© Health Economics Group 2016 Page 1 of 3

Incorporating Health Economics into Grant Proposals Health Economics Short Course

For more information and course dates, please visit our website: http://go.unimelb.edu.au/i8ba

Or email us: health-economics@unimelb.edu.au

Modelling Extrapolation and Modelling Longer Term Outcomes and Costs

Centre for Health Policy Melbourne School of Population and Global Health

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

QA

LYs

Months

Standard practice treatment

Extrapolation - Outcomes

Normally the primary outcome is

compared at a particular point in

time for significance

But to find the benefit of an intervention versus standard practice we

need to know the difference at all points in

time

End of follow-

up

We need to work out the area between

these curves

Extrapolation - Outcomes

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

QA

LYs

Months

Standard practice treatment Standard practice (P) treatment (P)

But we don’t just want to know the benefit until the end of the follow-up but also the benefit of the intervention post follow-up

The benefit post follow-up can be as big if not bigger than the within follow-up benefit

End of follow-up

Extrapolation - Costs

Costs are no different to

outcomes – there may be large cost

savings in the difference between

treatment arms post follow-up

0

2000

4000

6000

8000

10000

12000

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Cos

t per

mon

th $

Months

Standard practice treatment Standard practice (P) treatment (P)

So it could be important to extrapolate

Our new treatment might be expensive in the short run but save resources in the long run

© Health Economics Group 2016 Page 2 of 3

Intervention to reduce youth binge drinking?

•  QALYs in the short-term often do not change much •  Most of the health gains are unlikely to be seen for many

years •  So there is a need to estimate/predict them •  We need to make sure we have a way to change short-

term changes in risk factors into long-term changes in health

•  Often if we are improving long-term health this also means we are reducing long-term health expenditure (or at least delaying this expenditure) and so modelling the long-term cost implications of this is also important to determine the most approach treatment decision

Modelling

•  Conceptually extrapolation into the future makes sense… •  …similarly translating intermediate outcomes into final

outcomes •  …but how do we do it?

•  Ideally we want to predict the differences between the new treatment and standard practice for both costs and outcomes until death (for a lifetime)

Modelling Outcomes

•  Extrapolation –  Parametric modelling of survival analysis –  Regression modelling –  ‘Big data’ predictive approaches

•  Translate impact on intermediate outcomes into final outcomes using: –  Estimates from the literature e.g. relative risks on morbidity (QALYs)

and mortality –  Information from other data sources e.g. registries or longitudinal

studies –  Microsimulation modelling – combine evidence from a range of sources

to model life-course

Modelling Outcomes

•  Estimate QALYs for people with Type 2 diabetes, based on profile of complications of each patient over their remaining lifetime

•  Use UKPDS Outcomes Model which is based on an integrated set of parametric proportional hazard models to predict absolute risk of first occurrence of seven major diabetes-related complications, using: – patients’ characteristics –  time varying risk factors (e.g. HbA1c)

10

Simulating lifetime outcomes

Average QALY for UKPDS patients by blood glucose policy

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44

Years (from diagnosis of diabetes)

QA

LY (U

tility

wei

ght)

Intensive Therapy

Conventional therapy

Med

ian

dura

tion

of th

e tr

ial

Minimumfollow-up

Maximumfollow-up

11

Costs are also attached to health events

Benefits are mostly post

follow-up

Small difference

in trial Here the red that we can see represents the lifetime benefit

Modelling Costs

•  Often costs may be driven by health outcomes –  once we have extrapolated outcomes we can link this to costs e.g. cost

of a CVD event such as a stroke

•  Observational data sets are also very useful to estimate the resource implications of particular health outcomes –  predict the costs (healthcare utilisation) associated with different stages

of COPD using routinely collected data sources

•  Recruit an observational sample of those treated in the past (using standard practice) to better understand future cost (& health) implications of the disease

© Health Economics Group 2016 Page 3 of 3

Modelling - Uncertainty

•  Modelling into the unknown often means that there is considerable uncertainty (so we need to make sure we capture and present this)

•  But not predicting into the future means that we have

biased estimates and we are simply ignoring this uncertainty in terms of benefits & costs

•  So when we are designing a study we need to think about data which are going to help us predict into the future