Spatio-temporal modelling and mapping of malaria in Angola.
Date
2017
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
About half of the world's population is at risk of contracting malaria. The growing
number of malaria cases and deaths due to this disease in Africa has become a
major challenge to public health care sector. Malaria is reported to be the primary
cause of mortality in Angola, hence major focus needs to be put on intervention
and prevention methods that reduce the disease risk and mortality to a level which
is no longer a public risk problem. The common risk factors for malaria can be
linked to environmental, socio-economic, demographic and climatic factors.
The mortality rate due to malaria in Angola is analyzed using spatial disease mapping
models. Such models are widely used to study the disease incidence, spatial
distribution of diseases, prediction of the disease outcome and also to inform intervention
strategies in various regions across the world. The methodology used
is based on the Bayesian hierarchical modelling (BHM) framework. Four models
namely the Poisson-gamma model, Poisson log-normal model, conditional autoregressive
(CAR) model as well as the convolution model were used to study the
relative risk of malaria mortality at provincial level in Angola using the National
malaria control programme data from the period of 2003 to 2010. The Deviance
Information Criteria selection was applied to compare and select the best fitted
model. A total of 109; 320 deaths due to malaria were observed during the period of
2003-2010 in Angola. The lowest crude death rate was estimated as 124.14 per
100,000 in Lunda Sul province and the highest was 1583.63 per 100,000 in Luanda
province. The results revealed that when comparing the four fitted models, the
convolution model when we fitted to the data with both spatial structured and
unstructured random effects performed better than the other three models. The
structured and unstructured random effects were used to capture variation of risk
specific to a province and across provinces respectively. The risk maps revealed
variation of risk among provinces with very high relative risks in the South-East
parts of Angola. A full Bayesian approach was also applied to perform a spatial and spatio-temporal
modeling of malaria prevalence in Angola among children under 5 years using the
2006-2007 and 2011 Angola malaria indicator survey (AMIS) data. The Bayesian
logistic model was applied in the spatio-temporal analysis to investigate the relationship
between malaria prevalence and some reported socio-economic and demographic
factors for data collected over the years 2006-2007 and 2011. The
space-time effect of the association between malaria and these factors has practical
implications for informing strategies for malaria control. Other than temporal
variation, the risks factors were also found to vary spatially.
The study found that there was a significant difference in the effects of socioeconomic
and demographic variations on malaria between these two time periods.
Wealth has a negative relationship with malaria prevalence while age was found
to have a positive linear relationship with malaria prevalence which is indicative
that this covariate play an important role in contracting malaria. Children living
in urban areas and those who had bed nets were less likely to contract malaria as
compared to those who lived in rural areas and those who did not have bed nets.
The temporal analysis show that the prevalence for malaria was lower in 2011 as
compared to the 2006-2007 period.
Description
Master of Science in Statistics. University of KwaZulu-Natal, Pietermaritzburg 2017.