Biology Reference
In-Depth Information
the degree of correlation between a pair of observations depends on the
degree of spatial or temporal separation.
Helminth spatial models typically include covariates, including
remotely sensed environmental and climatic data,
167
often with the aim of
enhancing the predictive accuracy of the fitted model. Bayesian Markov
chain Monte Carlo methods are generally advantageous for fitting such
models since they permit simultaneous estimation of the effects of cova-
riates and spatial clustering.
168
Maximum likelihood methods are more
disjointed, often involving initial fitting of a regression model followed by
an assessment of spatial autocorrelation using residuals.
144,169
Spatial models have been extensively used to construct predictive
prevalence maps of STH and Schistosoma infections at both country
170
and
regional
171,172
levels. For these infections there is an abundance of cross-
sectional egg count-based community prevalence data from locations
across the globe. Furthermore, these data have been collated,
geo-referenced and stored in the Global Atlas of Helminth Infection
database
60
which has greatly facilitated access to a rich source of data to
feed geo-statistical models.
173,174
The model-predicted prevalence can be
used to identify target populations for treatment
167
and, combined with
demographic data, facilitate estimation of the global burden of infection
and morbidity, although estimation of the latter is fraught with uncer-
tainty.
3
The relationship between environmental covariates such as land
surface temperature and precipitation and the developmental and survival
rates of the free-living infective stages of the STH parasites
167
also permits
an understanding of possible distributional changes in infection preva-
lence under a range of climate change scenarios.
175
Spatial modeling approaches have also yielded insight into the effects
that the underlying species-specific biology has on the large-scale distri-
bution of infection. For example, Clements et al.
171
found a decreased
prevalence of A. lumbricoides in rural compared to peri-urban or urban
areas in the Great Lakes region of East Africa, in accordance with the
longstanding perception that transmission of A. lumbricoides and
T. trichiura is more pronounced in densely populated urban locations (in
contrast to hookworm which is more associated with rural settings).
176
Brooker et al.
167
elucidated the relationship between prevalence and land
surface temperature which peaks between 29 and 32
C, before declining
such that in areas where the land surface temperature exceeds 36
37
C
e
5%.
169
Saathoff et al.,
144
working in northern
KwaZulu-Natal, South Africa, found that vegetation density was strongly
associated with a higher prevalence of A. lumbricoides, results in accor-
dance with those of a previous larger scale spatial study in Cameroon.
177
Bayesian geo-statistical models have also been developed to predict
infection intensity (of S. mansoni) by fitting a negative binomial distri-
bution to the data and accounting for their spatial correlation.
178
prevalence is typically
<
