Page 92 - Contributed Paper Session (CPS) - Volume 3

CPS1952 Michele N. et al.
malaria chemoprevention has been shown to be most effective when delivered
over three months, these maps can be useful for targeting such interventions
(Cairns et al. 2012).
Despite their functionality, the threshold-based maps have several
limitations. Although the environment is a key driver of seasonality, there are
other contributors such as migration (Martinez 2018). The same environmental
factors could also affect different areas differently: rain, for example, can both
create and wash away mosquito breeding sites depending on the local
topology and rainfall intensity (Martinez 2018). Using environmental
thresholds does not allow for other potential drivers or account for the
variation of responses.
Another class of seasonality maps relates to concentration indices. To
quantify the distribution of malaria cases in each district over a year, Mabaso
et al. (2005) used Markham's concentration index which was previously used
to determine rainfall concentrations. Their concentration maps from the case
numbers estimated using a Bayesian spatiotemporal regression model
displayed clearer spatial patterns than those derived from raw case numbers.
Spatiotemporal models smooth out idiosyncratic deviations to enable us to
focus on the main seasonal trend. They are also useful for relating the
seasonality to input covariates and account for unknown spatiotemporal
effects.
In this paper, we present a modelling framework for a cohesive and
evidence-based analysis of malaria seasonality. Using a spatiotemporal
geostatistical model, we obtain maps of various seasonality measures
including the number of transmission periods in a year, as well as start and
end months of each transmission season. Unlike previous work, we also
present the uncertainty associated with each map. A seasonality index from
the rainfall literature is adapted to give a visual impression of both the
distribution and magnitude of malaria cases over a year. The methodology is
illustrated using administrative level data from the Latin America and
Caribbean (LAC) region.

2. Methodology

2.1 Original seasonality index
As suggested by Feng et al. (2013), “how seasonal” a location is can be
expressed as the product of an entropy measure ( ) and the relative

amplitude ( ):

= × ,

81 | I S I W S C 2 0 1 9