Malaria and Water Resource Development - Recent Advances and Issues in Environmental Science - page 163

Classification Tree
To investigate the potential complex interactions between the different determinants
in explaining the presence of the parasite, classification trees (CART) were used [34].
This technique can be used to investigate how the available determinants can be used
in creating homogenous subgroups, with either high or low prevalence's. The CART
models are fitted by binary recursive partitioning of a multidimensional covariate
space, in which the dataset is successively split into increasingly homogeneous sub-
sets until a specified criterion is stratified. The minimum error tree was selected. The
CART provides a predictor ranking (variable importance) based on the contribution
predictors make to the construction of the tree. This indicates how important the dif-
ferent independent variables are in determining the division. Importance is determined
by playing a role in the tree, either as a main splitter or as a surrogate. Variable im-
portance, for a particular predictor, is the sum across all nodes in the tree of the im-
provement scores that the predictor has when it acts as a primary or surrogate splitter.
It is thus possible that a variable enters the tree as the top surrogate splitter in many
nodes, but never as the primary splitter. Such a surrogate splitter will turn out as very
important in the variable importance ranking provided by CART. More details on this
technique can be found in [35].
Prevalence Fraction
The cross-sectional study allows to compute a prevalence ratio (PR) which is com-
puted as follows: p(D+ IE+)/p(D+ IE-) with p a probability, D+: positive case, E+:
living close to the dam and E-: living away from the dam. The “prevalence fraction
(exposed)” was calculated using the relation that PrFe = (PR-1)/PR. The PrFe ex-
presses the proportion of disease in exposed individuals that is due to the exposure,
assuming that the relationship is causal. Alternatively, the indicator can be viewed as
the proportion of infections in the exposed group that would be avoided if the exposure
were removed.
Ethical Considerations
Ethical approval for this study was obtained from Jimma University Research and
Ethics Committee. Communal consent was first obtained then children were recruited
after informed oral consent was sought from their parents or guardians of each child
before a child was enrolled in the study.
Of the 1,855 children below the age of 10 years examined in this study, 905
(48.8%) were girls and 950 (51.2%) were boys. The mean age of children was 4.7
years and the number of children surveyed from “at-risk” and “control” communities
was 1,081 (58.3%) and 774 (41.7%), respectively. Of the children in “at-risk” com-
munities, 528 (48.8%) were boys and 553 (51.2%) girls while in “control” communi-
ties, 377 (48.7%) were boys and 397 (51.3%) were girls. Overall, 194 (10.5%) of the
sampled children were positive for malaria, of which, 117 (60.3%) were positive for P.
vivax , 76 (39.2%) for P. falciparum and one (0.5%) for both P. vivax and P. falciparum .
Highest P. vivax (60.7%) and P. falciparum (57.9%) positivity rates were observed
in October. The P. vivax prevalence varied from 5.9% in children<1year of age to