Spatial and spatio-temporal modeling and mapping of self-reported health among individuals between the ages of 15-49 years in South Africa.

Abstract

Self-reported health has been commonly used as a measure of individuals health in public health studies. Health presents a complete physical, emotional, and social well-being. It also plays an important role in the development of the country, economically and socially. Poor health still remains a serious problem and it is linked to high burden of diseases in the world. As part of the Healthy People 2020 and Sustainable Development Goals (SDGs) in Sub-Saharan African (SSA), the goals of improving health has not been achieved. Hence, further investigation of the influential factors on health is relevant to improving health inequalities in SSA countries. Disease mapping provides a robust tool to assess geographical variation of disease and has been used in epidemiology and public health studies. The aim of this research is to use two distinct response outcome variables to investigate factors and geographical variations that are associated with self-reported health in South Africa. To accomplish the former and the latter, this research uses data from the National Income Dynamics Study (NIDS). The NIDS datasets are longitudinal data collected every two years from 2008. In this research, several structured additive regression (STAR) models were utilized within a Bayesian methodology, particularly the Bayesian hierarchical models. Models reviewed included Bayesian spatial and spatio-temporal cumulative logit models and logistic regression models, the primary interest was on the conditional autoregressive (CAR) models. Furthermore, the nonlinear effects of individuals age and body mass index (BMI) were part of the research interest. Two applications are discussed; one for the cumulative logit models for the ordinal response, the other for the logistic regression models of the binary response. In the case of the ordinal response, inference was based on the empirical Bayes approach, while for the binary case, a fully Bayesian procedure was used. Similar results were obtained between the two approaches. Findings reveal that age, gender, household income, education, exercising level, alcohol consumption level, smoking, employment, nutrition status, TB, and depression were associated with self-reported health. The BMI was found to have a nonlinear relationship with self-reported health. Also, the findings show that age has a positive linear effect on selfreported health. In addition, the findings reveal significant spatial variation, with higher poor health prevalence in the Siyanda, John Taoli Gaetsewe, Ngaka Modiri Molema, Dr Ruth Segomotsi Mompati, Dr Kenneth Kaunda, Frances Baard, Lejweleputswa, Xhariep, Thabo Mofutsanyane, Fezile Dabi, Mangaung, Chris Hani, Umgungundlovu, Sisonke, Zululand, Umkhanyakude and Gert Sibande districts. Nevertheless, low poor health prevalence was recorded in the West Coast, Cape Winelands, Overberg, Eden, Central Karoo, Uthungulu, iLembe, and eThekwini districts. Interventions to improve individuals health should include addressing of gender inequalities, education, and income inequalities but altogether with employment status and healthy living lifestyle, in particular, targeting districts identified to have highest poor health prevalence.