HIV epidemic drivers in South Africa: A model-based evaluation of factors accounting for inter-provincial differences in HIV prevalence and incidence trends

Introduction

South Africa’s HIV epidemic is highly heterogeneous, with population HIV prevalence levels ranging between 5.0% in the Western Cape and 16.9% in KwaZulu-Natal in 2012.¹ In such settings, it has been suggested that policymakers should focus HIV prevention efforts on the regions in which HIV incidence is greatest, in order to make efficient use of limited HIV resources.^2,3,4 It is therefore important to produce robust estimates of provincial HIV prevalence and incidence.

Models may also be required to estimate HIV prevalence at district levels for the purpose of district-level treatment coverage estimation and resource allocation. Although previous studies have estimated district HIV prevalence from antenatal HIV survey data,⁵ antenatal survey data are known to be biased.^6,7,8 It is likely that the extent of the bias differs between districts as a result of differences in the factors that account for the bias (e.g. patterns of health-seeking behaviour, contraception, epidemic stage and age distributions). It is therefore important to assess the extent of differences in antenatal bias between regions so that resource allocation is not unfairly skewed towards those districts in which the antenatal bias is greatest.

Lastly, an understanding of the factors that explain geographical differences in HIV prevalence is important in identifying epidemic drivers, which in turn is important in developing effective HIV prevention strategies. Previous studies have speculated that geographical variation in HIV prevalence within sub-Saharan Africa may be explained by differences in rates of marriage,⁹ male circumcision,^10,11,12,13 migration,¹¹ concurrency^13,14 and other sexually transmitted infections.¹⁵ However, few attempts have been made to identify the factors that account for inter-provincial differences in HIV prevalence in South Africa.

This study aims to estimate HIV incidence and prevalence trends in each of South Africa’s provinces, and to identify the key epidemiological factors that account for these differences.

Methods

The Thembisa model of the South African HIV epidemic was applied to each of the nine provinces in South Africa. Detailed descriptions of the national model¹⁶ and the province-specific adjustments to the national model¹⁷ have been published previously. Briefly, Thembisa is a combined demographic and epidemiological model, which simulates changes in the population profile and HIV disease burden over time, starting in 1985. Demographic estimates of...