Download Estimating the cost impact of NICE policy guidance: a comparison of

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts

Regression analysis wikipedia, lookup

Linear regression wikipedia, lookup

Data assimilation wikipedia, lookup

Transcript
Quantifying uncertainty in
long-term care costs following
introduction of new drug
therapy: the importance of
model choice
Paul McNamee, Alessandra Vanoli,
Debbie Hutchings, Ian McKeith, John
Bond
In the next 25 minutes...
• Aims, methods and results of the paper
• Conclusions and points for discussion
Aims of the paper
• To compare two approaches in modelling the net
NHS costs of drug treatment for Alzheimer’s
Disease
• Application of deterministic and probabilistic
modelling methods to estimate costs and explore
uncertainty about likely “budget impact” to the
NHS over 10 years
• Apply these methods to 3 different policies (ALL,
NG1, NG2)
Some background
• AD affects a large and growing number of older
people
• Drugs are effective at slowing disease progression
• Uncertainty over whether they are ‘cost-effective’
• NICE recommends assessment of “budget impact”
for NHS, but only partially done for AD drugs in
2001
• Recent guidance (March 05) - treatment should no
longer be prescribed for new patients
Methods -1
• Simulate cost impact for AD population
(England and Wales)
– ONS population data
– MRC ageing study for AD prevalence rates
– Systematic review of effects of treatment on
cognition
– Long term prognosis from observational studies
• OLS and GLM regression models estimate
cost effect of changes in disease progression
Methods – 2
Regression models
Ordinary least squares
(OLS)
• Identity link
• Normal (Gaussian)
distribution
Generalized Linear Model
(GLM)
• Log link
• Gamma or Negative
Binomial Distribution
N=371
dependent variable: weekly cost
Constant term allowed.
independent variables: sex, age, ADL, MMSE
Estimated incremental cost impact over
10 yrs 2002/03 prices £m
method Deterministic
based on OLS
pop.
estimates
ALL
550
NG1
NG2
drugs 851
115
drugs 233
298
drugs 449
With extra
clinic visits
NICE 2001
794
N/A
182
Drugs 42 pa
427
N/A
Estimated incremental 10 yr cost per
patient 2002/03 prices £
(% diff from OLS estimate)
method OLS det
OLS prob
GLM det GLM prob
pop.
All cohort
1792
1810
2107
(+18)
1608
(-11)
NG1
597
607
736
(+23)
334
(-45)
NG2
1550
1591
1785
(+15)
1389
(-13)
Conclusions
• Drug treatment for AD likely to lead to
additional costs to the NHS
• Choice of modelling method makes a
difference to magnitude of estimates
Discussion points
• A GLM framework for probabilistic
analysis more likely to produce outliers
• Lack of evidence on long term effects
• Estimates based on 10 year old observations
of costs of care
• Estimates assume full adoption for different
NICE policy decisions – but how realistic?
• How important is the “budget impact”
question to local and national decisionmakers?