Novel Epidemiological tools to inform malaria control and elimination in Melanesia

Abstract

Background: Malaria is a vector-borne parasitic disease that in 2017 was responsible for an estimated 219 million clinical cases of infection and an estimated 435000 deaths globally. Estimating the spatial distribution of malaria within endemic countries, and risk factors for transmission, is essential to the effective planning and allocation of malaria prevention interventions. The aims of this PhD thesis were to: 1) describe the epidemiology of malaria in Papua New Guinea (PNG) and summarise previous control strategies and outcomes in PNG over the last century; 2) compare the accuracy of multilevel generalised linear regression models (GLMs) with Bayesian decision network (BDN) models in the spatial prediction of prevalence of malaria in PNG; 3) predict the geographic niches of eight genotypes of Plasmodium falciparum in PNG to ascertain patterns of connectivity in the human population in terms of malaria transmission; and 4) examine the impact of human movement between high and low transmission intensity locations on malaria transmission using a mathematical model based on the example of two islands of Solomon Islands. Methods: Data for this research was obtained from published literature, a national malaria indicator survey conducted randomly selected villages in PNG in 2010 and 2011, and genotyped malaria indicator survey data collected in PNG and Solomon Islands between 2008 and 2009. Climate data at 1km resolution was obtained from the WorldClim and environmental remote sensing image data were obtained from Earthdata. Modelling approaches included: a comparison of GLMs with BDN models using point prevalence and ecological data to predict the spatial distribution of P. falciparum and P. vivax malaria in PNG; a Dirichlet regression model examining associations of P. falciparum genotype predominance with ecological covariates for the prediction of geographic niches of distinct parasite genotypes in PNG; and a Ross-Macdonald mathematical model using varying estimations of human migration rates and estimates of P. falciparum prevalence in two island of Solomon Islands for the estimation of the impact of human migration on malaria transmission. Results: In terms of P. falciparum and P. vivax spatial distribution in PNG, BDN models were found to have improved accuracy in spatial predictions when compared with generalised linear models. The predicted spatial distribution of P. falciparum and P. vivax based on BDN models followed a similar pattern to survey data with higher predicted prevalence on the islands to the East of PNG and northern coastline of the mainland, and lower predicted prevalence in the highlands and south coast. The results of the Dirichlet regression model identified geographic niches of eight distinct P. falciparum genotypes in PNG based on associations with population density, elevation, distance to the coastline, latitude and longitude, and their quadratic terms. The results of the mathematical model predicted that in the absence of sustained vector control post-elimination, resurgence of malaria may occur relatively quickly in low-transmission intensity locations where connectivity with high-transmission intensity locations exists due to human migration, such as in the islands of Solomon Islands. Conclusions: This PhD research provides a comprehensive review of literature on the control strategies for and challenges to, achieving goal of global malaria elimination, and a review of the current epidemiology of malaria, and major periods of malaria control in PNG. This thesis identifies novel epidemiological methods for improved prediction accuracy in the spatial distribution of malaria based on environmental and climate predictors, a method for inferring human connectivity in terms of malaria transmission in PNG using parasite genotype data and the application of a mathematical model to in examining the transmission dynamics of malaria transmission in two islands in Solomon Islands.