Thumbnail Image

Sexual network random effects models of migration and spread of HIV and other STIs in South Africa

South Africa is experiencing an explosive epidemic of Human Immunodeficiency Virus (HIV) and of Sexually Transmitted Infections (STI)s. Furthermore, South Africa has extraordinarily high rates of migration. The predominant type of migration is the circular migration in which young men migrate to work in urban areas leaving their sexual partners behind, to whom they return periodically. Conditions of migration bring men into sexual contact with prostitutes and other women at high risk of HIV /STIs. In this way, migrant men form sexual networks, which become a critical bridge for transmitting HIV /STIs between rural and urban areas. The thesis investigates the determinants of HIV and those of STIs, taking into account the migration status and sexual network clustering effect in the data. The data investigated is from cohorts of migrant men from Hlabisa district working in urban areas, non-migrant partner(s) of migrant men residing in Hlabisa district, non-migrant men and their non-migrant partner(s) residing in Hlabisa district in northern KwaZulu-Natal, South Africa. Initially, the expectation-maximization (EM) algorithm is used to estimate parameters of the logistic-mixed model investigating risk factors of STIs. The interval-censored time until HIV infection is investigated using the Cox proportional hazards model which includes sexual network random effects in addition to the fixed effect. The parameters of this model were initially estimated using the EM algorithm. The main parameter estimation was carried out using the Gibbs sampler, a Bayesian Markov chain Monte Carlo (MCMC) method. The results show that migration is a risk factor of HIV /STI. The results further show that age, marital status, age at first sexual intercourse, sexual contact partners, lifetime partners and other biomedical factors are important determinants of HIV /STIs. The study shows that ignoring sexual network random effects in the analysis of HIV /STIs biases the results. The Gibbs sampler is shown to be a plausible alternative to the EM algorithm in the analysis of correlated interval-censored data. It allows full Bayesian inference, which provides a natural framework with which to integrate the uncertainty about parameters and incorporate heterogeneity between sub-groups, without the need to evaluate high-dimensional integrals.
Type of thesis
Zuma, K. (2004). Sexual network random effects models of migration and spread of HIV and other STIs in South Africa (Thesis, Doctor of Philosophy (PhD)). The University of Waikato, Hamilton, New Zealand. Retrieved from https://hdl.handle.net/10289/13234
The University of Waikato
All items in Research Commons are provided for private study and research purposes and are protected by copyright with all rights reserved unless otherwise indicated.