Date of Award

2014

Degree Type

Thesis

Degree Name

Doctor of Philosophy

School

School of Engineering

Faculty

Health, Engineering and Science

First Advisor

Associate Professor Ute Mueller

Second Advisor

Associate Professor Jim Cross

Abstract

Malaria is one of the most prevalent and devastating health problems worldwide. It is a highly endemic disease in Ghana, which poses a major challenge to both the public health and socio-economic development of the country. Major factors accounting for this situation include variability in environmental conditions and lack of prevention services coupled with host of other socio-economic factors. Ghana’s National Malaria Control Programme (NMCP) risk assessment measures have been largely based on household surveys which provided inadequate data for accurate prediction of new incidence cases coupled with frequent incomplete monthly case reports. These raise concerns about annual estimates on the disease burden and also pose serious threats to efficient public health planning including the country’s quest of reducing malaria morbidity and mortality cases by 75% by 2015.

In this thesis, both geostatistical space-time models and time series seasonal autoregressive integrated moving average (SARIMA) predictive models have been studied and applied to the monthly malaria morbidity cases from both district and regional health facilities in Ghana. The study sought to explore the spatio-temporal distributions of the malaria morbidity incidence and to account for the potential influence of climate variability, with particular focus on producing monthly spatial maps, delimiting areas with high risk of morbidity. This was achieved by modelling the morbidity cases as incidence rates, being the number of new reported cases per 100,000 residents, which together with the climatic covariates were considered as realisations of random processes occurring in space and/or time.

The SARIMA models indicated an upward trend of morbidity incidence in the regions with strong seasonal variation which can be explained primarily by the effects of rainfall, temperature and relative humidity in the month preceding incidence of the disease as well as the morbidity incidence in the previous months. The various spacetime ordinary kriging (STOK) models showed varied spatial and temporal distributions of the morbidity incidence rates, which have increased and expanded across the country over the years. The space-time semivariogram models characterising the spatio-temporal continuity of the incidence rates indicated that the occurrence of the malaria morbidity was spatially and temporally correlated within spatial and temporal ranges varying between 30 and 250 km and 6 and 100 months, respectively. The predicted incidence rates were found to be heterogeneous with highly elevated risk at locations near the borders with neighbouring countries in the north and west as well as the central parts towards the east. The spatial maps showed transition of high risk areas from the north-west to the north-east parts with climatic variables contributing to the variations in the number of morbidity cases across the country. The morbidity incidence estimates were found to be higher during the wet season when temperatures were relatively low whilst low incidence rates were observed in the warm weather period during the dry seasons.

In conclusion, the study quantified the malaria morbidity burden in Ghana to produce evidence-based monthly morbidity maps, illustrating the risk patterns of the morbidity of the disease. Increased morbidity risk, delimiting the highest risk areas was also established. This statistical-based modelling approach is important as it allows shortterm prediction of the malaria morbidity incidence in specific regions and districts and also helps support efficient public health planning in the country.

Share

Paper Location

 
COinS