Role of Economic Modelling in a Cost

ROLE OF ECONOMIC
MODELLING IN A COSTEFFECTIVENESS ANALYSIS
Evaluating the cost-effectiveness of
antimalarials in South Africa
1
Charlotte Muheki Zikusooka
INTRODUCTION



In a world with limited resources, it is necessary to
have methods of determining the best way to allocate
those resources among alternative uses.
Definition: Economic Evaluation is a technique that
was developed by economists to assist decisionmaking when choices have to be made between
several courses of action.
There are 4 economic evaluation techniques; each
one involves the systematic identification,
measurement and valuation of costs and
consequences / benefits/ outcomes of the
alternatives under review.
2
BACKGROUND – SOUTH AFRICA
 Malaria in South Africa: in 3 out of 9 provinces
 Development of resistance to antimalarials; to
Chloroquine (CQ)& Sulphadoxine-pyremethamine (SP)
 Sky-rocketing malaria cases as a result.
 SP adult dose cost $0.10 compared to AL adult dose cost
$2.50 (at that time)
 Policy makers: need evidence-based information
showing the effectiveness, costs and cost-effectiveness of
a range of alternatives, in order to make choices.
 Traditional economic evaluation approaches become
inadequate in situations where the measurement of the 3
impact of an intervention is methodologically complex.
Ja
n-9
Ma 7
y-9
Se 7
p9
Ja 7
n-9
Ma 8
y-9
Se 8
p9
Ja 8
n-9
Ma 9
y-9
Se 9
p9
Ja 9
n-0
Ma 0
y-0
Se 0
p0
Ja 0
n-0
Ma 1
y-0
Se 1
p0
Ja 1
n-0
Ma 2
y-0
Se 2
p0
Ja 2
n-0
Ma 3
y-0
Se 3
p03
Notified malaria cases
MALARIA TRENDS IN KWAZULU-NATAL
7000
A: DDT reintroduced
6000
5000
B: IRS southern Mozambique
4000
3000
2000
C: Artemether-lumefantrine
implemented
1000
0
4
BACKGROUND
 Extended response to the malaria problem, including:
improved vector control (using new insecticide for
indoor residual spraying) and improved management of
malaria cases (effective antimalarial) and IRS in
neighboring districts in Mozambique
 Dramatic impact on malaria transmission & cases
 To which intervention do attribute the positive impact?
 How do conduct a cost-effectiveness analysis in this
complex situation?
 A range of methods were used, but in this presentation
we focus on modelling, using the decision-tree model
5
ECONOMIC MODELLING

Modelling: a way of representing the complexity of the real world in
a more simple and comprehensible form

In economic evaluation, modelling is “a systematic quantitative
approach for assessing the relative value of one or more decision
options”

The use of models in cost-effectiveness studies has been increasing
over the years despite the concerns and debates

Decision-analytic models used in economic evaluation include,
decision trees, Markov models, Monte Carlo (stochastic)
simulations, and Discrete Event simulations.

A decision tree “graphically depicts components of a decision
problem and relates actions to consequences”

Markov model is a decision-analytic model that involves a Markov
process, i.e. is a modelling technique derived from matrix algebra, 6
which describes the transitions a cohort of patients make among a
number of health states during the series of short or long cycles
Cured
SPcurerate
(DrugCostSP + DxOutptBaseline + OtherCostBaseline) / Cured
Cure
0.6
Treat with SP firstline
(DrugCostSP*2 + DxOutptBaseline*2 + OtherCostBaseline*2) / Cured
Cure
Treat with SP again
Not hospitalised
#
#
pCURE_noFurtherTx
Die
Fail
#
Hospitalised (Quinine)
#
0.9
0.98
(DrugCostSP*2 + DxOutptBaseline*2 + OtherCostBaseline*2 + HospCostBaseline) / Cured
Die
#
(DrugCostSP*2 + DxOutptBaseline*2 + OtherCostBaseline*2 + HospCostBaseline) / Dead
Cure
Malaria treatment at public HF
ALcurerate=0.99
Cured=1
Dead=0.
DrugCostAL=2.850654
DrugCostSP=1.273527
DxOutptBaseline=1.398646
DxOutptPostACT=7.207846
HospCostBaseline=174.93
HospCostPostACT=203.21
OtherCostBaseline=7.2813767
OtherCostPostACT=7.607708
pCURE_noFurtherTx=0.5
pHosp_AL=0.5
pHosp_SP=0.5
SPcurerate=0.12
(DrugCostSP*2 + DxOutptBaseline*2 + OtherCostBaseline*2) / Dead
Cure
#
Fail
(DrugCostSP*2 + DxOutptBaseline*2 + OtherCostBaseline*2) / Cured
Hospitalised (Quinine)
pHosp_SP
0.98
(DrugCostSP + DxOutptBaseline + HospCostBaseline) / Cured
Die
#
(DrugCostSP + DxOutptBaseline + HospCostBaseline) / Dead
Cured
ALcurerate
(DrugCostAL + DxOutptPostACT + OtherCostPostACT) / Cured
Cured
0.91
Treat with AL firstline
(DrugCostAL*2 + DxOutptPostACT*2 + OtherCostPostACT*2) / Cured
Cure
Treat with AL again
Not hospitalised
#
#
pCURE_noFurtherTx
Die
Fail
#
Hospitalised (Quinine)
#
(DrugCostAL*2 + DxOutptPostACT*2 + OtherCostPostACT*2) / Dead
Cure
#
Fail
(DrugCostAL*2 + DxOutptPostACT*2 + OtherCostPostACT*2) / Cured
0.9
0.98
(DrugCostAL*2 + DxOutptPostACT*2 + OtherCostPostACT*2 + HospCostPostACT) / Cured
Die
#
(DrugCostAL*2 + DxOutptPostACT*2 + OtherCostPostACT*2 + HospCostPostACT) / Dead
Cure
Hospitalised (Quinine)
pHosp_AL
0.98
(DrugCostAL + DxOutptPostACT + HospCostPostACT) / Cured
Die
#
(DrugCostAL + DxOutptPostACT + HospCostPostACT) / Dead
7
MARKOV MODEL - ILLUSTRATION
WELL
Uncompli
cated
malaria (S)
Uncompli
cated
malaria
(R)
DEAD
Severe
malaria
8
COST-EFFECTIVES OF ANTIMALARIALS IN S. AFR
 Comparing
SP and ACTs in 2 provinces (KZN & MPM)
A
decision tree was used to evaluate and compare
the relative impact on costs and health outcomes
associated with both uncomplicated and severe
malaria


SP monotherapy & artemether-lumefantrine (in KZN)
SP monotherapy & Artesunate + SP (in MPUM)
 Antimalarial


effectiveness:
KZN:
SP (12%)
MPUM: SP (95%);
AL (99%)
AS+SP (99%)
9
RESULTS: FROM DECISION TREE MODEL
KZN
Cost
Treat with AL
$18.80
Treat with SP
$115.80
Incr Cost
Eff
Incr Eff
C/E
0.9999
$97.00
0.9792
Incr C/E
$18.82
-0.0206
$118.26
(Dominated)
The strategy "Treat with SP" is dominated by "Treat with AL".
MPUM
Cost
Treat with ASP
$12.00
Treat with SP
$24.00
Incr Cost
Eff
Incr Eff
0.99987
$12.00
0.99882
C/E
Incr C/E
$12.05
-0.00105
The strategy "Treat with SP" is dominated by "Treat with ASP".
$24.05
(Dominated)
10
RESULTS
ACTs were clearly more cost-effective relative to SP
monotherapy, in both KZN and MPUM.
Sensitivity analyses of the decision tree model
consistently confirm that ACTs more cost-effective
relative to SP monotherapy even when the values of
different variables are varied over a wide range of
values.
C/E ratios were found sensitive to changes in some of
the variables.
 C/E ratio for the SP option is sensitive to changes in the values of
pHosp_SP (i.e. the probability that a patient who has failed to get
cured with SP (first time) will seek care at a hospital and will be
hospitalised to get second line treatment).
 The lower the value of pHosp_SP, the lower the C/E ratio and vice11
versa for the SP option.
KEY FINDINGS AND LESSONS
Despite being relatively more expensive, ACTs can be
cost-effective.
ACTs were not only more cost-effective than SP
monotherapy, but also resulted in substantial cost
savings in the Kwazulu Natal and Mpumalanga contexts.
The finding that ACTs are cost-saving in Mpumalanga is
particularly important in two ways:

SP monotherapy was still highly effective in Mpumalanga (90%) unlike in
Kwazulu Natal where its effectiveness had declined to only 12%;

There had been no changes in local vector control programme in
Mpumalanga (as was the case in Kwazulu Natal);
Similar findings have been reported by others: Sudre, P.,
J.G. Breman, D. McFarland, and J.P. Koplan (1992); Wilkins, J.J., N. Valentine, and 12
K. Barnes (2002); Institute of Medicine of the National Academies (2004);
Coleman, P.G., C. Morel, S. Shillcutt, C.A. Goodman, and A.J. Mills (2004)
ROLE OF MODELLING IN ECONOMICS
 Health outcomes are normally achievable in the very long run
 analysts need to find a method for extending the evidence of
effectiveness to cover the time interval for which
effectiveness could logically apply
 THUS, the need to use modelling in economic evaluation
remains of critical importance
 HOWEVER: Data needs for economic models are normally
extensive and such data are not easily available, resulting in
the need to rely on assumptions
 ALSO: Limitations with how to present more realistic disease
epidemiology and health system challenges in economic
models
 POLICY MAKERS: the ‘black-box’ effect of some models
13
makes their results less acceptable by policy makers
FINAL REMARKS
 Economist, epidemiologist, public health specialists not
trained in economic/mathematical models.
 Ability to use modelling techniques in analyses either
requires prior training on these techniques (which is not
always feasible) or working with specialists in modelling
(who usually do not have any training in diseases).
 Results of models are not very much trusted and accepted
by policy-makers and other.
 Use of models in disease modelling and economic
evaluations is useful and cannot be completely avoided.
 How can modelling techniques be improved and made
more accessible to non-mathematicians?
14
THANK YOU
FOR
15
YOUR ATTENTION