University of New York Medical School, New York, NY, USA
For analyses requiring a summary measure of health that integrates quantity of life and quality of life, QALY is arguably the gold standard.
But should it truly be the coin of the realm? Substantive concerns have been raised
Our conclusion (in preview): These conceptual and methods issues signal opportunities for making important incremental improvements in the QALY – rather than abandoning the construct.
Conceptualization and construction of health states
- Which domains?
- Which health levels within each domain?
Psychometric approaches for eliciting preferences
Statistical strategies for deriving overall value weights
for QALYconv
For example…… Highly simple model structure: QALYconv linearly additive
function of time in health states, with an exponential discounting factor to reflect time preference
Distributional and other ethical issues not formally integrated into model
Some suggest that the value component of QALY should be “experience-based” (from real-time perspective) rather than “ex ante”
In fact, QALYconv plays important role in regulatory and purchasing decisions in many industrialized nations But push back with NICE Non-QALY approaches being taken in France/Germany
Much less in U.S: Not (yet) by FDA Not (yet) by Center for Medicare & Medicaid Service Not (yet) by most private payers Recent studies suggest that resistance to CEA may be less than
suggested particularly given serious cost issues within public programs and employers
Should we abandon the QALY?
QALYconv has proved to be serviceable vehicle for quantifying joint mortality-morbidity impacts at individual and population level
To abandon QALYconv now is to sever link to hundreds of published studies & multiple ongoing investigations – including many capitalizing on data now collected in national surveys, clinical trials, observational studies…..
A more productive pathway: pursue program of research that takes QALYconv as starting point for “continuous QALY improvement” over time
Prominent “health measurement systems” (e.g., HUI2/3, EQ-5D, QWB, and also SF-36) are the major engines behind preference-based assessments, including CEAs and population surveillance
Multiple applications now in national surveys of health (e.g.,MEPS, U.S.-Canada Joint Survey, Medicare HOS)
But, basic issues about what exactly to measure remain active topics for investigation All major systems view health as multidimensional concept, but
domains vary across systems Even for similar domains (e.g., Mobility/Ambulation) item content
differs across systems
Variations in domain structure allows selection of particular QALYconv deemed best for application at hand – but does not promote comparability across studies. Solutions? Work toward “consensus domain structure” as one aspect of
community-based deliberative processes to identify and codify citizen perspectives on health measurement, or else
Cross-walk QALY scores between measurement systems To appraise, and improve, item content (within a
domain), apply well-defined psychometric methods Mixed qualitative-quantitative approaches to improve content
and construct validity e.g. HUI2/3 and SF36v1/v2 Promising development: application of item response theory
(irt) methods, e.g., Pickard et al. to study 3-Level vs. 5-Level EQ-5D
Across the major health measurement systems, derivation of the value component of QALYconv (the “Q part”) varies in important ways:
Method for eliciting preferences
- Standard gamble vs. time-tradeoff vs. visual analog scale
- Assumed duration of health state (1 day, 1 year, 10 years)
Deriving aggregate QALY score for a multidimensional health state
- Multi-attribute utility modeling (HUI2/3)
- Econometric modeling (EQ-5D, QWB, SF-6D)
States worse than death
- EQ-5D and HUI2/3 allow for them
- QWB, SF-6D, HALex do not
Cross-walk scores across measurement systems Predicting one set of QALY scores from another via
regression modeling (e.g., an EQ-5D to SF-6D mapping) Hierarchical irt modeling (a la Fryback et al) to map (for
example) from EQ-5D to irt-derived latent variable continuum to SF-6D
or……. Initiate consensus process leading to
“Reference Case” QALY, establishing baseline expectations about health state definition and valuation
Additional issues raised about QALYconv No health state duration or sequencing
effects incorporated Investigate preferences over multi-state health profiles,
with states drawn from current measurement systems (e.g., from the EQ-5D)
Value component of QALYconv based (typically) on community-derived health state preferences Instead, draw community preferences from those who’ve
experienced the states of health (Nord proposes SAVE) Instead, of using ex ante community preferences, use
experience-based valuations (Dolan and Kahneman) (The challenge is how to operationalize for efficient
application to health program evaluation)
Fairness matters – but it is not a matter that can be settled by QALYconv
In response….
Equity Weighting: Factor fairness directly into the preference weighting process (e.g., approaches advanced by Nord, by Wagstaff, and by Johannesson)
Constrained Optimization Modeling: Maximizing QALY improvement, subject to meeting equity conditions (e.g., as illustrated by Stinnett and Paltiel and by Chen and Bush)
Community-Based Deliberative Processes where the implications of cost-effectiveness decisions based on
QALYconv can be examined for fidelity to public values (e.g. Citizen’s Councils)
Must one decide between “building a bridge over troubled QALYs” and “sailing off into the less-charted waters” of non-QALY approaches?
False Choice! Instead, cross that bridge and boldly set sail to new
lands,
(BUT treat the conventional QALY as the point of departure for the development of new models in order to capitalize on what has been learned across many years. This will also allow us to maintain continuity/comparability in tracking of trends in population health and in CEAs.)
Paul KindUniversity of York, York, UK
Jennifer Elston Lafata PhDCenter for Health Services Research, Henry Ford Health System,
Detroit, MI, USA
Karl Matuszewski MS, PharmDElsevier/Gold Standard, Tampa, FL, USA
Dennis Raisch BSPharm, MS, PhDUniversity of New Mexico, Albuquerque, MN, USA
Who are they?Faceless bureaucrats?Company management?Health plan CEOs?Mysterious, nameless committees?Doctors & hospitals?Patients?All of the above?
Matrix of potential users/uses of QALYsPatients, provider(s), employer/ins, govt.Observing health status, comparing to
reference norm, changes over timeAggregation levels (individual, groups, pop.)
Lack of consensus on HrQoL by instrument developers, economists
There is hope, there is no better alternative
Pick a measure/methodology Incorporate into studies Report results Educate all constituencies
1. One health-based input to decisions (of many)
2. Can be used at various levels in health system
3. Reference method is required
Michael Drummond DPhilUniversity of York, York, UK
Diana Brixner BSPharm, PhDUniversity of Utah, Salt Lake City, UT, USA
Marthe Gold, MD, MPHCity University of New York Medical School, New York, NY USA
Paul KindUniversity of York, York, UK
Alistair McGuire PhDLondon School of Economics, London, UK
Erik Nord PhDNorwegian Institute of Public Health, Oslo, Norway
Context for consensus
Many “perspectives” Desire to move economic analyses
along in a manner that overcomes outsider scepticism
Therefore: Best agreement on high level principlesAreas of departure described in the agenda
for further research
1. QALYs are but one input to health care
decision-making There are always additional factors that society and
decision-makers must take into account, i.e.Equity, social justiceBudget
These factors need to be incorporated within healthcare decision making in a manner that allows:Transparency Legitimacy
2. QALYs can be used in different ways at multi-levels
of health care Traditionally QALYs have been used for resource
allocation between groups in the population QALYs can serve additional purposes in systems
where there is no universal budgetThe context may be narrower when the
government is not the funder of health care e.g.,○ different treatments for the same condition○ focusing on health outcomes rather than $s.
3. Health is a determinant of well-
being The QALY is a measure and valuation of
health, but intersectoral decision-making may benefit from the broader context of the concept of “well-being”
(In the interim the QALY remains highly serviceable for population-based health care decision making)
4. Both ex-ante preferences and
experience should count Health state valuations should reflect both
the experience of people who have familiarity with them, as well as ex-ante preferences that reflect forgone capabilities and opportunities
But, how do we take account of differences between patients and the “inexperienced”
Can differences be melded into a common language to serve broader resource allocation decisions?
5. Distributive issues must be addressed
satisfactorily Cost/QALY is a measure of efficiency,
not of fairness Failure to attend to this distinction will
sink the credibility of economic analyses
(Distributive issues can be addressed within the QALY measure itself, or within the accompanying decision-making process)
6. Different methods for valuing QALYs yield different results
7. The QALY approach to summing health gains over time is simplistic
The Panel on Cost-Effectiveness in Health and
Medicine, 1996 Recommendations for Reference Case analyses stopped short of endorsing a single approach. Agreement: Generic HRQL 0-1 scale preference based include domains important to the problem under
consideration; include effects of morbidity on productivity and
leisure PCEHM research recommendations
Support research that assesses the performance of different measurement strategies in relationship to others
Compare valuation strategies Develop a catalogue of weights
8. The time has come for all good health economists to
rally behind a Reference Case QALY
As an outcome measure for decision making, the QALY continues to fall short of its potentialCEAs remain non-comparable across diseases and
interventionsAcademic debate rolls on and distracts from progress
in moving economic analyses into mainstream decision making
A process should be put in place to define a reference case for QALY measurement
Acclaim a measure (without excluding others)
Use cross-walks that allow interpretability between studies using different measures
Set up a consensus group to determine criteria for a preferred approachAn extant measure?Cross walks?New measure?
Reference Case QALY II: Ways Forward
Michael Drummond DPhil Professor of Health EconomicsCentre for Health EconomicsUniversity of YorkYork, UK
Publication of the consensus workshop proceedings in Value in Health (vol. 12, suppl. 1, 2009) available free at: http://www.ispor.org/meetings/Invitational/WorkshopPhila1107.asp
Accompanying editorials, raising additional challenges
Official pushback on QALYs, most notably in Germany.
NICE’s supplementary guidance on ‘end of life’ therapy.
Research agenda from the workshop.
Some additional thoughts.
The relevance of non-health objectives to health-care decision-making.
Case studies on the use of QALYs at different levels in the health-care system.
Case studies on the use of QALYs in decentralized, privately funded health-care systems.
The impact of health on broader well-being. The role of an expanded QALY,
incorporating dimensions other than health.
The relationship between the valuations of those experiencing, or having experience of, health states and the valuations of the general public.
Methods for briefing members of the public on the experiences of those in particular health states.
Comparisons between the main methods for valuing health in respect of their incorporation of distributional concerns.
Weighting schemes reflecting distributional concerns.
Qualitative research into the community’s views about distributional concerns.
Evaluation of health gains versus the evaluation of health states.
Research into the assumption of linearity of preferences over time and the ways of obtaining valuations of pathways or profiles.
Development of a reference case, or series of reference cases, for estimating QALYs.
Any decision-making procedure for allocating healthcare resources needs to weigh benefits and costs, plus their distribution.
All approaches are value-laden, whether the values are made explicit or not.
The main issue for health economics and outcomes research is the extent to which quantification and aggregation contributes to the decision-making process.