Biology Reference
In-Depth Information
6.2. Global Endemicity of P. vivax
6.2.1. P. vivax Parasite Rate Data
To map P. vivax malaria endemicity within the limits of stable transmis-
sion, in addition to the updated geographic boundaries and corresponding
populations at risk of P. vivax , an updated georeferenced database of P. vivax
parasite rate ( Pv PR) survey data were compiled. PR was used here because
it is the most ubiquitous of the malariometric measures of risk ( Hay et al.,
2009 ). Pv PR data represent the proportion of a randomly sampled popula-
tion to have detectable P. vivax parasitaemia when screened via microscopy
or RDTs and is the most consistently measured index of malaria endemic-
ity. The final database consisted of unique surveys obtained from published
and unpublished literature sources spanning the period 1985-2010. The
database included information on the survey origin, how the location was
determined (georeferencing method), time period, age group, sample size
and diagnostic method used.
The Pv PR database was made up of 9970 spatiotemporally unique
records from 432 different sources. Data were available from 53 countries,
12 of which were in the Americas, 19 in the Africa+ region, 15 in Asia and
7 in Asia-Pacific. There were 44 Pv MECs not represented in the database,
most of which were in Africa, with the exception of Argentina, Azerbaijan,
Belize, Bhutan, Korea DPR, El Salvador, Georgia, Guyana, Iran, Kyrgyzstan,
Panama, Paraguay, Republic of Korea and Uzbekistan. Details of the PR
data that were input into the model from each region are given below.
6.2.2. Modelling P. vivax Endemicity
To generate a continuous surface of P. vivax endemicity using PR data, a flex-
ible modelling framework based on model-based geostatistics (MBG) ( Diggle
et al., 1998 ; Diggle and Ribeiro, 2007 ) was used. With areas of stable transmis-
sion converted into a 5 × 5 km grid, MBG models allow for endemicity values
to be predicted at each pixel as a function of the geographically varying mean
of survey values and a weighted average of neighbouring data values. MBG
models are well suited for predicting endemicity values, in this case PRs, for a
number of reasons. First, the mean PR values may be defined as a function of
multiple environmental covariates that influence malaria transmission. Second,
a covariance function may be employed to define the spatial heterogeneity of
the PR data and, in turn, define the appropriate weight for each data point
when generating a prediction. Third, uncertainty can be based on the nature
and density of data surrounding a pixel. Fitting MBG models with Bayesian
Search WWH ::




Custom Search