The interplay between environment and sexual health: Investigating the association of climate change and HIV in vulnerable populations of sub-Saharan

HIV and climate change are both long-term adversities which exacerbate poverty, gender inequalities and exploitation of children. Additionally, marginalised individuals such as teenage girls, young women, migrants, sexual and gender minorities, as well as individuals with limited resources are often affected the most. Finally, both phenomena are difficult to tackle due to issues with political engagement and a large and spatially heterogenous distribution. HIV and climate change affect certain regions of the globe at different levels, with sub-Saharan Africa being the region the most heavily affected by both phenomena. As countries strive to achieve new UNAIDS 95-95-95 goals established in December 2020, climate change-related events could jeopardise epidemic control. This can hence drive the synergistic relationship between environment, human behaviours, vulnerabilities and HIV transmission, incidence and severity.
There is very limited statistical evidence on the associations between environment and HIV. There are currently only four existing studies which looked at this relationship. All those studies confirm that environmental changes might have a negative impact on HIV incidence. However, those studies do not provide a comprehensive overview of all the potential variables associated with climate change and HIV. They also focus on only one weather event (such as warm weather, drought or heavy rainfall), mostly focus on regional data and only on adults. The links between climate change and HIV severity remain completely unexplored. The main aim of this project is to identify a potential relationship and understand the complex pathways between climate change and HIV by creating and analysing a range of quantitative model to discover which mediators have the greatest impact on that relationship. The secondary aim is to decide whether vulnerable individuals (such as deprived individuals, women, children) suffer from the impact of climate change on HIV more than the rest of the population at household and regional levels. Based on the proposed aims, a number of research questions will be investigated:
1) How does climate change impact HIV seroconversion rates in sub-Saharan Africa?
2) Is the impact of climate change on HIV seroconversion exacerbated in vulnerable populations in sub-Saharan Africa?
3) How does climate change impact HIV severity in HIV-positive patients in sub-Saharan Africa?
4) Is the impact of climate change on HIV severity in HIV-positive patients exacerbated in vulnerable populations in sub-Saharan Africa?
To answer the research questions, PHIA datasets will be used. They are nationally representative, weighted surveys of HIV incidence in 13 sub-Saharan African countries which can be extrapolated to cases on the national level. They contain the largest numbers of new HIV infections in sub-Saharan Africa to date. Across 13 countries, data from 143,853 households is available. In the datasets, there are 368,373 individuals with HIV status biomarker available. Environmental data will be used from a number of sources: monthly rainfall will be extracted from Climate Hazards group InfraRed Precipitation with Station data and temperature data will be extracted from The Climate Data Store Copernicus data.
Bayesian multilevel modelling will be conducted to assess conditional indirect effects (moderated mediation analysis). The data analytical approach will consist of two steps. The first step will estimate the indirect association between the environmental variables and HIV serostatus/severity (Research Questions 1 and 3) through a number of mediators (such as food insecurity, inadequate drinking water, inadequate living conditions, limited healthcare access, sexual violence, transactional sex, drug abuse and risky sexual behaviours). The second step will involve integrating previously named parameters with the addition of the moderators (gender, age, education) in separate distinct models (Research Qs, 2&4)