Post on 13-May-2023
The financial and clinical implications of adult malaria diagnosis
using microscopy in Kenya
D. Zurovac1,2, B. A. Larson3,4,5, W. Akhwale6 and R. W. Snow1,2
1 Malaria Public Health and Epidemiology Group, Kenya Medical Research Institute/Wellcome Trust Research Laboratories, Nairobi,Kenya
2 Centre for Tropical Medicine, University of Oxford, John Radcliffe Hospital, Oxford, UK3 Center for Clinical Research, Kenya Medical Research Institute, Nairobi, Kenya4 Center for International Health and Development, Boston University, Boston, MA, USA5 Department of Agricultural and Resource Economics, University of Connecticut, Storrs, CT, USA6 Division of Malaria Control, Ministry of Health, Nairobi, Kenya
Summary objective A recent observational study undertaken at 17 health facilities with microscopy in Kenya
revealed that potential benefits of malaria microscopy are not realized because of irrational clinical
practices and the low accuracy of routine microscopy. Using these data, we modelled financial and
clinical implications of revised clinical practices and improved accuracy of malaria microscopy among
adult outpatients under the artemether–lumefantrine (AL) treatment policy for uncomplicated malaria in
Kenya.
methods The cost of AL, antibiotics and malaria microscopy and the expected number of malaria
diagnosis errors were estimated per 1000 adult outpatients presenting at a facility with microscopy
under three scenarios: (1) current clinical practice and accuracy of microscopy (option A), (2) revised
clinical practice with current accuracy of microscopy (option B) and (3) revised clinical practice with
improved accuracy of microscopy (option C). Revised clinical practice was defined as performing a
blood slide for all febrile adults and prescribing antimalarial treatment only for positive results.
Improved accuracy of routine microscopy was defined as 90% sensitivity and specificity. In the sensi-
tivity analysis, the implications of changes in the cost of drugs and malaria microscopy and changes in
background malaria prevalence were examined for each option.
results The costs of AL, antibiotics and malaria microscopy decreased from $2154 under option A to
$1254 under option B and $892 under option C. Of the cost savings from option C, 72% was from
changes in clinical practice, while 28% was from improvements in the accuracy of microscopy. Com-
pared with 638 malaria overdiagnosis errors per 1000 adults under option A, 375 and 548 fewer
overdiagnosis errors were estimated, respectively, under options B and C. At the same time, the number
of missed malaria diagnoses remained generally low under all options. Sensitivity analysis showed that
both options B and C are robust to a wide range of assumptions on the costs of drugs, costs of blood
slides and malaria prevalence.
conclusions Even with the imperfect microscopy conditions at Kenyan facilities, implementation of
revised clinical practice (option B) would substantially reduce the costs and errors from malaria over-
diagnosis. Additional interventions to improve the accuracy of microscopy (option C) can achieve
further benefits; however, improved microscopy in the absence of revised clinical practice is unlikely to
generate significant cost savings. Revision of guidelines to state explicitly age-specific indications for the
use and interpretation of malaria microscopy is urgently needed. Further prospective studies are required
to evaluate the effectiveness and costs of interventions to improve clinical practice and the accuracy of
malaria microscopy.
keywords malaria, microscopy, cost, diagnosis errors, clinical practice
Tropical Medicine and International Health doi:10.1111/j.1365-3156.2006.01674.x
volume 11 no 8 pp 1185–1194 august 2006
ª 2006 Blackwell Publishing Ltd 1185
Introduction
Current malaria diagnostic and clinical practices in sub-
Saharan Africa result in massive overdiagnosis and over-
treatment of malaria (Amexo et al. 2004; Reyburn et al.
2004; Zurovac et al. 2006). Higher risks of malaria
infection, disease, and death in non-immune children
below 5 years of age (Trape & Rogier 1996; Snow &
Marsh 1998; Snow et al. 2003a) support diagnostic
approaches to treat presumptively all febrile children in
most malaria endemic areas. However, presumptive treat-
ment of older children and adults, which accounts for a
major age-specific share of antimalarial drugs (Snow et al.
2003b), will become less acceptable when significantly
more expensive Artemisinin-based combination therapies
(ACTs) are adopted and implemented as the recommended
first-line therapies for uncomplicated malaria (WHO 2001;
Malenga et al. 2005). Malaria microscopy has been tradi-
tionally seen as one of the potential solutions to increase
diagnostic specificity and overcome the problem of malaria
overdiagnosis (WHO 2000).
However, during our earlier observational study in
Kenya, we demonstrated that potential benefits of micr-
oscopy are not realized among older children and adults
for three main reasons (Zurovac et al. 2006). First, blood
slides were frequently performed for patients without
fever. Secondly, the accuracy of routine microscopy was
low with errors tending towards overreporting of positive
slide results. Thirdly, the results of the routine negative
blood slides, which were correct 92% of the time, were
ignored, and the majority of patients with negative blood
slides were prescribed antimalarials.
As of 2006, Kenya will implement a specific ACT,
artemether–lumefantrine (AL), as the recommended first-
line therapy for uncomplicated malaria. The continuation
of current clinical and laboratory practices with AL will
inevitably waste precious resources. To support imple-
mentation of new Kenyan AL policy, this article analyses
the financial and clinical implications of improving the
accuracy of malaria microscopy and changing common
clinical practices among adult outpatients presenting to a
typical health facility with microscopy.
To undertake this analysis, we used data on adults
(‡15 years) from a recent study evaluating outpatient
malaria case management in patients above 5 years of age
at 17 health facilities with functional microscopy in two
Kenyan districts (Zurovac et al. 2006). The small number
of observations in older children (5–14 years) in our earlier
study did not allow a detailed analysis for this age group.
As adults accounted for 79% of all malaria treatments in
our previous study, demonstrating the financial and clinical
utility of revising clinical practice for adults is an important
first step before considering appropriate policy for older
children. This analysis also provides information that is
directly relevant to the process of revising the National
Guidelines for Diagnosis, Treatment and Prevention of
Malaria for Health Workers in Kenya (MoH 1998).
Methods
Definitions of policy options
We organize our analysis as the evaluation of three policy
options:
• Option A is the continuation of current clinical
practice in an era of AL. This option provides the
benchmark from which other options are evaluated.
• Option B requires a change in clinical practice so that
(i) blood slides are performed for all febrile adults and
no blood slides are performed for non-febrile adults
and (ii) all adults with routine blood slides reported as
negative under current practices are not prescribed AL
and all adults with routine positive blood slides are
prescribed AL.
• Option C is option B combined with a substantial
improvement in the accuracy of malaria microscopy
(from 66.7% sensitivity and 64.7% specificity to 90%
for both).
We do not argue that option C is feasible to implement
in Kenya in the near future, but it provides an alternative
benchmark for understanding the outcomes associated
with options A and B. Implementing option B would
clearly involve interventions to implement revised clinical
practices, while option C would involve interventions to
revise clinical practice and to realize improvements in the
accuracy of routine microscopy. The costs of such inter-
ventions are not included in this analysis because their
content, effectiveness and costs are essentially unknown.
However, the incremental changes in costs and risks of
moving from option A to option B and option C show the
maximum incremental benefits of such investment.
Determination of initial model parameter values
Results from our earlier observational study are summar-
ized as four adult treatment branches in Figure 1, where
branch 1 is for outpatients presenting without a fever who
do not have a blood slide performed; branch 2 is for
outpatients presenting without a fever who have a blood
slide performed; branch 3 is for outpatients presenting with
a fever who do not have a blood slide performed and branch
4 is for outpatients presenting with a fever who have a blood
slide performed. We use the basic information provided in
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
1186 ª 2006 Blackwell Publishing Ltd
Figure 1 to develop the initial model of existing clinical
practice (option A). Figure 2 shows the revision of clinical
practice proposed as option B. Figure 2 also summarizes the
revision of clinical practice envisioned as option C.
Three assumptions for comparing options A, B and C
need to be mentioned here. First, we allow the same
percentage of non-febrile adults without a blood slide to
be prescribed AL (k1 ¼ 0.10) for options B and C as
option A. We do not suggest that this is good clinical
practice, but we acknowledge that some non-febrile
adults are likely to be prescribed AL if either option B or
option C was implemented. Secondly, adults who would
fall into branch 2 for option A are moved to branch 1
for options B and C. Because of this shift, we assume
Adult Outpatients (n)†
% BS Not Done(1–g) = 0.476
% Without Fever(1–p ) = 0.22
% BS Doneg = 0.542
% Rx with ABfor all patientsin Branch 1m
1 = 0.87
% BS Negative(1–spr
2R) = 0.697
% BS Positivespr
2R
= 0.303
% Rx with ALb
2 = 0.652
% Rx with ALa
2 = 1.0
“Branch 1” “Branch 2”
% BS Not Done(1–h) = 0.229
% With Feverp = 0.78
% BS Doneh = 0.771
% Rx with ALl
3 = 0.49
% BS Negative(1–spr
4R) = 0.609
% BS Positivespr
4R
= 0.391
% Rx with ALb
4 = 0.835
% Rx with ALa
4 = 0.957
“Branch 3”“Branch 4”
% Rx with ALl
1 = 0.10
% Rx with ABfor all patientsin Branch 2m
2 = 0.64
% Rx with ABfor all patients in Branch 3m
3 = 0.47
% Rx with ABfor all patientsin Branch 4m
4 = 0.50
Figure 1 The clinical process for adultoutpatients for option A*. *In each box, the
variable is defined, notation for the variable
is specified and the basic values obtained
from Zurovac et al. (2006) are provided.All percentages are reported in decimal
format (i.e. 22% is denoted as 0.22).
�The sample refers to 286 adult outpatients
‡15 years obtained from Zurovac et al.(2006). BS, blood slide; Rx, treatment; AL,
artemether–lumefantrine; AB, antibiotic.
Adult Outpatients (n)
% BS Not Done(1–g) = 1.0
% Without Fever(1–p ) = 0.22
% BS Doneg = 0.0
% Rx with ABfor all patientsin Branch 1m
1 = 0.87
“Branch 1” “Branch 2”
% BS Not Done(1–h) = 0.0
% BS Doneh = 1.0
% BS NegativeOption B: (1–spr
4R) = 0.609
Option C: (1–spr4
R) = 0.802
% BS PositiveOption B: spr
4R
= 0.391
Option C: spr4R
= 0.198
% Rx with ALb
4 = 0.0
“Branch 3” “Branch 4”
% Rx with ALl
1 = 0.10
% Rx with ABfor all patientsin Branch 4m
4 = 0.50
% Rx with ALa
4 = 1.0
% With Feverp = 0.78
Figure 2 The clinical process for adult
outpatients for options B and C with nota-
tion and parameter values*. *In each box,the variable is defined, the notation for the
variable is specified and the basic parameter
assumptions for options B and C are pro-
vided. All percentages are reported in deci-mal format (i.e. 22% is denoted as 0.22).
BS, blood slide; Rx, treatment; AL, artem-
ether–lumefantrine; AB ¼ antibiotic.
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
ª 2006 Blackwell Publishing Ltd 1187
that these adults would also be prescribed antibiotics at
the higher rate l1 ¼ 0.87, when compared with l2 ¼0.64. Thirdly, adults who would fall into branch 3 for
option A are moved to branch 4 for options B and C.
Because of this shift, we assume that these adults would
also be prescribed antibiotics at the rate l4 ¼ 0.50 when
compared with l3 ¼ 0.47.
For ‘costs’, our accounting stance is that of a typical
government health facility, and we assume initially that the
health facility bears the full financial costs of microscopy,
AL and antibiotics. Although patients may pay some
portion of diagnostic and treatment costs, and donor funds
are used by the Kenyan Government to purchase drugs,
issues related to ‘who’ pays which portion of the costs is a
distributional issue that does not change the conclusions of
this article on the total amount of such costs. For notation,
cs, ca and cb represent per adult cost of a blood slide, AL
and antibiotics, respectively. As reported in Table 1,
our initial cost estimates are cs ¼ $0.40, ca ¼ $2.4 and
cb ¼ 0.27. While Table 1 documents how these cost
estimates were obtained, we note here that the basic
conclusions of this analysis are not sensitive to these values.
Modelling expected costs
Using the basic four-branch framework for describing
clinical practice from Figure 1, expected costs are
EðCÞ ¼X4
i¼1
piCi ð1Þ
where costs C in eqn (1) include the cost of blood slides, AL
and antibiotics, i ¼ 1, 2, 3 and 4 are the four branches in
Table 1 Initial parameter assumptions not
included in Figures 1 and 2Description
Modelnotation
Initialvalue
Cost of blood slide per adult outpatient cs 0.40 USD*Cost of artemether–lumefantrine per adult outpatient ca 2.4 USD�Cost of antibiotics per adult outpatient cb 0.27 USD�Expert slide positivity rate for non-febrile adult outpatients§ sprEnf 0
Routine slide positivity rate for non-febrile adult outpatients– sprRnf 0.303
Expert slide positivity rate for febrile adult outpatients§ sprEf 0.122
Assumptions for option A and option BSensitivity** SEN 0.667
Specificity** SPEC 0.647
Routine slide positivity rate for febrile adult outpatients�� sprRf 0.391
Positive predictive value�� PPV 0.208Negative predictive value�� NPV 0.933
Assumptions for option C
Sensitivity SEN 0.900Specificity SPEC 0.900
Routine slide positivity rate for febrile adult outpatients�� sprRf 0.198
Positive predictive value�� PPV 0.556
Negative predictive value�� NPV 0.985
* Gross provider cost per malaria slide, which includes costs of microscope, supplies,training, staff time, supervision and overhead for laboratories (Goodman 1999).
� Based on the WHO agreement with the supplier of AL, the instrument that Kenyan MoH
uses to procure the drug.
� Based on the average cost of adult treatment course, which Kenyan MoH pays for themost commonly prescribed antibiotics in study districts.
§ Based on the 0% and 12.2% expert slide positivity rate among non-febrile and febrile
adult outpatients, respectively, who had routine blood slide performed during our earlier
study.– Based on the 30.3% routine slide positivity rate among non-febrile adult outpatients who
had routine blood slides performed during our earlier study.
** The sensitivity and specificity of routine microscopy was 66.7%, and 64.7%, respect-ively, among adult outpatients who had routine blood slide performed during our earlier
study.
�� Computed from eqns (8) to (10) in the text based on sprEf ¼ 0.122 and associated SEN
and SPEC.
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
1188 ª 2006 Blackwell Publishing Ltd
Figure 1 and pi the probabilities for each branch, with
p1 ¼ (1 ) p)(1 ) g), p2 ¼ (1 ) p)g, p3 ¼ p(1 ) h) and
p4 ¼ ph. Costs Ci for each branch are computed as
Ci ¼ nðkica þ licbÞ for i ¼ 1; 3 ð2Þ
and
Ci ¼ nðcs þ kica þ lic
bÞ for i ¼ 2; 4 ð3Þ
with
ki ¼ 1� sprRi� �
bi þ sprRi ai i ¼ 2; 4 ð4Þ
From (2), costs Ci for branches 1 and 3 are simply the
number of patients prescribed AL and antibiotics (nki andnli, respectively) multiplied by their respective unit costs
(ca and cb). For branches 2 and 4, costs include the addition
of blood slide costs (cs), but the percentage of adults
prescribed AL from (4), ki, is now a function of the routine
slide positivity rate (sprRi ), the percentage of negative slides
that are ignored (bi) and the percentage of positive slides
that are respected (ai).To estimate expected costs for option A, all the
parameter assumptions used to estimated E(C) based on
eqns (1)–(4) are provided in Figure 1 and Table 1. For
options B and C, all the parameter assumptions used to
estimate E(C) are provided in Figure 2 and Table 1.
Modelling diagnosis errors
Errors from overdiagnosis occur because some adults
who are prescribed AL do not have malaria, while
errors from underdiagnosis occur because adults who are
not prescribed AL do have malaria. For any number of
adult outpatients, n, presenting to a health facility, let
E(no) represent the expected number overtreated and
E(nu) represent the expected number undertreated
patients.
The expected number of adults overtreated for malaria is
EðnoÞ ¼X4
i¼1
pinoi ð5Þ
where
noi ¼ nkið1� sprEi Þi ¼ 1; 2; 3 ð6Þ
and
no4 ¼ n 1� sprR4� �
b4NPVþ sprR4 a4ð1� PPVÞ� �
ð7Þ
For non-febrile adults, overtreatment errors noi for
branches 1 and 2 from eqn (6) depend on the number of
adults prescribed AL, nki and the slide positivity rate of
expert microscopy (sprEi Þ. The assumption here is that sprEi
is a best estimate of ‘true’ malaria; so on average 1 � sprEipercentage of adults in branches 1 and 2 do not have
malaria. As reported in Table 1, the expert slide positivity
rate for non-febrile adults (sprEnfÞ is 0%, suggesting that
non-febrile adults do not have malaria. Data do not exist to
estimate separate values for sprE1 and sprE2, so that sprEnf is
used for both branches 1 and 2 (i.e. sprE1 ¼ sprE2 ¼ sprEnfÞto compute eqn (6). For branch 3, eqn (6) uses the expert
slide positivity rate for adults with fever, sprEf to estimate
the number of overtreated adults for branch 3 (i.e.
sprE3 ¼ sprEf Þ.For branch 4, however, as shown in eqn (7), the
number of adults overtreated with AL, no4, depends on
the number of adults treated with routine negative slides
and the number of adults treated with routine positive
slides. From (7), (1 � sprR4 )b4 shows the percentage of
adults with routine negative slides who are prescribed
AL. The negative predictive value (NPV) shows the
percentage of routine negative tests that were also
evaluated as negative by expert microscopists. Thus,
n(1 � sprR4 )b4 NPV shows the number of adults with
routine negative slides who are prescribed AL but do not
have malaria. Similarly, n sprR4 a4ð1 � PPVÞ shows the
number of adults with routine positive slides who are
prescribed AL but do not have malaria. In sum, the term
in brackets [ ] shows the probability of an overtreatment
error for branch 4.
To complete the model and apply it under varying
circumstances, the basic relationships between background
malaria rates, based on sprEf ; and the accuracy of routine
blood slides needs to be incorporated into the model. For
reference, we note here that
sprR4 ¼ sprRf ¼ sprEf SENþ ð1� sprEf Þð1� SPECÞ ð8Þ
PPV ¼ (sprEf =sprRf ÞSEN ð9Þ
NPV ¼ 1� sprEf� �
= 1� sprRf� �� �
SPEC ð10Þ
For fixed sensitivities and specificities of microscopy
practices, SPEC and SEN, eqns (8)–(10) show how sprRf ,
NPV and positive predictive value (PPV) are all logically
linked to the background malaria rate in febrile adult
outpatients (sprEf ). Table 1 provides initial parameter
values for sprEf , SEN and SPEC based on the data for adults
from Zurovac et al. (2006) and the resulting sprRf , NPV
and PPV from eqns (8)–(10). Table 1 also provides the
revised values for sprRf , NPV and PPV for option C, with
90% sensitivity and specificity.
Following similar logic, the expected number of adults
undertreated for malaria is estimated as
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
ª 2006 Blackwell Publishing Ltd 1189
EðnuÞ ¼X4
i¼1
pinui ð11Þ
where
nui ¼ nð1� kiÞsprEi i ¼ 1; 2; 3 ð12Þ
and
nu4 ¼ n 1� sprR4� �
ð1� b4Þð1�NPVÞ þ sprR4 ð1� a4ÞPPV� �
ð13Þ
The model for expected costs and expected errors from
overtreatment and undertreatment of adults, as outlined in
eqns (1)–(13), can be estimated for any parameter values
and assumptions, such as the number of output patients
(n), the percentage of febrile outpatients out of total
outpatients (p), the percentages of non-febrile and febrile
outpatients receiving blood slides (g and h, respectively),
background ‘true’ malaria in febrile and non-febrile adults
(sprEf and sprEnfÞ, the accuracy of routine microscopy (SEN
and SPEC), clinical decisions on respecting test results,
joint treatment with antibiotics and the unit costs of blood
slides, AL and antibiotics prescribed.
Ethical approval
The KEMRI national ethical review committee provided
ethical clearance for the study that generated basic data for
this research (reference no. 681).
Results
Expected costs and diagnosis errors per 1000 adult
outpatients
Because expected costs from eqn (1) and expected diag-
nosis errors from eqns (5) and (11) are linear in the total
number of adult outpatients (n), we estimate expected costs
and expected diagnosis errors for options A, B and C for
1000 adult outpatients arriving at a government health
facility with microscopy. These results can then be multi-
plied by any other number to develop estimates for
multiples of 1000 adults.
As reported in Table 2, the expected cost of option A
based on eqns (1)–(4) and initial parameter assumptions
provided in Figure 1 and Table 1 is estimated to be $2154
per 1000 adult outpatients presenting at a health facility.
The majority of such costs are from AL. For option B,
which again is based on eqns (1)–(4) but with assumptions
provided in Figure 2, expected costs fall to $1254 per 1000
adults. For option C, expected costs fall further to $892 per
1000 adults.
Table 2 also reports expected diagnosis errors for the
three options. For option A, 638 adults are overtreated
with AL, while 17 adults are undertreated. For option B,
the expected number of adults overtreated falls to 264,
while the number undertreated increases to 32. For
option C, given the substantial improvement in the
sensitivity and specificity of microscopy, overtreatment
errors fall to 90 adults and undertreatment errors fall to
10 adults.
For the initial assumptions used here, which are based
on recent information on actual practice in government
health facilities in Kenya and reasonable cost estimates for
per adult treatments, we conclude that option B is clearly
preferred to option A for malaria case management in
Kenya in the coming era of AL as a first-line therapy. The
expected cost of option B is 42% lower than option A,
while the costs of AL in particular fall by 54%.
Overtreatment errors fall by 375 adults per 1000 outpa-
tients, although undertreatment errors increase by 14
adults per 1000. While undertreatment errors increase
somewhat, as long as the risks from undertreatment are
no more than 26.7 times as large as the risks from
overtreatment, the total risks from option B are less than
that for option A.
Table 2 Differences in expected costs and diagnosis errors between policy options: results per 1000 adult outpatients
Option A Option B Option C
% cost difference,
option A to B
% cost difference,
option A to C
Expected costs – total $2154 $1254 $892 )42% )59%Expected cost of blood slides $287 $312 $312 9% 9%
Expected costs of AL $1719 $785 $422 )54% )75%Expected costs of antibiotics $148 $156 $156 6% 6%Diagnosis errors Change in errors,
option A to B
Change in errors,
option A to C
Expected number overtreated 638 264 90 )375* )548*Expected number undertreated 17 32 10 14* )8*
* Numbers do not add exactly because of rounding errors.
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
1190 ª 2006 Blackwell Publishing Ltd
Option C is obviously better than both options A and B.
Expected costs are 59% less for option C compared with
option A, and the costs of AL are 75% less than that for
option A. Diagnosis errors fall substantially, with errors
from overtreatment falling to just 90 per 1000 adult
outpatients and errors from undertreatment falling to 10
per 1000 adult outpatients. Of the expected total cost
savings from option C, Table 2 shows that 72% of such
savings is from the changes in clinical practice, while the
remaining 28% is from the improvements in the accuracy
of microscopy. Diagnosis errors are less for option C, with
a reduction in underdiagnosis errors of 22 adults per 1000
when moving from option B to option C, suggesting that
improvements in diagnostic accuracy are obviously useful.
Sensitivity analysis
The model developed here in eqns (1)–(14) is applicable to
any situation defined by costs, clinical practices and
accuracy of microscopy. While the initial analysis was
based on actual data on clinical practices and microscopy
accuracy found in the two districts in Kenya studied in
Zurovac et al. (2006) and reasonable estimates of costs and
the joint prescribing of antibiotics summarized in Figures 1
and 2, the model can be organized into a simple spread-
sheet to consider if our basic analysis is sensitive to any of
the assumptions.
Table 3 shows that the conclusions of our analysis from
Table 2 are robust to assumptions related to costs of blood
slides, AL or antibiotics. As shown in Table 3, increasing
substantially the cost of blood slides or antibiotics raises
costs for all options, but option A remains significantly
higher. Clearly, if AL becomes very inexpensive, say $0.50
per adult treated as shown in Table 3, costs of all options
fall dramatically although options B and C still reduce
substantially the number of adults overtreated. Similarly, if
the cost of blood slides or other diagnostic tools increases
substantially, the costs of all options increase but option B
remains preferred to option A. The same conclusion holds
if AL costs fall and blood slide costs increase simulta-
neously.
Table 3 also shows the impacts on costs and diagnosis
errors if this analysis is applied to areas with either lower
or higher background prevalence of uncomplicated malaria
in adult outpatients. In areas with lower prevalence, the
costs of treatment fall somewhat for all options and fewer
undertreatment errors occur. In areas of higher prevalence,
such as 25%, as evaluated in Part B of Table 3, options B
and C remain substantially less costly than option A, with
significantly fewer errors from overdiagnosis. The number
of missed diagnoses in adults increases, as background
prevalence increases to 25%, with an additional 30 adults
are underdiagnosed for option B when compared with
option A. The additional risks to these 30 adults per
Table 3 Sensitivity of results to basic
assumptions: results per 1000 adultoutpatients
Part A: Sensitivity of results to cost assumptions
Costs
Initial
assumptions
Cost of ablood slide
increases from
$0.4 to $1.00
Cost of AL
falls from
$2.40 to $0.5
Cost of antibiotics
increases from
$0.27 to $1.00
Option A $2154 $2584 $793 $2554
Option B $1254 $1722 $632 $1678
Option C $892 $1360 $557 $1315
Part B: Implications of background malaria prevalence
Prevalence ¼ 6% Prevalence ¼ 12.2% Prevalence ¼ 25%
Costs
Option A $2150 $2154 $2161Option B $1218 $1254 $1329
Option C $799 $892 $1083
Diagnosis errors
Expected number overtreatedOption A 676 638 559
Option B 281 264 229
Option C 95 90 81Expected number missed
Option A 8 17 35
Option B 16 32 65
Option C 5 10 20
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
ª 2006 Blackwell Publishing Ltd 1191
1000 outpatients are likely to be small in malaria endemic
areas.
Discussion
Given the imperfect microscopy conditions at Kenyan
health facilities, our analysis shows that implementing the
two basic changes in clinical practice as outlined for option
B – only perform blood slides for febrile adults and respect
the results of the tests – would substantially reduce
expected treatment costs, AL drug costs and errors from
overdiagnosis of adult outpatients presenting to a health
facility with functional microscopy when compared with
current practice. While undertreatment errors would
increase by an estimated 14 per 1000 adult outpatients,
because the risks of malaria disease are low in adults, the
benefits of option B based on substantial treatment cost
savings and reduction in overtreatment errors far outweigh
risks associated with a minor absolute increase in under-
treatment errors. Substantial improvements in the accuracy
of microscopy when combined with changes in clinical
practice (option C) would further reduce expected treat-
ment costs and diagnosis errors when compared with
current practice. However, a substantial improvement in
the sensitivity and specificity of routine microscopy, in the
absence of revised clinical practice, does little to reduce
treatment costs or overtreatment errors.
Sensitivity analysis show that both options B and C are
robust to a wide range of assumptions on the costs of
drugs, costs of blood slides and malaria prevalence.
Moreover, successful implementation of either option B or
option C can generate substantial aggregate cost savings on
AL drugs used for adult patients within the government
health sector in Kenya. While functional microscopic
services exist at 24% of all government facilities, 40% of
all adult outpatient visits take place in these facilities
(D. Zurovac & J. Ngigi, unpublished data). If we assume
that these 40% of the adult outpatient visits account for
40% of total treatment costs for adults, implementing only
the changes in clinical practice envisioned as option B
would reduce treatment costs for adults by 17% (24% for
option C) and the total cost of AL drugs prescribed for
adults would fall by 22% (30% for option C). These
figures are probably conservative because cost-per-outpa-
tient visit at facilities with microscopy, which are usually
larger health facilities such as health centres and hospitals,
are likely to cost more than visits to rural dispensaries.
The successful implementation of either option B or
option C would also not alter the overall workload of
laboratories. From Figure 1, our data suggest that 72% of
all adults coming for an initial visit already have a blood
slide performed, and the change in practice envisioned for
option B would require that 78% of adults have malaria
microscopy. This 6% increase in the number of blood
slides is unlikely to burden the capacity of laboratories to
perform blood slides.
Consistent with the findings reported in Malawi
(Jonkman et al. 1995), we also doubt that the reduction in
use of antimalarials associated with either option B or
option C would substantially increase the prescription of
symptomatic and other antimicrobial treatments. Any
additional cost of such prescriptions is unlikely to eliminate
the treatment cost savings from improved clinical practices.
The majority of outpatients in our study sites were
prescribed inexpensive antipyretics and antibiotics anyway,
and we included such drug substitution in our analysis, as
summarized in Figures 1 and 2. As a related issue, patients
(or health workers) may perceive that the prescription of
medications is a signal for good quality of care. As most
adult outpatients would still leave a health facility with
antipyretic and antibiotic prescription, this issue is unlikely
to be of major importance. In most cases, the main
difference to current practices would be an omission of
antimalarial treatment from the existing polypharmacy
prescriptions.
The analysis in this article does not focus on the
interventions that would be needed, and the costs associ-
ated with such interventions, to implement option B or C.
Both options B and C require that health workers change
their current clinical practices. The major change required
is that health workers request a blood slide only for febrile
adults and prescribe AL only for patients with positive test
results. A basic pre-requisite for implementing these
changes in clinical practice is that clinical guidelines
explicitly state age-specific indications for the use of
malaria microscopy and provide instructions as to how
blood slide results should be interpreted. The existing
Kenyan national malaria guideline, which is currently
under revision, does not provide these instructions (MoH
1998). The results of our analysis suggests that, even with
the imperfect microscopy conditions found in government
health facilities, the changes in clinical practices incorpor-
ated into option B for adult patients can be safely
incorporated into national guidelines for AL therapy in
Kenya.
Once clear guidelines exist, however, changing practices
so that health workers adhere to such guidelines will
clearly present challenges. Jonkman et al. (1995) is the
only study in Africa that demonstrated a significant
reduction of unnecessary antimalarial prescriptions in
adults after the introduction of microscopy. This study,
however, was undertaken in a specific setting of an
outpatient department of a referral hospital in Malawi,
where health workers were required to prescribe
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
1192 ª 2006 Blackwell Publishing Ltd
antimalarial drugs only to parasitaemic patients (Jonkman
et al. 1995). Larger studies to evaluate the costs and
effectiveness of potential interventions to improve adher-
ence to unambiguous guidelines under operational condi-
tions are urgently needed to provide programmatic
evidence on the factors influencing case management
practices using malaria microscopy.
Finally, several challenges can be expected in the process
of improving the accuracy of malaria microscopy. The
target specified in option C to achieve 90% sensitivity and
specificity of routine microscopy is likely to require a series
of laboratory focused interventions such as retraining of
laboratory technicians, strengthening of the supervision
and establishing currently non-existent quality control
systems. As with the clinical practices, studies evaluating
the costs and effectiveness of various programmatic inter-
ventions to improve and maintain high accuracy of routine
malaria microscopy are urgently needed. At the same time,
investments to improve microscopy needs to be evaluated
in the wider context of multiple diseases for which
microscopy is used as well as new technologies such as
rapid tests for diagnosing malaria.
Acknowledgements
This study received financial support from the Roll Back
Malaria Initiative, AFRO (AFRO/WHO/RBM # AF/ICP/
CPC/400/XA/00), The Wellcome Trust, UK and the Kenya
Medical Research Institute. RWS is a Senior Wellcome
Trust Fellow (#058992). We thank Dr Mike English and
Mathew Fox for comments on the article and Lydia
Mwangi and Lucy Muhunyo for their assistance in data
handling as well as to all health workers and patients who
participated in the original study. This article is published
with the permission of the director KEMRI.
References
Amexo M, Tolhurst R, Barnish G & Bates I (2004) Malaria
misdiagnosis: effects on the poor and vulnerable. Lancet 364,
1896–1898.
Goodman C (1999) The economic evaluation of malaria diagnosis.
Working Paper Prepared for Informal Consultation on ‘‘Malaria
diagnostics at the turn of the century’’, organized by WHO and
USAID, Geneva, 25–27 October 1999.
Jonkman A, Chibwe RA, Khoromana CO et al. (1995) Cost-
saving through microscopy-based versus presumptive diagnosis
of malaria in adult outpatients in Malawi. Bulletin of the
World Health Organisation 73, 223–227.
Malenga G, Palmer A, Staedke S et al. (2005) Antimalarial treat-
ment with artemisinin combination therapy in Africa. British
Medical Journal 331, 706–707.
Ministry of Health (MoH) (1998) National Guidelines for Diag-
nosis, Treatment and Prevention of Malaria for Health Workers.
Ministry of Health, Nairobi, Kenya.
Reyburn H, Mbatia R, Drakeley C et al. (2004) Overdiagnosis of
malaria in patients with severe febrile illness in Tanzania: a
prospective study. British Medical Journal 329, 1212–1215.
Snow RW & Marsh K (1998) New insights into the epidemiology
of malaria relevant for disease control. British Medical Bulletin
54, 293–309.
Snow RW, Craig MH, Newton CRJC & Steketee RW (August
2003a) The public health burden of Plasmodium falciparum
malaria in Africa: deriving the numbers. Working Paper 11,
Disease Control Priorities Project. Fogarty International Center,
National Institutes of Health, Bethesda, Maryland. http://
www.fic.nih.gov/dcpp.
Snow RW, Eckert E & Teklehaimanot A (2003b) Estimating the
needs for artesunate-based combination therapy for malaria
case management in Africa. Trends in Parasitology 19, 363–
369.
Trape JF & Rogier C (1996) Combating malaria morbidity and
mortality by reducing transmission. Parasitology Today 12,
236–240.
World Health Organization (WHO) (2000) New Perspectives:
Malaria Diagnosis. Report of a Joint WHO/USAID Informal
Consultation 25–27 October 1999. (WHO/CDS/2000.14).
WHO, Geneva.
World Health Organization (WHO) (2001) Antimalarial Drug
Combination Therapy. Report of a technical consultation.
(WHO/CDS/RBM/2001.35). WHO, Geneva.
Zurovac D, Midia B, Ochola SA, English M & Snow RW (2006)
Microscopy and outpatient malaria case management among
older children and adults in Kenya. Tropical Medicine and
International Health 11, 1–9.
Corresponding Author Dejan Zurovac, Malaria Public Health & Epidemiology Group, Centre for Geographic Medicine, Kenya
Medical Research Institute/Wellcome Trust Research Laboratories, P.O. Box 43640, 00100 GPO, Nairobi, Kenya. Tel.:
+254 20 2720163; Fax: +254 20 2711673; E-mail: dzurovac@wtnairobi.mimcom.net
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
ª 2006 Blackwell Publishing Ltd 1193
Implications financieres et cliniques du diagnostic microscopique de la malaria chez l’adulte au Kenya
donnees de base Une etude d’observation menee sur la microscopie dans 17 services de sante au Kenya a revele que les benefices potentiels de la
microscopie de la malaria n’etaient pas realises a cause de pratiques cliniques irrationnelles et de la faible precision de la microscopie de routine. Sur base
de ces donnees, nous avons modelise les implications financieres et cliniques de la correction de la pratique clinique et de l’amelioration de la precision de
la microscopie de la malaria chez les patients ambulants adultes sous les directives de traitement utilisant artemether-lumefantrine pour la malaria non
compliquee au Kenya.
methodes Les couts de artemether-lumefantrine, des antibiotiques, de la microscopie de la malaria et du nombre attendu d’erreurs de diagnostic ont
ete estimes par 1000 patients ambulants adultes se presentant dans un service pratiquant la microscopie sous trois scenarios: 1) Option A: pratique
clinique en cours et precision de la microscopie, 2) Option B: pratique clinique corrigee avec la precision microscopique en cours et 3) Option C:
pratique clinique corrigee avec amelioration de la precision de la microscopie. La correction de la pratique clinique a consiste a prescrire la pratique d’un
frottis sanguin pour tout cas febrile adulte et la prescription d’un traitement antimalarique seulement pour les resultats positifs. L’amelioration de la
precision de la microscopie de routine a ete definie pour 90% de specificite et de sensibilite. Dans l’analyse de sensibilite, les implications dans les
changements du cout des medicaments et de la microscopie de la malaria et les changements sous-jacents dans la prevalence de la malaria, ont ete
examinees pour chacune des options.
resultats Le cout de l’artemether-lumefantrine, des antibiotiques et de la microscopie de la malaria ont baisse de 2154$ pour l’option A a 1254 $ pour
l’option B et a 892 $ pour l’option C. 72% des couts evites dans l’option C s’appliquaient aux changements de la pratique clinique et 28% a
l’amelioration de la precision de la microscopie. Sous l’option A, 638 erreurs (faux positifs) de diagnostic par 1000 adultes ont ete effectuees. 375 erreurs
de diagnostic en moins ont ete estimees sous l’option B et 548 en moins sous l’option C. Le nombre de malaria non diagnostiquees restait en general
faible quelque soit l’option. L’analyse de sensibilite a revele que les options B et C etaient toutes les deux robustes pour une large variete d’assomptions
sur le cout des medicaments, des frottis de sang et de la prevalence de la malaria.
conclusion Meme avec des conditions imparfaites de microscopie dans les services de sante au Kenya, l’implementation de pratique clinique corrigee
(option B) reduirait substantiellement les couts et les erreurs de diagnostic en surplus de la malaria. Des interventions supplementaires dans l’ameli-
oration de la precision de la microscopie (option C) peuvent permettre des benefices supplementaires, quoiqu’il soit improbable que l’amelioration de la
microscopie en l’absence de pratique clinique corrigee permette de reduire les couts. Des directives expliquant clairement les indications specifiques a
l’age pour l’utilisation et l’interpretation de la microscopie de la malaria devraient etre urgemment corrigees.
mots cles malaria, microscopie, cout, erreurs de diagnostic, pratique clinique
Las implicaciones financieras y clınicas del diagnostico microscopico de malaria en adultos en Kenia
antecedentes Durante un estudio observacional realizado recientemente en 17 centros sanitarios con microscopıa en Kenia, se demostro que los
beneficios potenciales de la microscopia de malaria no se estaban dando debido a una practica clınica irracional y a una baja precision en la microscopıa
de rutina. Utilizando estos datos, modelamos las implicaciones clınicas y financieras de unas practicas clınicas revisadas y un aumento en la precision de
la microscopia para malaria, entre adultos que acudıan a consultas externas por polıtica de tratamiento con artemeter-lumefantrina para malaria no
complicada en Kenia.
metodos Se estimaron el coste de artemeter-lumefantrina, de los antibioticos y la microscopia para malaria ası como el numero esperado de errores en
el diagnostico de malaria por cada 1,000 pacientes adultos que se presentaban en un servicio sanitario con microscopio bajo tres escenarios:1) practica
clınica y precision de la microscopıa actuales (Opcion A); 2) practica clınica revisada y precision de la microscopıa actual (Opcion B); y 3) practica
clınica revisada con mejora en la precision de la microscopıa (Opcion C). Se definio la practica clınica revisada como el hacer una gota gruesa a todos los
adultos febriles y prescribirles tratamiento antimalarico solo a los que obtuviesen un resultado positivo. Una mejora en la precision de la microscopıa de
rutina se definio como una sensibilidad y especificidad del 90%. En el analisis de sensibilidad se examinaron para cada opcion las implicaciones de
cambio en el coste de los medicamentos y la microscopıa de malaria ası como los cambios en la prevalencia de la malaria de fondo.
resultados El coste de artemeter-lumefantrina, antibioticos y la microscopia de malaria disminuyo de $2,154 bajo la Opcion A a $1,254 bajo la
Opcion By $892 bajo la Opcion C. De los ahorros en costes de la Opcion C, un 72% correspondıa a cambios en las practicas clınicas y 28% era de
mejoras la precision de la microscopıa. Bajo la Opcion A se sobrediagnosticaron como malaria 638 de cada 1,000 adultos; en la Opcion B y C se
estimaron 375 y 548 menos errores de sobrediagnostico respectivamente. El numero de diagnosticos de malaria omitidos fue en general bajo en todas las
opciones. El analisis de sensibilidad mostro que ambas opciones By C eran robustas a un amplio rango de supuestos sobre el coste de los farmacos, el
coste de las gotas gruesas y la prevalencia de malaria.
conclusiones Aun con las condiciones imperfectas de microscopıa existentes en las instalaciones de Kenia, la implementacion de una practica clınica
revisada (Opcion B) reducirıa sustancialmente los costos y errores debido al sobrediagnostico de la malaria. Intervenciones adicionales para mejorar la
precision de la microscopıa (Opcion C) podrıan alcanzar mayores beneficios; aunque la mejora de la microscopıa en ausencia de una practica clınica
revisada es poco probable que genere ahorros significativos en los costos. Es urgente revisar las directrices para que expongan explıcitamente las
indicaciones, especificadas por edad, del uso y la interpretacion de la microscopıa para malaria.
palabras clave malaria, microscopıa, costo, error de diagnostico, practica clınica
Tropical Medicine and International Health volume 11 no 8 pp 1185–1194 august 2006
D. Zurovac et al. Financial and clinical implications of malaria microscopy
1194 ª 2006 Blackwell Publishing Ltd