The process of extracting latent factors from survey data using factor analysis is used commonly in the psychology literature to diagnose underlying psychological conditions like depression from questionnaires. This research aims to apply the same techniques (namely principal-component factor analysis) to identify the components of risk perception that drive the variation in often seemingly unrelated attitudes of farmers in the Karoo. Multiple analyses on samples extracted from two waves of longitudinal data are conducted in order to verify the variable structure of risk components in the community and Cronbach's alpha is used as a reliability measure to test the degree to which the variables that constitute each extracted component measure a single construct. These tests provide insight into the efficacy with which the survey questions are measuring intended aspects of a population and has implications for survey design. Because the data spans four years in the population under consideration and new variables are included in the second wave of questioning, the effect that new, severe threats have on the structure of risk components are also estimated.
The presentation will reflect on the pragmatic, political, epistemological and methodological possibilities and challenges of conducting collaborative longitudinal qualitative research with families in 21 study communities in South Africa and Zambia in the context of a large-scale cluster randomised HIV prevention trial. Through describing the will describe the design and conduct of the qualitative cohort study nested within this one trial context, the presentation will explore the benefits and limitations of forms of knowledge produced through this approach. Further, it will think through broader possibilities for and implications of integrating collaborative ethnographic and social scientific research into population-based clinical trials and public health research.
Tuberculosis is one of the most common causes of death in the era of HIV/AIDS. According to the World Health Organisation (WHO), South Africa houses third largest global burdens of tuberculosis (TB) after India and China, and 28% of the world’s cases involve HIV-TB co-infection. Recent studies in South Africa indicate that <10% of patients attending clinics for TB-related symptoms were screened for TB and TB screening among people living with HIV is around 80%. Among those tested for TB, the availability of results in patient files is limited. Thus, accurate estimates for rates of TB infection remain unknown, particularly for high-risk populations. Recently, provider-initiated TB case-finding has become an integral part of HIV care in resource-poor settings. However, studies comparing symptom-based screening with sputum tests suggest that current symptom screening has poor sensitivity. This presentation will illustrate the use of latent class analysis to empirically identify distinct patterns of self-reported pulmonary TB symptoms from the second wave of Mzantsi Wakho study – the world’s largest cohort of adolescents living with HIV. In turn, given low rates of TB screening and accurate testing in the South Africa this will help to estimate rates of pulmonary TB infection among this high-risk population as well as explore a simple symptom screening checklist that may be crucial for more effective case finding and follow-up treatment in high-burden, low-resource contexts like South Africa.