CEA Fundamentals: Valuing Outcomes

Learning Objectives and Outline

Learning Objectives

Understand the concepts of summary measures of health, specifically, quality-adjusted life years (QALYs)
Describe the general differences between direct and indirect methods for estimating health state utilities
Curate model parameters for quantifying “benefits” (the denominator in the C/E ratio)

Outline

Clinical Outcomes

Summary Measurements

Alternatives to ICERs

ICER review

Numerator (costs)

Valued in monetary terms

Examples:

$USD
₦NGN
KES
R

\frac{\colorbox{#CfAE70}{$C_1 - C_0 \quad (\Delta C)$}}{E_1 - E_0 \quad (\Delta E)}

Denominator (benefits)

Valued in terms of clinical outcomes

Examples:

# of HIV cases prevented
# of children seizure free
# of quality-adjusted life years gained

\frac{C_1 - C_0 \quad (\Delta C)}{\colorbox{#CfAE70}{$E_1 - E_0 \quad (\Delta E)$}}

Benefits

What’s important for the question at hand
Most analyses report several different outcomes
QALYs/DALYs enable comparability across disease areas

Valuing Health Outcomes

Clinical Outcomes

What are they?

Clinical outcomes allow us to measure particular events in a decision tree or health economic model.

For Example:

# of HIV cases prevented

# of children seizure free

# of healthy pregnancies

# of hospital visits

# of disease deaths

# of cancer progressions

When are they used?

Policy holders need the information
The outcome holds high importance
Short-term models

How do you choose one?

“Is implementing screening for colon cancer cost-effective?”

What concerns might arise for each of these?

Number of cases of colon cancer Screening will increase the number of cases caught.
Number of cases of STAGE 1 colon cancer Could heavily favor Screening
Number of cancer deaths Deaths aren’t the healthcare burden
Number of hospitalizations from colon cancer Minimal Data
Number of screenings completed Focuses on implementation
etc.

All these are still great outcomes, they just highlight important considerations when choosing the one for your research question

A Well Defined Clinical Outcome

Specific

Each outcome should describe one clear event or condition.
If you combine too many things into one outcome, it can be hard to understand why strategies are different.
When possible, use separate outcomes for different events (for example: survival, complications, hospitalizations).

Prevent biasing a strategy

Make sure the outcome truly shows what is happening clinically, not just what is cheap or expensive.
Some bad events may look “good” in a model because they cost less. (For example, death has low cost, but it is clearly a bad outcome.)

Answers your research question

Chosen outcomes should generate information that is meaningful for the research objective and decision context.
Focus model structure and computational effort on outcomes that influence decision‑making ot drive differences in outcomes.

Data is available

As with much of economic modeling, clinical outcomes rely on accurate data

Summary Measurements

Why summary measures of health?

QALYs and DALYs both provide a summary measure of health

Allow comparison of health attainment/disease burden

Across diseases
Across populations
Over time
Etc.

Intro to QALYs

Origin story: welfare economics

Utility = holistic measure of satisfaction or well-being

With QALYs, two dimensions of interest:

Length of life
measured in life-years

Quality of life
measured by utility weight, usually between 0 and 1

QALYs

A metric that reflects both changes in life expectancy and quality of life (pain, function, or both).

Formula:

Quality Adjusted Life Years =
Sum of weight * duration of life

Example: Patient with coronary heart disease (with surgery)

QALYS = (3yrs * 1.000) + (1.5yrs * 0.5) + (5yrs * 0.75) + (0.5yrs * 0.25)

= 7.625 QALYs

Source: Harvard Decision Science

Example: Patient with coronary heart disease (without surgery)

QALYS = (3yrs * 1.000) + (1.5yrs * 0.5) + (5yrs * 0.5) + (0.5yrs * 0.25)

= 6.375 QALYs

Source: Harvard Decision Science

Example: Patient with coronary heart disease

Life Years

With surgery: 10 LYs
Without surgery: 10 LYs
Benefit from surgery intervention:
10LYs – 10LYs
= 0 LYs

QALYs

With surgery: 7.625 QALYs
Without surgery: 6.375 QALYs
Benefit from surgery intervention:
7.625 QALYs – 6.375 QALYs
= 1.25 QALYs

Utility weights – How are they obtained?

Utility weights for most health states are between 0 (death) and 1 (perfect health)

Direct methods:

Standard gamble
Time trade-Off
Rating scales

Indirect methods:

EQ-5D
Other utility instrument: SF-36; Health Utilities Index (HUI)

Direct methods - Standard Gamble (SG)

What risk of death would you accept in order to avoid [living with an amputated leg for the rest of your life] and live the rest of your life in perfect health?

Find the threshold p that sets EV(A) = EV(B)
Assume respondent answered that they would be indifferent between A and B at a threshold of pDeath = 0.10
Then U(Amputation) = p*U(Death) + (1-p)*U(Perfect Health) = 0.10*0 + (1-0.10)*1 = 0.9 = threshold of indifference between surgery & no surgery (I will live with this rather than have a high risk of dying)

The STANDARD GAMBLE technique asks patients (FOR EXAMPLE) - I WANT YOU TO IMAGINE
Your leg gets amputated above the knee
This doesn’t impact your life expectancy, but you now live your life with an amputated leg
Now, there’s a NEW TREATMENT available where they can re-attach your leg
But the treatment is RISKY & those undergoing it have a probability of dying
The standard gamble technique asks you to find the THRESHOLD of indifference between SURGERY & NO SURGERY
In other words, let’s say you have a 30% probability of dying: Would you do the treatment?

What about a 20% probability of dying?
What about a 10%? - Ajay says yes, .90 is your utility value for the health state of living with an amputated knee Etc.

Okay, I will live with this rather than have any high risk of dying”***

This is the STANDARD GAMBLE TECHNIQUE

Direct methods - Standard Gamble (SG)

What risk of death would you accept in order to avoid [living with stroke for the rest of your life] and live the rest of your life in perfect health?

As a result of a stroke, you

Have impaired use of your left arm and leg
Need some help bathing and dressing
Need a cane or other device to walk
Experience mild pain a few days per week
Are able to work, with some modifications
Need assistance with shopping, household chores, errands
Feel anxious and depressed sometimes

Direct methods - Time Trade-Off (TTO)

An alternative to standard gamble
Instead of risk of death, TTO uses time alive to value health states
Does not involve uncertainty in choices
Task might be easier for some respondents compared to standard gamble

Direct methods - Time Trade-Off (TTO)

What portion of your current life expectancy of 40 years would you give up to improve your current health state stroke to ‘perfect health’?

U(Post-Stroke) * 40 years = U(Perfect Health) * 25 years + U(Dead) * 15 years

U(Post-Stroke) * 40 years = 1 * 25 years + 0 * 15 years

U(Post-Stroke) = 25/40 = 0.625

SG vs TTO

Standard Gamble

Represents decision-making under uncertainty
Only valuing the health state
Risk posture is captured (risk aversion for death)
Utility values usually > TTO for same state

Time Trade-Off

Is decision-making under certainty
Might inadvertently capture time preference (i.e., we might value health in the future less than we do today)
Risk posture is NOT captured
Utility values usually < SG for same state

Rating scales

Example:

On a scale where 0 represents death and 100 represents perfect health, what number would you say best describes your health state over the past 2 weeks?“.

Problem:

It does not have the interval property we desire. For example, a value of “90” on this scale is not necessarily twice as good as a value of “45”

Visual Analogue Scale (VAS)

The Visual Analog Scale (VAS) is a commonly-used rating scale

Source: https://assessment-module.yale.edu/im-palliative/visual-analogue-scale

Direct methods – Rating scales

Easy to use: Rating scales often used where time or cognitive ability/literacy prevents use of other methods
Very subjective and prone to more extreme answers! Usually, utilities for VAS < TTO < SG

Indirect methods - EQ-5D

System for describing health states
5 domains: mobility; self-care; usual activities; pain/discomfort; and anxiety/depression
3 levels: 243 distinct health states (e.g. 11223)
Valuations elicited through population based surveys with VAS, TTO

Indirect methods

HUI
Health Utility Index

EQ5D
EuroQol health status measure

SF-6D
Converts SF-36 & SF-12 scores to utilities

QWB
Quality of well-being scale

I’m not going to get into detail for each of these since you don’t need to know the granularity of each right now until you start exploring for your studies in the future, but just as an overview of some of the more well-known ones:
HUI (Health Utility Index), based on a large, Canadian community sample (copyrighted, not free to use), patients are asked to consider their health over the past 2, 3, & 4 weeks
EQ5D, European & US-based, asks participants to consider their health during the day of the interview; free but need to request permission & if you need to translate into other languages then it can be quite pricey
SF-6D, 4 to 1-week recall period, converts the SF-36 & SF-12 (validated, generic quality of life measures) into utility scores, conversion done by University of Sheffield developer John Brazier; published articles on the methodology to convert these
QWB (Quality of Well-Being scale), asks participants to consider their health over the previous three days, primary care patients in San Diego, CA; contact UCSD health outcomes assessment program for access

Alternatives to ICERs

ICERs are the most common approach for describing CEA results

Advantages

Summarizes all aspects of decision problem except WTP (which comes from decision-maker)

Disadvantages

Ratios can act poorly/are unstable; especially when there is a lot of uncertainty in the denominator, the incremental QALYs are small, or close to zero.
For cost-saving interventions, the ICER is meaningless.

An Alternative

Maybe all of your interventions are cost-effective under accepted WTPs & your goal, rather, is to quantify the short and long-term health benefits of interventions.

If willing to choose a fixed willingness-to-pay threshold (e.g., λ= 100,000 / QALY), can write down an equation for the contribution of health and cost to utility.

Net Health Benefit (NHB)

Net Monetary Benefit (NMB)

Objective: Select the strategy with the highest NHB/NMB

Net Health Benefit (NHB)

NHB_s = E_s - \frac{C_s}{λ} E_s is effectiveness of strategy s

C_s is cost of s

λ is WTP threshold

Effectiveness is already in QALYs, no conversion needed
Cost is in $, needs conversion to QALYs (so need to divide by $/QALY)

Net Monetary Benefit (NMB)

NMB_s = E_s * λ - C_s

where

E_s is effectiveness of strategy s

C_s is cost of s

λ is WTP threshold

Cost is already in $, no conversion needed
Effectiveness is in QALYs, needs conversion to $ (multiply by $/QALY)

Example: NHB

E_1 = 0.07 years

E_2 = 0.10 years

C_1 = 1,500

C_2 = 2,800

λ = 50,000 per year of life saved

ICER = \frac{2800 - 1500}{0.10 - 0.07}

=43,333

Example: NHB

NHB_1 = 0.07−\frac{1500}{50000} =0.040

NHB_2 = 0.10−\frac{2800}{50000} =0.044

IncNHB = 0.044−0.040=0.004

Suggests a small difference in QALYs between the interventions (.004*365 = 1.46 days)
The cost adjusted NHB accounts for the tradeoffs between health outcomes and costs into a single measure.

Example: Net Monetary Benefit (NMB)

NMB_1=(0.07×50000)−1500 = 2000

NMB_2=(0.10×50000)−2800=2200

IncNMB =2200−2000=200

Quantifies this improvement’s value in monetary terms.
Treatment 2 provides $200 more net value than Treatment 1 after accounting for both its additional health benefit and higher cost.

NHB/NMB

The strategy with the highest incremental NHB or NMB is the preferred option, given a specified WTP threshold.

A higher NHB means more health gained after accounting for costs.

Key Takeaways

Benefits

Clinical outcomes should be clearly defined and relevant to the decision question.
QALYs combine quantity and quality of life into a single summary measure, enabling comparisons.
Net benefit approaches provide alternatives to ICERs but require an explicit willingness‑to‑pay threshold.

Questions?

Next Lecture: DALYs