Temporal Trends and Determinants of HIV Testing at Antenatal Care in Sub-Saharan Africa: A Pooled Analysis of Population-Based Surveys (2005-2021)

Abstract

Background: In sub-Saharan Africa (SSA), integrating HIV testing into antenatal care (ANC) has been crucial toward reducing mother-to-child transmission of HIV. With the introduction of new testing modalities, we explored temporal trends in HIV testing within and outside of ANC and identified sociodemographic determinants of testing during ANC.

Methods: We analyzed data from 139 nationally representative household surveys conducted between 2005 and 2021, including more than 2.2 million women aged 15-49 years in 41 SSA countries. We extracted data on women's recent HIV testing history (<24 months), by modality (ie, at ANC versus outside of ANC) and sociodemographic variables (ie, age, socioeconomic status, education level, number of births, urban/rural). We used Bayesian generalized linear mixed models to estimate HIV testing coverage and the proportion of those that tested as part of ANC.

Results: HIV testing coverage (<24 months) increased substantially between 2005 and 2021 from 8% to 38%, with significant variations between countries and subregions. Two percent of women received an HIV test in the 24 months preceding the survey interview as part of ANC in 2005 and 11% in 2021. Among women who received an HIV test in the 24 months preceding the survey, the probability of testing at ANC was significantly greater for multiparous, adolescent girls, rural women, women in the poorest wealth quintile, and women in West and Central Africa.

Conclusion: ANC testing remains an important component to achieving high levels of HIV testing coverage and benefits otherwise underserved women, which could prove instrumental to progress toward universal knowledge of HIV status in SSA.

FIGURE 1.

Flowchart of considered, included, and excluded nationally representative population-based surveys for the analyses of recent HIV testing and proportion of HIV testing in ANC settings in sub-Saharan Africa (2005–2021).

FIGURE 2.

Recent HIV testing coverage (ie, last 24 months), in ANC settings and outside of ANC, among women aged 15–49 years in sub-Saharan Africa countries, between 2005 and 2021. We present country-mean estimates of the proportion of women who received an HIV test in the past 24 months (gray curves), broken down between testing during an ANC visit (red curves), and outside of an ANC visit (blue curves). Crosses, circles, and triangles represent survey mean estimates of recent HIV testing coverage, recent testing at ANC, and recent testing outside of ANC, respectively. Shaded areas represent 95% uncertainty intervals. Countries in white on the inset map of Africa were not included in the model for the lack of survey data. Countries in Western, Central, Eastern, and Southern Africa are colored in orange, indigo, red, and turquoise, respectively. Centr. Afr. Rep., Central African Republic; Congo, Republic of Congo; DRC, Democratic Republic of Congo; Sao Tome, Sao Tome and Principe.

FIGURE 3.

Odds ratio mean estimates and 95% intervals for the logistic regression model predicting recent HIV testing among all women aged 15–49 years in sub-Saharan Africa (A) and for the logistic regression model predicting testing during ANC among women who reported recently being tested for HIV (B), by covariates. Purple circles are used for demographic predictors, red triangle for socioeconomic variables, and blue square for other covariates. The mean estimates are displayed above the circles. Horizontal lines are 95% confidence intervals. For each variable, the reference category is indicated in parenthesis.

FIGURE 4.

Change in the probability of recent HIV testing among all women aged 15–49 years in sub-Saharan Africa (A) and of testing during ANC among women who reported recently being tested for HIV (B), given a change in the covariates. This figure shows the estimated mean marginal effects of 6 covariates on the probability of HIV testing at ANC on a linear scale, obtained by postestimation simulation from the model. Purple circles are used for demographic predictors, red triangle for socioeconomic variables, and blue square for other covariates. The mean effect estimates are displayed above the circles. Horizontal lines are 95% credible intervals.