Malaria and Water Resource Development - Recent Advances and Issues in Environmental Science - page 167

(prevalence = 9.9%) and those living in “control” communities (prevalence = 5.5%).
The group of children living in “control” communities was further split in children of
age below 5 years (prevalence = 3.5%) and above 5 years (prevalence = 7.5%). Ac-
cording to the overall discriminatory power (i.e., the relative importance) in the CART
analysis, month emerged as the strongest overall discriminating risk factor for a P.
vivax infection (Score (Sc) = 100), followed by village type (Sc = 20.21) and age (Sc
= 8.73) and sex (Sc = 2.19). The classifi cation tree corresponds well with the P. vivax
trends in Figure 1. Indeed, the trends show that in December the P. vivax prevalences
in “at-risk” and “control” communities are both low and that the difference between
“at-risk” and “control” communities are especially clear in October-November.
Figure 3 shows the classifi cation tree for P. falciparum reproduced by CART. Chil-
dren are fi rst split into children living in “at-risk” communities (prevalence = 5.5%)
and those living in “control” communities (prevalence = 2.2%). The group of children
living in “at-risk” communities was further split in children sampled in November (preva-
lence = 2.7%) and those sampled in October and December (prevalence = 6.3%).
According to the overall discriminatory power in the CART analysis, village type
emerged as the strongest overall discriminating risk factor for malaria P. falciparum
infection (Sc = 100), followed by month (Sc = 41.3). The other variables, age, and
sex had a zero-Score. The classifi cation tree corresponds well with the P. falciparum
trends in Figure 1. The trends show that P. falciparum prevalences are lower in “con-
trol” communities and that in “at-risk” communities the prevalences were lower in
November.
Figure 3. Classification tree of the risk factors for P. falciparum infection.
The prevalence fraction (exposed) PrFe, for malaria ( P. vivax and P. falciparum
together), measuring the effect of the dam, was calculated using PR p(D+ IE+)/p(D+
IE-) = 7.7/4.4 = 1.75 and is PrFe = (PR-1)/PR. = 0.43. This means that 43% of the
malaria occurring in children can be attributed to the dam, assuming that the relation-
ship is causal.