Models and Methods in Health Data Science
Lecture 6: Health Economics and the QALY
Rachel Oughton
20/02/2023
1 Overview
In this lecture, we introduce some key ideas in Health
Economics:
- the Quality-adjusted life-year (QALY)
- Cost-effectiveness
- Probabilistic sensitivity analysis (PSA) for cost-effectiveness
2 Health Economics
Health economics is about maximizing the amount of
health that can be achieved with the money that is
available.
- Economic evaluation has been in use in healthcare systems around the
world since the 1990s
- National Institute for Clinical Excellence (NICE) was formed in 1999
(in the UK).
By economic evaluation we mean a comparison of
different options in terms of their cost and their
consequences.
Health economics
NICE and similar organisations use economic evaluation to
learn about
- pharmaceuticals
- surgical procedures
- diagnostic technologies
- medical devices
- screening programmes
- interventions
- …
in order to best allocate the finite resources available.
2.1 Comparing treatments
In healthcare, comparing treatments for cost-effectiveness
involves
- a health-related objective function (ie. the
outcome you are trying to optimize) and
- a health care budget constraint.
The key questions are:
- How much more does this treatment/option cost than the current
version?
- Is it more effective (and if so, how much more)?
Within a particular disease/condition, this might be
relatively simple to answer.
Example: Cost-effectiveness
Investigating a new asthma drug
- Drug A (current standard) costs £1000 per patient per year
- Drug B (new drug) costs £1200 per patient per year
Evidence from a clinical trial suggests
- Patients on drug B had 30% more episode-free days than those on drug
A.
If we have £200 per person to year to spend, we should opt
for drug B.
Example: cost-effectiveness
But suppose
- we have limited funds
- there’s also a new drug that improves symptoms for sufferers of
another disease.
Now we have new questions:
- How do we compare these new drugs?
- How do we compare the improvements they bring?
- How do we choose which drug to pay for?
- Which spending decision will achieve the ‘most health’?
We need a measure of health that applies to
any possible health-spending decision.
2.2 Measuring Health
The discussion about measuring ‘health’ has been ongoing
since the 1920s.
Following the formation of the NHS
- Life expectancy increased
- Demand for healthcare grew
- More sophisticated (and expensive) treatments became available
- It was impossible to deliver every treatment
2.2.1 The Rosser index
The Rosser index, for evaluating effectiveness of a hospital stay.
From Rosser and Watts (1972)
The possible states (combinations of disability and distress) were
ranked, to create a univariate index.
The Rosser index and Alan Williams
The Rosser index attracted the attention of Prof. Alan Williams, a
health economist who was seconded to the Treasury from 1966-1968
Alan Williams’s idea:
- To focus not on people’s clinical symptoms, but on
how people live
- Seek the views of the ‘general public’, not just medics
- a single number that encompasses both
- the value of additional life expectancy and
- the value of improved quality of life
3 The quality-adjusted life-year (QALY)
The quality-adjusted life-year (QALY) measures
health as a combination of the duration of life and the
health-related quality of life (HRQoL).
- Duration of life is measured in years
- HRQoL is measured on a scale anchored by 0 and 1
- 0 represents the state of death
- 1 is a state of perfect or best-imaginable health.
- It is theoretically possible to have negative values for HRQoL
The quality-adjusted life-year
So, a QALY value of 1 represents one year of perfect health, but it
could also represent
- two years with a HRQoL of 0.5, or
- four years with a HRQoL of 0.25.
See Whitehead and Ali (2010) for some
discussion and analysis around negative QALY values and the QALY
generally.
3.1 Some examples
Expected value of health-related quality of life for patients with
severe angina and left main vessel disease, taken from Williams (1985).
Example: hip replacement
Expected value of health-related quality of life for patients
following a hip replacement (from Fordham et al.
(2012)).
Exercises
Suppose a patient has a chronic condition. With drug A, his
health-related quality of life is 60% of that of a fully healthy person,
and he has a life expectancy of five years (red area).
From the figures below:
- Describe the effect of Drug B in terms of duration and
health-related quality of life (HRQoL).
- Calculate the gain in QALYs if the patient is switched to drug B
(blue/green area).
How are HRQoL values elicited?
The health-related quality of life of a health state is not known or
directly measurable
There are three main methods (broadly speaking)
- Visual analogue scale
- Time trade-off
- Simple gamble
The aim is to order health states by preference, and
to assign HRQoL values to them.
3.2.1 Visual Analogue scale
People are asked to place various health states on a scale from 0 to
1
Simple, but not very accurate:
- Rating states by preferences is difficult
- Scaling bias means people avoid the extremes of the
scale
- People are better at making choices than at assigning
values
3.2.2 Time trade-off
Give the person the choice between
- Living the remainder of their life \(\left(t_{rem}\right)\) in an imperfect
health state, OR
- Living in full health for a shorter time \(\left(t_{full}\right)\).
Vary \(t_{full}\) until the person
is indifferent between the two states. Call this \(t^*_{full}\)
The QALYs must be equal, therefore
\[ 1 \times {t^*_{full}} =
HRQoL\times{t_{rem}} \] and our estimate of HRQoL is \[HRQoL \approx
\frac{t^*_{full}}{t_{rem}}\]
Here, \(HRQoL \approx \frac{8}{10} =
0.8.\)
3.2.3 Standard gamble
Offer the person the choice between:
- Remaining in a particular health state with certainty OR
- Taking a gamble in which you will either die (probability \(p\)), or be in full health (probability
\(1-p\))
The probability \(p\) is varied
until the two options are considered equally desirable (call that
probability \(p^*\)).
The QALYs must be equal, therefore
\[ p^*\times{0} +
\left(1-p^*\right)\times1 = 1\times{HRQoL} \] and our estimate of
HRQoL is \[HRQoL \approx 1-p^*\]
Here, \(p^*=0.3\). This means we
estimate the HRQoL as \(1-0.3=0.7\).
3.2.4 EQ-5D
A measure developed by the EuroQol group
- Widely used by clinical researchers and pharma companies
- Translated into over 200 languages
- Standardized and non-disease-specific
The requirements were for it to be
- An easy ‘add-on’ to studies using existing instruments
- Able to function as a postal questionnaire for self-completion
- Not too demanding (it takes a few minutes to complete)
- Relevant to respondents of all health states
- Capable of producing a single index value
- Consistent with health states judged worse than death (ie. \(HRQoL < 0\))
Dimensions of the EQ-5D
The EQ-5D-5L defines health in terms of five dimensions:
- mobility
- self-care
- usual activities
- pain/discomfort
- anxiety/depression.
Each dimension is divided into five levels (in the EQ-5D-5L):
- No problem
- Slight problem
- Moderate problem
- Severe problem
- Unable to / extreme problems
From this, there are 3125 possible health states.
EQ-5D process
- The respondent answers the 5 dimension questions
- This produces a 5-digit health state summary, eg. 31121
- The respondent places themselves on the visual analogue scale
- “How would you describe your health state today?”
- The 5-digit health state summary can be converted into a single
summary number reflecting the respondent’s HRQoL.
- Uses a formula involving parameters (‘value sets’)
- Different value sets available for different countries/regions
- Value sets usually calibrated using representative samples and TTO
method
Find out more:
EQ-5D and the EuroQol group: past, present and future, Devlin and Brooks (2017)
Comparing
responsiveness of the EQ-5D-5L, EQ-5D-3L and EQ VAS in stroke patients,
Golicki et al. (2015)
3.3 Weaknesses with QALYs
There are many arguments against QALYs, and Whitehead and Ali (2010) give a good
summary.
Some of the main ones are:
- this works with the ‘average’ patient, allowing for no variability
between patients (age, sex, severity of disease, level of deprivation
etc.)
- A QALY gain from 0.1 to 0.3 is equivalent to one from 0.7 to
0.9.
- Takes no account of productivity or activity
- A year of perfect health gained by a 6-year-old has the same value
as one gained by a 30-year-old or a 90-year-old.
- the aim is to maximise the total health of the population -
in theory this could be to the detriment of some.
If you find this interesting, you could look into the
disability adjusted life-year (DALY) as a proposed alternative,
and discounting as a way to address time issues.
4 Cost-effectiveness
Two key quantities in health economics are cost per
QALY and the willingness-to-pay.
The cost per QALY of a treatment is the monetary
cost of that treatment per QALY gained.
So, if a treatment costs £30,000 per person and generates an expected
gain of 2.5 QALYs, the expected cost per QALY is
\[ \frac{30000}{2.5} =
£12000.\]
The willingness-to-pay threshold is the amount of
money a health funder or healthcare provider is willing (or able) to pay
for an increase of one QALY.
In England, NICE has [approximately] a £20,000 cost per QALY
threshold (with some caveats).
The ICER
The incremental cost-effectiveness ratio (ICER) for two
options \(A\) and \(B\) is
\[ICER = \frac{Cost_A - Cost_B}{Outcome_A
- Outcome_B}.\]
The extra cost per extra unit of
health (eg. QALY).
Directly comparable to the willingness-to-pay
threshold.
Notice that if
- \(Cost_A > Cost_B\) and \(Outcome_A < Outcome_B\) (ie. treatment
\(A\) costs more than treatment \(B\) for worse results)
- \(Cost_A < Cost_B\) and \(Outcome_A > Outcome_B\) (ie. if
treatment \(A\) costs less and has a
better effect)
Then \(ICER < 0\). In both of
these cases the outcome is obvious.
4.1 Visualising cost-effectiveness
Incremental net benefit
The incremental net benefit is \[Z = \lambda\left(Outcome_A -
Outcome_B\right)-\left(Cost_A - Cost_B\right)\] where
- \(\lambda\) is the
willingness-to-pay threshold
- \(Z\) is measured in monetary
units.
The larger \(Z\) is, the better our
new treatment \(A\) is judged to
be.
If \(Z<0\) then \(A\) is judged to be worse than \(B\).
5 Example: health behaviour intervention
Kruger et al. (2014) perform a
cost-effectiveness analysis of an online health behaviour intervention
aimed at young people, called “U@Uni”.
- U@Uni focussed on aspects
of life such as smoking, binge-drinking, eating
fruit and vegetables and physical exercise.
- A 6 month trial was performed at the University of Sheffield
- Used to estimate the resulting improvement (if any) to HRQoL.
- Over the 6 month period
- Projected over life times
- Evidence from the trial was combined with
- general population data on health behaviours
- published evidence about the effect of these health behaviours on
mortality
Aim: estimate the short and long term cost effectiveness of
U@Uni, in terms of
cost-per-QALY.
The analysis: main steps:
- Costing analysis to estimate the cost of U@Uni.
- Within-trial analysis to estimate the short-term (6 month)
cost-effectiveness of U@Uni
- Economic modelling analysis to estimate the long-term
(lifelong) cost-effectiveness.
The modelling involved
- Combining individual follow-up data with cross-sectional data on
health behaviours
- Using this information to estimate the long-term effect.
- Using published data linking all four health behaviours to mortality
risk.
5.1 Estimating the cost - within trial
The main cost of the trial was staff time.
- Survey all involved staff about time spent on development and
implementation of U@Uni
- Combined with salary information to estimate the total cost of the
development and implementation.
\[\text{Within trial cost per student} =
\frac{\text{Total development and implementation cost}}{\text{Number of
students in trial}}\]
Estimating the roll-out cost
Rolling U@Uni out to
other universities would have a different cost:
- local development
- implementation
- Potentially more students
\[\text{Rollout cost per student} =
\frac{\text{Cost of any local developement + implementation + monitoring
cost}}{\text{Average number of new students}}\]
5.2 Within trial cost-effectiveness
The starting characteristics of the trial participants are shown
below
5.3 Within trial cost-effectiveness
- Follow-up surveys at 1 month and 6 months
- Cost of healthcare accessed for each participant
- QALYs accrued by each participant were estimated using HRQoL data
collected at the start, 1 month and 6 months.
This data used (as covariates in linear regression) to estimate cost
pp and HRQoL at each time point.
Cost and HRQoL information used to estimate cost per
QALY
5.4 Long term (lifetime) economic modelling
More modelling for long term
- Models of changes to health behaviour over long term
- Models of effect of health behaviours on long term HRQoL
- Markov model to predict death given health behaviours
5.5 Results
Costs
- The cost of development and implementation per person in the
trial was estimated to be £283.29.
- The cost of rolling out U@Uni to other universities was estimated to be £19.16
per person.
QALYs
- Estimated 0.0013 QALYs gained per person within trial
- Estimated long term gain of 0.0128 QALYs.
5.5.1 Cost-effectiveness
Within trial estimate of around £250,000 per
QALY.
Well above willingness to pay of £20k/QALY
Cost-effectiveness: long term and rolled-out
Long term within trial cost per QALY around £22,844
Long term
rolled-out cost per QALY around £1545
If you’d like to read more about the details of the models used, see
Kruger et al. (2014).
6 Probabilistic sensitivity analysis (PSA) for CEA
In this section:
- Uncertainty around our model
- Using probabilistic modelling for cost-effectiveness
- An example: flu vaccination
Uncertainty
In our examples so far we are reliant on various
quantities:
- The costs of interventions / treatments
- The QALYs gained/lost in particular scenarios
- The proportion of people behaving/reacting in certain ways
In reality we are uncertain about these things.
Build a probabilistic model, capturing this
uncertainty with distributions
6.1 PSA: a simple example
We want to model people catching a disease from a population of size
\(N\).
We use a Binomial distribution - \(Bin(N, p)\)
what if we aren’t certain about \(p\)?
We are unlikely to know \(p\)
exactly, so we model our uncertainty.
\[p\sim{Beta\left(\alpha,
\beta\right)}\]
Sampling infection numbers
We can now adopt a two-stage sampling process:
- Sample a value for \(p\) from \(Beta\left(\alpha,\beta\right)\).
- Sample a value for \(x\) from \(Bin\left(N, p\right)\)
Repeat this many times to generate a sample of \(x\) values that reflect our uncertainty
about \(p\).
Parameter uncertainty, not structural
uncertainty
Sampling infection numbers
We have repeated steps 1 and 2 a thousand times, for a population of
100 and \(p\sim{Beta\left(6,34\right)}\).
6.2 Example: Flu vaccine
A simplified version of Turner et al.
(2006) and Baio and Dawid
(2015).
Goal:
Evaluate the cost-effectiveness of a flu vaccination programme.
We will have two cohorts of size \(N\):
- Cohort \(V\), who are offered the
vaccine
- Cohort \(0\), who aren’t offered
the vaccine
Plan:
- Build a model of this scenario
- Simulate the outcomes for both cohorts (many times)
- Perform our CEA.
6.2.1 Vaccination
For the \(V\) cohort, \[N^V_V \sim Bin\left(N, \phi\right).\]
People choose to be vaccinated with probability \(\phi \sim {Beta\left(11.31,
14.44\right)}\)
So \[\begin{align*}
N^V_0 & = N - N^V_V\\
N^0_0 & = N.
\end{align*}\]
Vaccination: cost
Vaccination is associated with a cost \(\psi_V\) per person, which we model by
\(\psi_V \sim logN\left(1.95,
0.0606\right)\).
The total cost due to vaccination is \(\psi_V
{N^V_V}\).
Vaccination: adverse effects
There are sometimes adverse effects from the vaccination, which we
model using
\[\begin{align*}
AE & \sim Bin\left(N^V_V, \beta_{AE}\right) \\
\beta_{AE} & \sim Beta\left(3.5, 31.5\right)
\end{align*}\] and the loss of QALYs per person experiencing
adverse effects is \[\begin{align*}
\omega_{AE} & \sim{logN\left(-0.634, 0.0717\right)}.
\end{align*}\]
The total loss of QALYs due to adverse effects from the vaccine is
therefore \(\omega_{AE}AE\).
6.2.2 Infection
Non-vaccinated people are infected with the flu with probability
\(\beta_I\), so we have
\[\begin{align*}
I^V_0 & \sim{Bin\left(N^V_0, \beta_I\right)}\\
I^0_0 & \sim{Bin\left(N^0_0, \beta_I\right)}.
\end{align*}\]
For people who have been vaccinated, the probability of being
infected is damped by \(\rho_v\), so we
have
\[I^V_V \sim Bin\left(N^V_V,
\beta_I\left(1-\rho_v\right)\right).\]
The total number of people infected with the flu from the \(V\) cohort is \(I^V_{tot} = I^V_V + I^V_0\).
Infection parameters
The parameters \(\beta_I\) and \(\rho_v\) are modelled by
\[\begin{align*}
\beta_I & \sim {Beta \left(13.01, 172.38\right)}\\
\rho_V & \sim{ logN \left(-0.374,0.00524\right)}
\end{align*}\]
GP visits
Of those who are infected, some number visit their GP:
\[\begin{align*}
GP^V & \sim Bin\left(I^V_{tot}, \beta_{GP}\right)\\
GP^0 & \sim Bin\left(I^0, \beta_{GP}\right)
\end{align*}\]
where \(\beta_{GP}\sim{Beta\left(5.8,
13.80\right)}\). We model the cost per GP visit as \(\psi_{GP}\sim{logN\left(3,
0.0606\right)}\).
We calculate the total cost of GP visits as \(C_{GP}^V = GP^V\psi_{GP}\; \text{ and }\;
C_{GP}^0 = GP^0\psi_{GP}.\)
6.2.3 Further complications
Of those who visit the GP
- \(P\) people develop pneumonia
- loss of \(\omega_P\) QALYs per
person (pp)
- \(H\) people are hospitalized
- loss of \(\omega_H\) QALYs pp
- cost of \(\psi_H\) pp
- \(D\) people die
- loss of \(\omega_D\) QALYs pp
These are all modelled in a similar way to the parameters
above.
Total costs and QALYs
We can now calculate the cost pp and QALYs lost pp for each
cohort:
Cohort \(V\)
- \(\text{Cost} =
\frac{1}{N}\left(\psi_VN^V_V + \psi_{GP}GP^V +
\psi_HH^V\right)\)
- \(\text{QALYs lost} =
\frac{1}{N}\left(\omega_{AE}AE^V + \omega_II_{tot}^V + \omega_PP^V +
\omega_HH^V + \omega_DD^V\right)\)
Cohort 0
- \(\text{Cost} =
\frac{1}{N}\left(\psi_{GP}GP^0 + \psi_HH^0\right)\)
- \(\text{QALYs lost} =
\frac{1}{N}\left(\omega_II_{tot}^0 + \omega_PP^0 + \omega_HH^0 +
\omega_DD^0\right)\)
We do this for each simulation.
6.2.4 Results
In the following results, we have performed 1000 simulations with
\(N=100000\).
First of all, we can look at the difference in how many people are
infected.
In cost-effectiveness space
Red line: WTP threshold = £20000 / QALY Blue line: WTP threshold =
£100 / QALY
## Warning: Removed 2 rows containing missing values (`geom_point()`).
Cost-effectiveness acceptability curve
Using the incremental net benefit
\[Z = \lambda\left(Outcome_A -
Outcome_B\right)-\left(Cost_A - Cost_B\right)\]
We can plot the proportion of simulations that conclude the flu
vaccine programme to be cost-effective (ie. for which \(Z>0\)) against \(\lambda\).
The red line is at 0.95.
6.2.5 Simplifications
In our simple example, we have assumed
- The vaccination has no effect on the course of influenza once
infected
- People who don’t visit the GP don’t develop complications
and have ignored
- many possible complications
- factors like age, gender, pre-existing medical conditions
- costs not incurred to the healthcare provider (eg. from time off
work)
7 Summary
In this lecture we have introduced health
economics
- The QALY is a measure of health that combines
duration and health-related quality of
life (HRQoL)
- The QALY allows us to make comparisons of the ‘amount of health’
gained (or lost) by a particular procedure.
- We have looked at cost-effectiveness analysis.
- We can account for parameter uncertainty in our cost-effectiveness
model using probabilistic sensitivity analysis.
Later in the week we will look more at how QALYs are used to make
health economic decisions.
8 References
Baio, Gianluca, and A Philip Dawid. 2015. “Probabilistic
Sensitivity Analysis in Health Economics.” Statistical
Methods in Medical Research 24 (6): 615–34.
Brazier, John, Colin Green, Christopher McCabe, and Katherine Stevens.
2003. “Use of Visual Analog Scales in Economic Evaluation.”
Expert Review of Pharmacoeconomics & Outcomes Research 3
(3): 293–302.
Devlin, Nancy J, and Richard Brooks. 2017. “EQ-5D and the EuroQol
Group: Past, Present and Future.” Applied Health Economics
and Health Policy 15: 127–37.
Fordham, Richard, Jane Skinner, Xia Wang, John Nolan, Exeter Primary
Outcome Study Group, et al. 2012. “The Economic Benefit of Hip
Replacement: A 5-Year Follow-up of Costs and Outcomes in the Exeter
Primary Outcomes Study.” BMJ Open 2 (3): e000752.
Golicki, Dominik, Maciej Niewada, Anna Karlińska, Julia Buczek, Adam
Kobayashi, MF Janssen, and A Simon Pickard. 2015. “Comparing
Responsiveness of the EQ-5D-5L, EQ-5D-3L and EQ VAS in Stroke
Patients.” Quality of Life Research 24: 1555–63.
Kruger, Jen, Alan Brennan, Mark Strong, Chloe Thomas, Paul Norman, and
Tracy Epton. 2014. “The Cost-Effectiveness of a Theory-Based
Online Health Behaviour Intervention for New University Students: An
Economic Evaluation.” BMC Public Health 14 (1): 1–16.
Rosser, RM, and VC Watts. 1972. “The Measurement of Hospital
Output.” International Journal of Epidemiology 1 (4):
361–68.
Turner, DA, AJ Wailoo, NJ Cooper, AJ Sutton, KR Abrams, and KG
Nicholson. 2006. “The Cost-Effectiveness of Influenza Vaccination
of Healthy Adults 50–64 Years of Age.” Vaccine 24 (7):
1035–43.
Whitehead, Sarah J, and Shehzad Ali. 2010. “Health Outcomes in
Economic Evaluation: The QALY and Utilities.” British Medical
Bulletin 96 (1): 5–21.
Williams, Alan. 1985. “Economics of Coronary Artery Bypass
Grafting.” Br Med J (Clin Res Ed) 291 (6491): 326–29.