POLICYFORUM
0.8
Fraction of antimalarial recipients treated in the private sector
0.7
0.6
0.4
0.3
10
1
Annual private sector
antimalarial demand
(millions)
Nigeria
50
100
DRC
0.5
Sudan
Ethiopia
Kenya
Togo
CAR
Ghana
Niger
Chad
Madagascar
Guinea
Malawi
Sierra Leone
Uganda
Burkina Faso
Côte d'Ivoire
Cameroon
Somalia
Rwanda
Zimbabwe
Burundi
Benin
Mali
Tanzania
Mozambique
0.2
Senegal
Gambia
Mauritania
Guinea-Bissau
Djibouti
Swaziland
Angola
Zambia
Gabon
Liberia
Congo
Eq. Guinea
Namibia
0.1
0
0.1
0.2 0.3 0.4
Fraction of private-sector antimalarials for Pf+ patients
0.5
Private-sector patients receive antimalarials for malaria or fever. Distribution of sub-Saharan African countries by the
fraction of antimalarial recipients estimated to be positive for P. falciparum (x axis), the relative importance of the private
sector for febrile treatment (y axis), and the overall size of the private-sector antimalarial market (bubble size). Dotted lines
represent median values. Values for each country are given in the SM, table S3.
by these drugs. Full details of data collection
and analysis can be found in the supplemen-
tary materials (SM). In summary, data on
reported febrile illness in children younger
than 5 years and the fraction of those receiv-
ing antimalarial drugs from the private sector
were assembled from all population-repre-
sentative household surveys conducted since
2000 in malaria-endemic countries in sub-
Saharan Africa for which individual survey
responses were available (n = 96). The com-
bined data set included records on 680,964
children from 43 countries for whom reports
of fever status were recorded; all but two of
these surveys did not include fever or treat-
ment-seeking behaviors for ages older than
5. All surveys used multistage cluster ran-
domized sampling from subnational admin-
istrative divisions, allowing fever prevalence
to be recorded separately at a regional level.
Repeated-measures regression models (19)
were fitted to adjust survey responses for
timing of the survey during the year and to
account for variation over surveyed years
by extrapolating trends to 2013. Literature
reviews were conducted to facilitate extrapo-
lations of fever incidence (fig. S2) and treat-
ment-seeking behavior (fig. S3) to those ≥5
years old, based on surveys recording these
measures for both age groups.
Rates for both age groups were applied to
a high-resolution map of populations across
Africa (20) in order to estimate the number
of individuals receiving an antimalarial drug
for fever annually from the private sector in
each administrative unit as the product of the
population, the annual fever rate, the frac-
tion receiving antimalarials for fever, and the
fraction of antimalarials reportedly received
in the private sector.
The expected number of antimalari-
als received by individuals infected and not
infected with malaria was then estimated by
multiplying the antimalarial demand in each
age group by the prevalence of Plasmodium
falciparum malaria (PfPR) in febrile individ-
uals in that age group for each administra-
tive unit. Malaria prevalence in fevers was
estimated from a PfPR map in the general
population in 2010 (21) adjusted accord-
ing to an empirical relation between popu-
lation and febrile prevalence derived from
household survey data (22). This multiplica-
tion assumed that the decision to seek treat-
ment in the private sector is independent of
true malaria infection status, an assumption
corroborated by nearly identical positivity
rates in those seeking treatment for fever in
the public and private sectors in Tanzania
(23). The inverse of the malaria prevalence
in febrile private-sector treatment
seekers represented the fraction
of antimalarials taken by individ-
uals with nonmalarial fevers. To
capture the uncertainty surround-
ing these estimates and derive an
interquartile range (IQR), values
were drawn from distributions
for each input variable through
10,000 Monte Carlo repetitions
and combined into a distribution
of outcomes. Details on assump-
tions and conversions are provided
in SM (tables S1 to S3 and figs. S1
to S3).
