Psychometry and Quality of Life

Outline
• Concept of HRQoL, preferences and utilities
• Methods to measure utilities
• QALYs

2
• Some measures of health outcome are restricted to a single
aspect or dimension of health: for example, effects on
mortality and survival. However, although length of life is
clearly an important aspect of health, the quality in which it is
lived is also important.
• There are very many measures which describe the different
dimensions or attributes of health-related quality of life
(HRQoL).
 Why
1. utilities?
Global measure of health status, incorporate quality of life,
value judgments

2. Allow comparison
Between interventions
Across programs

3. Strong theoretical foundation

→ can be used in economic evaluation

4
→ Health-related Quality of
Life (HRQoL) HRQoL

HRQoL is a
multidimensional measure
of the effect of a health
condition or intervention on
an individual’s overall well-
being, encompassing
physical and occupational
function, psychological
state, social interaction
and somatic sensation.
(Schipper J 1996)
5
 HRQoL, Preferences & Utilities
• Many methods measure HRQoL.
• CUA requires preference-based measures:
A single, overall summary score that numerically reflects the value of a
given health state.
• Terms:
– “preference” and “utility” are generally treated as synonymous. but are NOT
– “utilities”: preferences obtained by methods that involve choices made under
uncertainty (Drummond 2015).
– Preference is the umbrella term that describes the overall concept;
utilities and values are different types of preferences.
5
 Utilities and HRQoL
Disciplin What is How? Scores Weights Application
ary measured?
origins

Utility Utilitarian- GLOBAL Direct/ 0-1 (some- Preference As follows…

ism, health status indirect times <1) weights
economics utility
elicitation

Selected attributes, Question- Variable Usually Assess

HRQoL Social occasionally overall naire none outcomes
science HRQoL (profile) in RCT,
s cohort
studies
etc.

7
Slide courtesy of Murray Krahn
 A Taxonomy
Health Status Measures

Functional Status
HRQoL
(6 minute walking distance)

Psychometric
(SF-36)
Preference-based

Non-utility Utility

Direct Indirect
8
 Utility
• An individual’s
– “strength of preference”
– for outcomes
– under conditions of uncertainty

• Sox: “a number for comparing gambles”such that the gamble

with the highest expected utility should be preferred.
• Quantitative
• “Summary” quality of life measures
• Comprehensive
9
 Utilit
• Concept from economics y
• Developed by von Neumann and Morgenstern
• Consider 3 health states
– Ankle sprain utility = 0.6
– Partial recovery utility = 0.8
– Full recovery utility = 1.0

• Ordinal utility - ranks in terms of preference

– Full rec > (is preferred to) Partial rec > Ankle sprain

• Cardinal utility, interval scale

– Gives value of utility to different options 9
MEASURING UTILITIES
 How to Measure Utilities
• WHO?
Patients vs. general public (Direct vs. proxy assessment)
• WHAT?
Own health vs. standardized health
• HOW?
– Holistic vs. decomposed
– Direct vs. indirect measurement
– Interviewer vs. paper vs. computer
 Direct Preference Measures

Question framing
Response method Certainty (values) Uncertainty
(utilities)

Scaling Rating scale

Visual analogue scale etc

Choice Time trade-off Paired Standard gamble

comparison
etc
Utilities? 12
 A Taxonomy

Health Status Measures

Functional Status Quality of life

Psychometric
Preference-based
(SF-36)

Non-utility Utilit
y

Direct Indirect
Rating Scale(1-100)/VAS(10 cm
line)/catagory scaling(1-10)

• Imagine you have rheumatoid arthritis: You have constant partially controlled pain, but can do
most daily tasks, though with much difficulty. The joints of your hands are now deformed,
but function with assistive devices.

Rate your quality of life?

• “Feeling thermometer”
• Doesn’t assess
preference under
condition of
uncertainty
• Not a “utility” Health state
? description

Mark a line, or assign a

• Advantages: Quick, easy, number corresponding to
your current health state
cheap
• Disadvantage- end-of-
scale bias & context bias
 Time Trade-Off (TTO) Method

You determine the point of indifference:

Y years in perfect health * uwell = x years with disease * udisease

udisease = (Y * uwell )/X

“Y” years of life in “perfect health”
Healthy

OR
“X” years of life with “disease”
Statei

Dead 16
Y X Time
 Time Trade-Off (TTO) Method
Choose one of the following alternatives:
1. live with Cancer for 2 years
OR
2. a new treatment which yields
– best possible quality of life
– 1½ year life expectancy

Calculate the utility for cancer!

17
Time Trade-Off (TTO) Method- Classical for BTD

• (X) 10 years in Hhealth state = “Y” years in full health

• Identify the value of “Y” that results in
indifferences

→U (H) = Y/10

Example: U (H) = 0.5

0.5= Y/10
y= 5
18
 A Taxonomy

Health Status Measures

Functional Status) Quality of life

Psychometric
Preference-based
(SF-36)

Non-utility Utilit
y

Direct Indirect
 Direct Utility Measurement

• In health, usually scaled from

– 0 = death to
– 1 = best possible health

• Standard Gamble
• Offered two alternatives:

* Option-1(Gamble): magic pill with 2 possible outcomes: either

patient lives in perfect health (probability P); or dies immediately
(probability 1—P)

* Option-2 (Sure thing): certain outcome of health state “x (e.g. to

live with cancer)”

• What chance of full health for indifference?

 The Standard Gamble
Sure Thing
u(Cancer)

Gamble u(Best Health)

0.75

u(Dead))
0.25

Which would you prefer?

• The utility score for this person for this disease state or condition would be
calculated as the probability (p) of living a normal life after the operation.
 SG in Decision Tree Format

Sure Thing Cancer uCancer

Best Health
uBestHealth= 1
Gamble p
Dead
uDead = 0
1-p
At indifference:

uCancer = p*uBestHealth+ (1-p)*uDead uCancer =

p*1+ (1-p)*0 23
SG Example:

• “Would you take the magic pill if the chance of death were . . % and the chance
of full health for 25 years is . . %, or you would prefer to remain in your
present state of health for the next 15 years?“.

 Relative risk attitude?

• Individual face-to-face interviews with the subjects, complete
with carefully scripted interviews and helpful visual aids (Furlong et
al. 1990). Other, more efficient techniques are, however, being
developed. These include interactive computer approaches (Lenert
2001), paper-based approaches (Ross et al. 2003), and group
interviews with paper based response .
Risk Aversion
Sure Thing
Best Health
u(15 years)

Best Health
Gamble u(25 years)
0.60
Dead
u(0 years)
0.40

Although 15 = 0.60 x 25
u(15) > 0.60 x u(25)
 Risk Seeking

Don’t Buy Ticket Keep money u($0)

Lose
u($-5)
Buy Ticket 0.999
Win
u($1000)
0.001

Expected value = - 4.00

(0.999*(-5)+ 0.001*(1000)=-3.995/-4
Expected utility = ?
 Risk Attitude
•
Risk sneeuktrianlgity==“I lnince r eaar s u it ni lgity mf ua nrgcitn ioanl utility” (concave
• Risk neutrality = linear utility function convex
• Risk seeking = “Increasing marginal utility” (convex)

1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25
 Patterns of Risk Attitude
Gain Loss
Low Probability Risk Seeking Risk Averse
Example: Lottery Example: Insurance

High Probability Risk Averse Risk Seeking

Example: Clinical trial Example: Highly fatal illness
participation

• What’s a gain and what’s a loss? E.g. immunization

• Risk attitude can change with context, valuation
 Empiric Observations
• SG scores often different from TTO, correlation often poor
SG>TTO>VAS
• SG scores tend to be high
• SG and TTO can be internally inconsistent.
• Is SG harder ? Little empiric evidence!

• The reason given for the differences between cells 3 and 4 is risk
attitude, which is only captured in cell 4. The reason for the
difference between cells 1 and 3 presumably lies in the difference
between choosing and scaling.
 Health Status &
Utility
• Often don’t correlate
• But no a priori reason to expect this correlation
• Respondent effects sometimes observed
– Health care proxies  low ratings
– Race, gender
 Utility Assessment
• Population
– Selection
– Sampling
– Recruitment
• Biases
– Anchoring
– Framing
– Labelling
– Several others
• Adaptation
 A Taxonomy
Health Status
Measures

Functional Status Quality of life

Psychometric Preference-based

Non-utility Utilit
y

Direct Indirect
 Indirect Measures
• Generic measures: QWB, HUI, EQ-5D Components:
– Health state classification system
• Definition of attributes
• Levels of attributes
– Weighting system
– Transformation to utility scores
• Advantages:
– Feasibility
– Community Preference Scales
33
QWB, Quality of Well-Being; HUI, Health Utility Index; EQ, EuroQol
 Overview
Attributes Levels per Preference Scoring Number of
attribute Measure Health
States
QWB Mobility, social activity, 3-5 RS Additive 945
physical activity,
symptoms
HUI II-III Vision, hearing, speech, 5-6 RS (SG for 4 Multiplicative 972,000
ambulation, dexterity, “marker”
emotion, cognition, pain states)

SF-6D Physical, role 4-6 SG Additive + 18,000

(SF36) limitations, social, pain, constant
mental, vitality

EQ-5D Mobility, self care, usual 5 TTO/VAS/DCE Additive + 3,125

activity, pain, anxiety, constant
depression
34
QWB, Quality of Well-Being; HUI, Health Utility Index; EQ, EuroQol
 EQ-5D
• Patients’ responses of the EQ-5D-5L describe
their own “health state”
• A unique health state can be defined by
combining 1 level from each of the 5
dimension of HRQoL

35
 EQ-5D Scoring
• Each state gets a 5 digit “score” (i.e. 3,125 (=55) possible health states)
– Each digit = score on individual domain
– 11111 = no problems
– 33333 (now:55555) = worst score on all domains
– Unconscious, dead included but not measured
• Health states mapped to preference scores
• Country-specific
– scoring algorithms
– Validation method (VAS or TTO with/without DCE)

• Endorsed by many countries (e.g.UK, NICE) as the “reference 36

standard” for measuring utilities
EQ-5D Transformation

37
 Which Method?
• Response Scale
• Cognitive Burden
• Theoretical Rigor
– Preference (RS, TTO, SG)
– Choice Based (TTO, SG)
– Under uncertainty (SG)

• Panel: “not clear that incorporation of risk attitude is

necessary for CEA” 38
QALYS
 QALYs
Quality Adjusted Life Year

Central concept:
• Health is a 2 attribute function:
– Length of life
– Quality of life
 QALYs
Combine duration and quality of life:

HRQOL

Quality adjusted
Life Years

Duration of Life
 QALYs gained from an intervention

An ideal outcome: Longer life and higher quality of life! 42

To satisfy the QALY concept, weights must be
1) Based on preferences
2) Anchored on perfect health and dead
– well specified and understood
– Full Health = 1, dead = 0 (convenient but not necessary)
3) Measured on an interval scale
– Every interval of equal size must have same value
– 0.2-0.4 = 0.8-1.0
43
Torrance et. al.
 QALY Concerns
• Failure to give priority to those worst off

• Discrimination against people with limited treatment potential

• Failure to account for qualitative differences in outcomes

(life saving vs. health improving)
→ Social value of health
 Are QALYs Good Enough?
• Recommended by Panel on cost-effectiveness, NICE in UK,
CADTH in Canada
• Suggest good trade-off between feasibility and theory
• Remain controversial
• Widely used
• Alternatives all have significant disadvantages
Thank you