Determinants of non-testing for HIV among women during antenatal care follow up in sub-saharan Africa: a hierarchical analysis of recent Demographic and Health Survey datasets Authors: Tadesse Tarik Tamir, Alebachew Ferede Zegeye, Enyew Getaneh Mekonen, Bikis Liyew, Belayneh Shetie Workneh Published: 2024 ABSTRACT IntroductionHIV/AIDS is a global public health concern that is closely related to other sociocultural problems. The prevention of mother to child transmission cascade often begins with and is integrated into antenatal care in order to ensure a high rate of case detection and optimal treatment coverage. Although guidelines suggest that all pregnant women should have HIV testing as part of their regular screening checks during antenatal care, a significant proportion of pregnant women were not tested for HIV during antenatal care follow-up in sub-Saharan Africa. Hence, this study was aimed at assessing the determinants of women not tested for HIV during antenatal care follow-up in sub-Saharan Africa.MethodA cross-sectional study was conducted to assess the determinants of non-testing for HIV among women as part of antenatal care in sub-Saharan Africa, utilizing secondary data from the 2015–2022 Demographic and Health Surveys. The study included a weighted sample of 788,421 women who had antenatal care follow-up during their most recent pregnancy within five years preceding the survey. A multilevel mixed effect logistic regression analysis was employed to identify the determinants of non-testing for HIV among these women. The adjusted odds ratios with 95% confidence intervals were calculated to determine the associations between the outcome and explanatory variables. Statistical significance was determined using a p-value of less than 0.05.ResultThe hierarchical analysis of this study identified several significant factors associated with women not being tested for HIV during antenatal care in sub-Saharan Africa. These factors include young age, low or no education, not being in a union, unemployment, and low household wealth index. Additionally, limited media exposure, stigmatizing attitudes towards people living with HIV, male-headed households, and rural residence were significant at the individual and household levels. At the community level, low media exposure and high illiteracy rates were significant, while at the country level, high fertility rates and low literacy rates were also associated with lower rates of HIV testing during antenatal care.ConclusionThis study identified significant individual, community, and country-level factors associated with women not testing for HIV during antenatal care in sub-Saharan Africa. Key factors include young age, low education, unemployment, not being in a union, low household wealth, limited media exposure, stigmatizing attitudes towards people living with HIV, male household head, rural residence, low community media exposure, high community illiteracy, high fertility rates, and low literacy rates. These findings highlight the need for targeted, context-specific interventions to improve HIV testing rates and enhance maternal and child health outcomes in the region. Introduction HIV/AIDS is an infectious disease caused by infection with the human immunodeficiency syndrome virus that compromises immune function, which in turn leads a patient to contract new illnesses [1]. HIV/AIDS is a global public health concern that is closely related to other sociocultural problems [2]. Nearly all pregnancy- and childbirth-related deaths worldwide (99%) occur in low- and middle-income nations, with sub-Saharan Africa alone accounting for 66% of the total [3]. According to joint united nations program on HIV and AIDS (UNAIDS), the increased accessibility of antiretroviral therapy (ART) has resulted in the saving of 9 million life-years in sub-Saharan Africa alone [4, 5]. However, sub-Saharan Africa continues to be the continent where the greatest number of pregnant women are living with HIV, making up about 85 percent of the global burden [5]. The prevention of mother to child transmission (PMTCT) cascade often begins with and is integrated into antenatal care (ANC) in order to ensure a high rate of case detection and optimal treatment coverage [6, 7]. Through this integration, it is anticipated that pregnant women receiving ANC will undergo routine HIV testing. If they test positive, they will receive prompt PMTCT interventions, including HIV treatment. The UNAIDS has set a 2030 deadline for zero new HIV infections, which includes improving mother health and preventing new HIV infections in children born to HIV-positive mothers [8, 9]. The World Health Organization (WHO) currently recommends routine HIV testing and treatment for all pregnant women as part of prenatal care in order to meet these objectives [10]. In addition to promoting positive results for mothers and newborns, the goal is to make targeted prenatal healthcare accessible to all [10]. According to the previous studies, HIV testing among women was influenced by a number of factors, including their level of education, marital status, place of residence, wealth status, media exposure, age of mothers, stigmatizing attitudes, understanding of HIV/AIDS, and hazardous sexual behavior [11–13]. Despite the recommendation that all pregnant women should undergo HIV testing as part of their regular screening during antenatal care (ANC), a significant proportion of pregnant women in sub-Saharan Africa were reported as not being tested for HIV during ANC follow-up in 2016 [14]. Therefore, this study aimed to assess the determinants of non-testing for HIV among women during ANC follow-up in sub-Saharan Africa. Methods Study design, setting and data source This study employed a cross-sectional design to examine the determinants of HIV non-testing among women during antenatal care (ANC) in sub-Saharan Africa. The analysis utilized secondary data from the 2015–2022 Demographic and Health Surveys (DHS), a collection of nationally representative and comparable datasets across countries. The study included 19 sub-Saharan African countries with a standard DHS dataset since 2015 and specific data on HIV testing during ANC (Fig. 1). The DHS uses stratified cluster sampling technique. From the outset, enumeration areas (sampling clusters) are selected, stratified by geographic region, urban/rural status, and other relevant factors. Then, a sample of households within each cluster is chosen, with stratification based on factors like household size and socioeconomic characteristics. This study included a total weighted sample of 788,421 women who had ANC follow-up during their most recent pregnancy within five years preceding the survey (Table 1). The datasets were accessed from the Monitoring and Evaluation to Assess and Use Results Demographic and Health Survey (MEASURE DHS) program, available at https://www.dhsprogram.com/data/available-datasets.cfm.Fig. 1Geographical location of sub-Saharan African countries included in the study. Source shape file (URL: https://africaopendata.org/dataset/africa-shapefiles)Table 1Sample size and years of survey for determinants of non-testing for HIV among women during ANC follow up in sub-Saharan Africa (n = 788,421)CountryYear of surveySample size [n (%)]Not tested for HIV [n (%)]Angola201547581 (6.0)11026 (23.2)Benin2017/1829054 (3.7)13841 (47.6)Burundi2016/1769329 (8.8)4281 (6.2)Cameroon201841845 (5.3)4886 (11.7)Ethiopia201631442 (4.0)12891 (41.0)Gabon2020/2118692 (2.4)1241 (6.6)Gambia201941525 (5.3)14982 (36.1)Guinea201827257 (3.5)18839 (69.1)Kenya202234384 (4.4)1100 (3.2)Liberia2019/2026210 (3.3)7270 (27.7)Mali201836410 (4.6)26093 (71.7)Malawi201586015 (10.9)2921 (3.4)Rwanda2019/2046814 (5.9)470 (1.0)Serra Leone201944613 (5.7)8201 (18.4)Tanzania202238905 (4.9)2523 (6.5)Uganda201672329 (9.2)4182 (5.8)South Africa201616477 (2.1)320 (1.9)Zambia201849862 (6.3)1178 (2.4)Zimbabwe201529677 (3.8)536 (1.8)Total2015–2022788421 (100.0)136783 (17.3) Geographical location of sub-Saharan African countries included in the study. Source shape file (URL: https://africaopendata.org/dataset/africa-shapefiles) Table 1Sample size and years of survey for determinants of non-testing for HIV among women during ANC follow up in sub-Saharan Africa (n = 788,421)CountryYear of surveySample size [n (%)]Not tested for HIV [n (%)]Angola201547581 (6.0)11026 (23.2)Benin2017/1829054 (3.7)13841 (47.6)Burundi2016/1769329 (8.8)4281 (6.2)Cameroon201841845 (5.3)4886 (11.7)Ethiopia201631442 (4.0)12891 (41.0)Gabon2020/2118692 (2.4)1241 (6.6)Gambia201941525 (5.3)14982 (36.1)Guinea201827257 (3.5)18839 (69.1)Kenya202234384 (4.4)1100 (3.2)Liberia2019/2026210 (3.3)7270 (27.7)Mali201836410 (4.6)26093 (71.7)Malawi201586015 (10.9)2921 (3.4)Rwanda2019/2046814 (5.9)470 (1.0)Serra Leone201944613 (5.7)8201 (18.4)Tanzania202238905 (4.9)2523 (6.5)Uganda201672329 (9.2)4182 (5.8)South Africa201616477 (2.1)320 (1.9)Zambia201849862 (6.3)1178 (2.4)Zimbabwe201529677 (3.8)536 (1.8)Total2015–2022788421 (100.0)136783 (17.3) Sample size and years of survey for determinants of non-testing for HIV among women during ANC follow up in sub-Saharan Africa (n = 788,421) Variables of the study The outcome variable of this study was the HIV testing status of women as part of antenatal care. It was categorized as "1" for women who were not tested and "0" for those who were tested. Independent variables were grouped into three levels: individual, community, and country. Individual-level variables included age, educational status, marital status, employment status, wealth index, media exposure, stigmatizing attitude towards people with HIV, and sex of the household head. Community-level factors included place of residence, community media exposure, community illiteracy level, and community poverty level. Country-level factors included fertility rate, literacy rate, and income level. To account for potential time effects arising from using DHS data collected in different years, the DHS survey year was included as a control variable (Fig. 2).Fig. 2Hierarchy of variables of the study Hierarchy of variables of the study This study used individual record (IR) data extracted from the DHS dataset. The data were accessed from the DHS official website (https://www.dhsprogram.com/data/available-datasets.cfm) through a formal request. The sole requirement for access to the datasets was registration. The IR data from 19 countries included in the analysis were upended together to create one data file using Stata version 14. The variables of the study were weighted using weighting variables of IR data [sampling weight (v005), primary sampling unit (v021), and strata (v023)] with consideration of the sampling design. Statistical analysis To analyze the factors influencing HIV non-testing among women during ANC using data from the DHS, we employed a multilevel mixed-effects logistic regression model. This approach, which accounts for the hierarchical structure of the data (individual, community, and country levels), allowed us to investigate the influence of factors at each level while controlling for variations between countries. We compared this multilevel model to a standard logistic regression model using a likelihood ratio test, confirming a significant improvement in model fit when using the multilevel approach, thus validating its suitability for our analysis. To examine the effects of various factors, we fitted four mixed-effect logistic regression models: a null model (Model 0) with no explanatory variables, a model with individual-level factors and control variable (Model I), a model with community-level factors and control variable (Model II), a model with country-level factors and control variable (Model III), and a full model (Model IV) incorporating all levels. We then analyzed the random effects (measuring variability) and fixed effects (measuring association) of each model. Random effects were quantified using variance, intra-class correlation coefficient (ICC), median odds ratio (MOR), and proportional change in variation (PCV), while fixed effects were presented as adjusted odds ratios with 95% confidence intervals. Statistical significance was determined using a p-value threshold of 0.05. The three-level multilevel mixed-effects logistic regression model can be represented as follows [15, 16]:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$yijk=\beta 0k+\beta 1kxijk+rijk+uijk$$\end{document}yijk=β0k+β1kxijk+rijk+uijk Where yijk​ is the dependent variable for the ith individual in the jth community in the kth country, β0k​ and β1k​ are the intercept and slope coefficients for the kth country, xijk​ is the independent variable for the ith individual in the jth community in the kth country, rijk​ is the random effect for the jth community in the kth country, uijk​ is the random effect for the ith individual in the jth community in the kth country. Model selection was based on comparing the log likelihood (LL), deviance (-2LL), and Bayesian Information Criterion (BIC) values of each model. Model IV was selected as the best-fit model based on these criteria. To ensure the robustness of our findings, we conducted several checks. We assessed multicollinearity, a situation where independent variables are highly correlated, by calculating the variance inflation factor (VIF) for each variable. The VIF values fell within acceptable ranges, indicating that multicollinearity was not a significant concern. Additionally, we employed multivariable regression techniques to control for potential confounding factors, variables that could influence the relationship between independent and dependent variables. This approach allowed us to isolate the effect of each independent variable while accounting for the influence of other relevant factors. Results Socio-demographic and economic characteristics of study subjects A total weighted sample of 788,421 women who attended antenatal care follow-up during pregnancy were enrolled in this study. The sample comprised 34.5% urban and 65.5% rural residents. The mean age of the study subject was 28.1 ± 6.9 years and 39.4% of the subjects had a primary school educational status. Regarding marital status, the majority (83.1%) of study subjects were in a union. Importantly, 76.1% of women had no stigmatizing attitude towards people positive for HIV. About two-thirds (67.5%) of study subjects had no media exposure, and more than half (54.18%) of them were members of communities with high poverty levels. Out of women aged 15–19 years, 15,719 (21.71%) were not tested for HIV as part of the ANC. Of the total subjects having no formal education, 76,242 (36.23%) were not tested for HIV during their antenatal care follow-up. Nearly forty percent (39.69%) of the study subjects who had a stigmatizing attitude towards people living with HIV were not tested for HIV during ANC follow-up. Additionally, more than two-thirds (67.5%) of the participants were residents of low-income countries where non-testing for HIV was 19.1% (Table 2). Table 2Socio-demographic and economic characteristics of study subjects (N = 788,421)VariablesTested for HIVTotal [N (%)]Yes [N (%)]No [N (%)]Age (mean = 28.1 ± 6.9)15–1956687 (78.3)15719 (21.7)72405 (9.2)20–24167609 (83.4)33333 (16.6)200942 (25.5)25–29170909 (83.0)35082 (17.0)205991 (26.0)30–34129507 (84.0)24683 (16.0)154191 (19.6)35–49126926 (81.9)27966 (18.1)154892 (19.7)Total651638 (82.7)136783 (17.3)788421 (100)Educational statusNo education134210 (63.8)76242 (36.2)210452 (26.7)Primary274139 (88.3)36394 (11.7)310534 (39.4)Secondry207677 (90.2)22580 (9.8)230257 (29.2)Higher35612 (95.8)1566 (4.2)37178 (4.7)Total651638 (82.7)136783 (17.3)788421 (100)Current marital statusIn union533586 (81.4)121503 (18.6)655089 (83.1)Not in union118052 (88.5)15280 (11.5)133332 (16.9)Total651638 (82.7)136783 (17.3)788421 (100)Employment statusEmployed204254 (82.3)44079 (17.8)248332 (31.5)unemployed447058 (82.8)92693 (17.2)539751 (68.5)Total651312 (82.7)136772 (17.3)788083 (100)Wealth indexPoorest127725 (77.9)36211 (22.1)163936 (20.8)Poorer129220 (79.6)33161 (20.4)162381 (20.6)Middle129668 (81.8)28795 (18.2)158464 (20.1)Richer134611 (85.0)23701 (15.0)158313 (20.1)Richest130414 (89.74)14914 (10.3)145328 (18.4)Total651638 (82.7)136782 (17.3)788422 (100)Media exposureYes444956 (83.9)85291 (16.09)530247 (67.3)No206682 (80.1)51480 (19.9)258162 (32.7)Total651638 (82.7)136771 (17.3)788409 (100)Have stigmatizing attitude towards people with HIVYes111188 (60.3)73179 (39.7)184366 (23.9)No524294 (89.2)63284 (10.8)587578 (76.1)Total635482 (82.7)136463 (17.3)771944 (100)Sex of household headMale491796 (81.3)113451 (18.7)605247 (76.8)Female159842 (87.3)23332 (12.7)183174 (23.2)Total651638 (82.7)136783 (17.3)788421 (100)Place of residenceUrban232238 (85.3)39910 (14.7)272148 (34.5)Rural419400 (81.2)96873 (18.76)516273 (65.5)Total651638 (82.7)136783 (17.3)788421 (100)Community media exposureLow383773 (82.6)80599 (17.4)464372 (58.9)High267866 (82.7)56183 (17.3)324049 (41.1)Total651639 (82.7)136782 (17.3)788421 (100)Community illiteracy levelLow231618 (86.4)36551 (13.6)268169 (34.0)High420020 (80.73)100232 (19.3)520251 (66.0)Total651638 (82.7)136783 (17.3)788421 (100)Community poverty levelLow302193 (83.6)59064 (16.4)361257 (45.8)High349445 (81.8)77718 (18.2)427164 (54.2)Total651638 (82.7)136782 (17.3)788421 (100)Fertility rateLow488438 (84.3)90904 (15.7)579,342 (73.5)High163,200 (78.1)45879 (21.9)209,079 (26.5)Total651,638 (82.7)136783 (17.3)788,421 (100)Literacy rateLow217,489 (67.4)105,038 (32.6)322,527 (40.9)High434,150 (93.2)31,745 (6.8)465,894 (59.1)Total651,639 (82.7)136,783 (17.3)788,421 (100)Income levelLow430,499 (81.0)101,308 (19.1)531807 (67.5)Middle221139 (86.2)35475 (13.8)256614 (32.6)Total651,638 (82.7)136,783 (17.3)788421 (100)HIV Human Immune Deficiency Virus Table 2Socio-demographic and economic characteristics of study subjects (N = 788,421)VariablesTested for HIVTotal [N (%)]Yes [N (%)]No [N (%)]Age (mean = 28.1 ± 6.9)15–1956687 (78.3)15719 (21.7)72405 (9.2)20–24167609 (83.4)33333 (16.6)200942 (25.5)25–29170909 (83.0)35082 (17.0)205991 (26.0)30–34129507 (84.0)24683 (16.0)154191 (19.6)35–49126926 (81.9)27966 (18.1)154892 (19.7)Total651638 (82.7)136783 (17.3)788421 (100)Educational statusNo education134210 (63.8)76242 (36.2)210452 (26.7)Primary274139 (88.3)36394 (11.7)310534 (39.4)Secondry207677 (90.2)22580 (9.8)230257 (29.2)Higher35612 (95.8)1566 (4.2)37178 (4.7)Total651638 (82.7)136783 (17.3)788421 (100)Current marital statusIn union533586 (81.4)121503 (18.6)655089 (83.1)Not in union118052 (88.5)15280 (11.5)133332 (16.9)Total651638 (82.7)136783 (17.3)788421 (100)Employment statusEmployed204254 (82.3)44079 (17.8)248332 (31.5)unemployed447058 (82.8)92693 (17.2)539751 (68.5)Total651312 (82.7)136772 (17.3)788083 (100)Wealth indexPoorest127725 (77.9)36211 (22.1)163936 (20.8)Poorer129220 (79.6)33161 (20.4)162381 (20.6)Middle129668 (81.8)28795 (18.2)158464 (20.1)Richer134611 (85.0)23701 (15.0)158313 (20.1)Richest130414 (89.74)14914 (10.3)145328 (18.4)Total651638 (82.7)136782 (17.3)788422 (100)Media exposureYes444956 (83.9)85291 (16.09)530247 (67.3)No206682 (80.1)51480 (19.9)258162 (32.7)Total651638 (82.7)136771 (17.3)788409 (100)Have stigmatizing attitude towards people with HIVYes111188 (60.3)73179 (39.7)184366 (23.9)No524294 (89.2)63284 (10.8)587578 (76.1)Total635482 (82.7)136463 (17.3)771944 (100)Sex of household headMale491796 (81.3)113451 (18.7)605247 (76.8)Female159842 (87.3)23332 (12.7)183174 (23.2)Total651638 (82.7)136783 (17.3)788421 (100)Place of residenceUrban232238 (85.3)39910 (14.7)272148 (34.5)Rural419400 (81.2)96873 (18.76)516273 (65.5)Total651638 (82.7)136783 (17.3)788421 (100)Community media exposureLow383773 (82.6)80599 (17.4)464372 (58.9)High267866 (82.7)56183 (17.3)324049 (41.1)Total651639 (82.7)136782 (17.3)788421 (100)Community illiteracy levelLow231618 (86.4)36551 (13.6)268169 (34.0)High420020 (80.73)100232 (19.3)520251 (66.0)Total651638 (82.7)136783 (17.3)788421 (100)Community poverty levelLow302193 (83.6)59064 (16.4)361257 (45.8)High349445 (81.8)77718 (18.2)427164 (54.2)Total651638 (82.7)136782 (17.3)788421 (100)Fertility rateLow488438 (84.3)90904 (15.7)579,342 (73.5)High163,200 (78.1)45879 (21.9)209,079 (26.5)Total651,638 (82.7)136783 (17.3)788,421 (100)Literacy rateLow217,489 (67.4)105,038 (32.6)322,527 (40.9)High434,150 (93.2)31,745 (6.8)465,894 (59.1)Total651,639 (82.7)136,783 (17.3)788,421 (100)Income levelLow430,499 (81.0)101,308 (19.1)531807 (67.5)Middle221139 (86.2)35475 (13.8)256614 (32.6)Total651,638 (82.7)136,783 (17.3)788421 (100)HIV Human Immune Deficiency Virus Socio-demographic and economic characteristics of study subjects (N = 788,421) HIV Human Immune Deficiency Virus Multilevel mixed effect logistic regression analysis of non-testing for HIV as part of antenatal care The multilevel mixed-effects logistic regression analysis revealed several individual, community, and country-level factors significantly associated with a higher likelihood of not receiving an HIV test as part of ANC. At the individual level, younger women, those with lower educational attainment, women not in a union, unemployed women, women from households with a low wealth index, those with no media exposure, and those who have no stigmatizing attitude towards people with HIV had notably higher odds of missing out on HIV testing. Specifically, women aged 15–19 had 1.47 times higher odds (AOR = 1.47, 95% CI: 1.30, 1.65), those aged 20–24 had 1.18 times higher odds of not being tested (AOR = 1.18, 95% CI: 1.07, 1.30) compared to women aged 35 years and older. Additionally, women with no formal education were 3.65 times more likely to miss an HIV test (AOR = 3.65, 95% CI: 2.92, 4.57) than those with higher education, while those with only primary schooling and secondary schooling had 2.62 (AOR = 2.62, 95% CI: 2.09, 3.27) and 1.82 (AOR = 1.82, 95% CI: 1.46, 2.28) times higher odds, respectively. Marital status also impacted testing likelihood, with women currently in a union having 1.12 times higher odds (AOR = 1.12, 95% CI: 1.03, 1.22) of not being tested compared to those not in a union. Additionally, unemployed women had 1.14 times higher odds (AOR = 1.14, 95% CI: 1.07, 1.21) of not receiving an HIV test compared to employed women. The odds of not being tested for HIV during ANC follow-up were 2.31, 1.86, 1.56, and 1.35 times higher among women from households with the poorest (AOR = 2.31, 95% CI: 2.03, 1.63), poorer (AOR = 1.86, 95% CI: 1.64, 2.10), middle (AOR = 1.56, 95% CI: 1.38, 1.76), and richer wealth index (AOR = 1.35, 95% CI: 1.21, 1.51) compared to women from households with the richest wealth index. The odds of not being tested for HIV were 1.39 (AOR = 1.39, 95% CI: 1.29, 1.61) times higher among women with no media exposure when compared to women with media exposure. A stigmatizing attitude towards people living with HIV greatly increased the odds, with women holding such attitudes having 1.63 times higher odds (AOR = 1.63, 95% CI: 1.54, 1.73) of not being tested. In addition, the odds of not being tested for HIV were 1.17 times higher among women from households with male household heads (AOR = 1.17, 95% CI: 1.09, 1.25) compared to women from households with female household heads. At the community level, rural residents had 1.56 times higher odds (AOR = 1.56, 95% CI: 1.41, 1.75) of not being tested compared to urban residents. High community illiteracy was also a significant factor, with women living in such communities having 1.46 times higher odds (AOR = 1.19, 95% CI: 1.08, 1.31) of not being tested. At the country level, women in countries with high fertility rates had 1.11 times higher odds (AOR = 1.11, 95% CI: 1.03, 1.20) of not being tested. Women in countries with low literacy rates faced particularly high odds, being 6.34 times more likely (AOR = 6.34, 95% CI: 4.89, 14.53) to miss an HIV test compared to those in countries with moderate literacy rates. The result of the random effects at the country level revealed significant variability in the likelihood of non-testing for HIV among women during ANC across the countries. The variance in the null model is 3.65 (95% CI: 2.13–6.25), which decreased progressively to 1.48 (95% CI: 0.79–2.77) in Model IV. This reduction indicates that the inclusion of predictors in the models accounts for some of the variability between countries. The ICC also decreased from 41.48% (95% CI: 29.01–54.45) in the null model to 24.25% (95% CI: 14.57–37.52) in Model IV. This suggests that country-level factors become less influential as more predictors are included in the models. The MOR follows a similar trend, decreasing from 6.18 (95% CI: 4.00–10.75) in the null model to 3.17 (95% CI: 2.32–4.86) in Model IV, indicating a reduction in the country-level differences in non-testing for HIV. The Proportional Change in Variance (PCV) shows a substantial increase from Model I (20.82%) to Model IV (59.45%), demonstrating that the models explain a significant portion of the variance at the country level. At the community level, the variance in non-testing for HIV among women during ANC shows a slight fluctuation but generally decreased from 1.93 (95% CI: 1.79–2.08) in the null model to 1.32 (95% CI: 1.22–1.44) in Model IV. The ICC decreased from 63.20% (95% CI: 54.40–70.73) in the null model to 45.98% (95% CI: 37.83–54.35) in Model IV, indicating that community-level factors also become less influential with the inclusion of more predictors. The MOR decreased from 3.74 (95% CI: 3.56–3.94) in the null model to 2.98 (95% CI: 2.86–3.12) in Model IV, showing a reduction in community-level differences in non-testing for HIV. The PCV fluctuates but shows a notable reduction in variance in Model IV (31.61%). The model fit statistics indicate that the inclusion of more predictors improves the model fit. The log likelihood increases from -24680.20 in the null model to -23439.16 in Model IV, suggesting better model fit with more predictors. Similarly, the deviance decreased from 49,360.40 in the null model to 46,878.32 in Model IV, further indicating improved model fit. The also decreases from 49,405.51 in the null model to 47,182.26 in Model IV, confirming that Model IV provides the best fit among the compared models (Table 3). Table 3Measure of association of women not tested for HIV as part of antenatal careMeasures of Association (Fixed effects)FactorsModel I AOR(95% CI)Model II AOR(95% CI)Model III AOR(95% CI)Model IV AOR(95% CI)Survey year2015–20191.29 (1.04–1.61)1.27 (1.01–1.60)1.69 (1.60, 1.79)1.41 (1.13, 1.76)2020–20221.001.001.001.00Age15–191.36 (1.22, 1.50)1.47 (1.30, 1.65)*20–241.09 (1.01–1.18)1.18 (1.07–1.30)*25–290.95 (0.88–1.03)1.02 (0.93–1.11)30–340.93 (0.86–1.01)0.95 (0.88–1.04)35–491.001.00Educational statusNo education3.87 (3.09–4.83)3.65 (2.92–4.57)*Primary2.74 (2.19–3.42)2.62 (2.09–3.27)*Secondry1.86 (1.54–2.25)1.82 (1.46–2.28)*Higher1.001.00Current marital statusIn union1.11 (1.02–1.20)1.12 (1.03–1.22)*Not in union1.001.00Employment statusEmployed1.001.00Unemployed1.13 (1.06–1.20)1.14 (1.07–1.21)*Wealth indexPoorest3.00 (2.67–3.38)2.31 (2.03–2.63)*Poorer2.35 (2.09–2.63)1.86 (1.64–2.10)*Middle1.88 (1.68–2.11)1.56 (1.38–1.76)*Richer1.50 (1.35–1.68)1.35 (1.21–1.51)*Richest1.001.00Media exposureYes1.001.00No1.41 (1.31–1.49)1.39 (1.29–1.61)*Have stigmatizing attitude towards people with HIVYes1.65 (1.55–1.75)1.63 (1.54–1.73)*No1.001.00Sex of Household headMale1.19 (1.11–1.27)1.17 (1.09–1.25)*Female1.001.00Place of residenceUrban1.001.00Rural3.12 (2.77–3.45)1.56 (1.41–1.75)*Community media exposureLow0.98 (0.89–1.08)1.04 (0.95–1.14)High1.001.00Community illiteracy levelLow1.001.00High1.31 (1.19–1.45)1.19 (1.08–1.31)*Community poverty levelLow1.001.00High1.04 (0.95–1.15)1.00 (0.92–1.10)Fertility rateLow1.001.00High1.23 (1.18–1.31)1.11 (1.03–1.20)*Literacy rateLow7.47 (5.09–16.35)6.34 (4.89–14.53)Moderate1.001.00Income levelLow0.48 (0.15–1.52)0.30 (0.09–1.04)Middle1.001.00Measures of Variation (Random effects)ParametersNull modelModel IModel IIModel IIIModel IVCountry levelVariance (95% CI)3.65 (2.13–6.25)2.89 (1.55, 5.38)3.03 (1.73–5.23)1.72 (0.98, 3.01)1.48 (0.79–2.77)ICC (%, 95% CI)41.48 (29.01–54.45)38.54 (25.23–53.83)43.09 (30.23–56.95)24.87 (15.90–36.69)24.25 (14.57–7.52)MOR (95% CI)6.18 (4.00–10.75)5.03 (3.26–9.06)6.11 (4.01–10.71)3.48 (2.56–5.20)3.17 (2.32–4.86)PCV (%)Reference20.82 (27.23–13.92)16.99 (18.78–16.32)52.88 (53.99–51.84)59.45 (62.91–55.94)Community levelVariance (95% CI)1.93 (1.79–2.08)1.32 (1.22–1.43)1.64 (1.52–1.77)1.89 (1.76, 2.04)1.32 (1.22–1.44)ICC (%, 95% CI)63.20 (54.40–70.73)56.14 (45.49–66.25)62.04 (52.54–70.69)52.34 (45.59–59.01)45.98 (37.83–4.35)MOR (95% CI)3.74 (3.56, 3.94)2.97 (2.86–3.11)3.34 (3.23–3.54)3.69 (3.53–3.88)2.98 (2.86–3.12)PCV%Reference31.61 (31.84–32.02)15.03 (15.08–14.90)2.07 (1.68–1.92)31.61 (31.84–30.77)Model fitnessLog likelihood-24680.20-23495.49-24360.51-24644.48-23439.16Deviance49,360.4046,990.9848,721.0249,288.9646,878.32BIC49405.5147216.1248811.2549367.9247182.26AOR Adjusted odds ratio, BIC Bayesian Information Criterion, CI Confidence Interval, ICC intra-class correlation coefficient, MOR median odds ratio, PCV proportional change in variation*Statistically significant at P value less than 0.05 Table 3Measure of association of women not tested for HIV as part of antenatal careMeasures of Association (Fixed effects)FactorsModel I AOR(95% CI)Model II AOR(95% CI)Model III AOR(95% CI)Model IV AOR(95% CI)Survey year2015–20191.29 (1.04–1.61)1.27 (1.01–1.60)1.69 (1.60, 1.79)1.41 (1.13, 1.76)2020–20221.001.001.001.00Age15–191.36 (1.22, 1.50)1.47 (1.30, 1.65)*20–241.09 (1.01–1.18)1.18 (1.07–1.30)*25–290.95 (0.88–1.03)1.02 (0.93–1.11)30–340.93 (0.86–1.01)0.95 (0.88–1.04)35–491.001.00Educational statusNo education3.87 (3.09–4.83)3.65 (2.92–4.57)*Primary2.74 (2.19–3.42)2.62 (2.09–3.27)*Secondry1.86 (1.54–2.25)1.82 (1.46–2.28)*Higher1.001.00Current marital statusIn union1.11 (1.02–1.20)1.12 (1.03–1.22)*Not in union1.001.00Employment statusEmployed1.001.00Unemployed1.13 (1.06–1.20)1.14 (1.07–1.21)*Wealth indexPoorest3.00 (2.67–3.38)2.31 (2.03–2.63)*Poorer2.35 (2.09–2.63)1.86 (1.64–2.10)*Middle1.88 (1.68–2.11)1.56 (1.38–1.76)*Richer1.50 (1.35–1.68)1.35 (1.21–1.51)*Richest1.001.00Media exposureYes1.001.00No1.41 (1.31–1.49)1.39 (1.29–1.61)*Have stigmatizing attitude towards people with HIVYes1.65 (1.55–1.75)1.63 (1.54–1.73)*No1.001.00Sex of Household headMale1.19 (1.11–1.27)1.17 (1.09–1.25)*Female1.001.00Place of residenceUrban1.001.00Rural3.12 (2.77–3.45)1.56 (1.41–1.75)*Community media exposureLow0.98 (0.89–1.08)1.04 (0.95–1.14)High1.001.00Community illiteracy levelLow1.001.00High1.31 (1.19–1.45)1.19 (1.08–1.31)*Community poverty levelLow1.001.00High1.04 (0.95–1.15)1.00 (0.92–1.10)Fertility rateLow1.001.00High1.23 (1.18–1.31)1.11 (1.03–1.20)*Literacy rateLow7.47 (5.09–16.35)6.34 (4.89–14.53)Moderate1.001.00Income levelLow0.48 (0.15–1.52)0.30 (0.09–1.04)Middle1.001.00Measures of Variation (Random effects)ParametersNull modelModel IModel IIModel IIIModel IVCountry levelVariance (95% CI)3.65 (2.13–6.25)2.89 (1.55, 5.38)3.03 (1.73–5.23)1.72 (0.98, 3.01)1.48 (0.79–2.77)ICC (%, 95% CI)41.48 (29.01–54.45)38.54 (25.23–53.83)43.09 (30.23–56.95)24.87 (15.90–36.69)24.25 (14.57–7.52)MOR (95% CI)6.18 (4.00–10.75)5.03 (3.26–9.06)6.11 (4.01–10.71)3.48 (2.56–5.20)3.17 (2.32–4.86)PCV (%)Reference20.82 (27.23–13.92)16.99 (18.78–16.32)52.88 (53.99–51.84)59.45 (62.91–55.94)Community levelVariance (95% CI)1.93 (1.79–2.08)1.32 (1.22–1.43)1.64 (1.52–1.77)1.89 (1.76, 2.04)1.32 (1.22–1.44)ICC (%, 95% CI)63.20 (54.40–70.73)56.14 (45.49–66.25)62.04 (52.54–70.69)52.34 (45.59–59.01)45.98 (37.83–4.35)MOR (95% CI)3.74 (3.56, 3.94)2.97 (2.86–3.11)3.34 (3.23–3.54)3.69 (3.53–3.88)2.98 (2.86–3.12)PCV%Reference31.61 (31.84–32.02)15.03 (15.08–14.90)2.07 (1.68–1.92)31.61 (31.84–30.77)Model fitnessLog likelihood-24680.20-23495.49-24360.51-24644.48-23439.16Deviance49,360.4046,990.9848,721.0249,288.9646,878.32BIC49405.5147216.1248811.2549367.9247182.26AOR Adjusted odds ratio, BIC Bayesian Information Criterion, CI Confidence Interval, ICC intra-class correlation coefficient, MOR median odds ratio, PCV proportional change in variation*Statistically significant at P value less than 0.05 Measure of association of women not tested for HIV as part of antenatal care AOR Adjusted odds ratio, BIC Bayesian Information Criterion, CI Confidence Interval, ICC intra-class correlation coefficient, MOR median odds ratio, PCV proportional change in variation *Statistically significant at P value less than 0.05 Discussion This study revealed the determinants of women not tested for HIV during ANC follow-up in sub-Saharan Africa using data from recent Demographic and Health surveys. The multilevel mixed effect logistic regression analysis of this study found that young women age, no and low women education, women not in union, women unemployment, low household wealth index, women media exposure, a stigmatizing attitude towards people living with HIV, male household head, rural residence, low community media exposure, and high community illiteracy were significantly associated with HIV not tested as part of antenatal care. According to the findings of this study, compared to women 35 years and older, the odds of not having an HIV test as part of ANC were higher for women aged 15–19, 20–24, and 25–29 years. This is in line with previous findings [12, 17]. These findings implicate the necessity for ANC providers to inform women that women of all ages and relationship statuses can catch HIV and should be tested for the virus, which helps prevent mother-to-child transmission. In this study, education was significantly associated with women not being tested for HIV during ANC follow up. Several studies have been reporting the association of educational status of women with HIV testing [14, 17–19]. Taking women with higher educational status as a reference, the odds of not being tested for HIV as part of ANC were higher for women with no formal education, primary schooling, and secondary schooling. In addition, the odds of not being tested for HIV were higher among women living in a community with high illiteracy compared to women living in a community with low level of illiteracy. Pregnant women with higher levels of education were found to be more likely to take HIV tests, which suggests that lower-educated women could not fully understand the advantages of doing so [20]. Similarly, educated women may be more exposed to information about HIV/AIDS, be more aware of the benefits of testing for the virus, and be better able to decide whether to get tested for the virus [21]. Reducing vertical transmission of HIV may be achieved by stepping up HIV testing in rural health institutions and offering programs to remove obstacles to testing for women without formal education [20]. Regarding the marital status of women, the odds of not being tested for HIV were higher among women currently in union compared to women currently not in union. Similar findings were obtained from earlier studies [22, 23]. This is most likely a result of their perceived personal danger of contracting HIV from previous or present relationships [23]. The women unemployment was positively associated with not being tested for HIV as part of ANC. Accordingly, when compared to employed women, the odds of not being tested for HIV as part of ANC were higher among women not employed. This finding is supported by a scoping review of the literature that addresses the relationship between unemployment and the HIV continuum of care, employment status is positively associated with HIV testing [24]. This could be due to the fact that employed women are more knowledgeable and have awareness about the advantages of HIV testing and its importance for the prevention of mother-to-child transmission, which the intern might have insisted on for the testing. In addition, previous findings have also shown that, while job type matters, working women are more likely to receive HIV medical care, including testing, timely linkage to care, continuation of HIV care, and adherence to HIV medication [24, 25]. Consistent with previous findings [14, 17, 26], the odds of not being tested for HIV during ANC as wealth index decreased from highest wealth index to lowest wealth index. Accordingly, the odds of not being tested for HIV during ANC follow-up were higher among women from households with the richer, middle, poorer, and poorest wealth index compared to women from households with the richest wealth index. This study shares the implication from previous study [27] that Poverty is the major factor influencing the decision to test for HIV during ANC. This finding may be explained by the fact that women in the wealthier quintile use prenatal care services more frequently than women in the poorer quintile, reaching the WHO-recommended threshold of four prenatal controls. This suggests that these women are more likely to have prenatal HIV testing as a component of mother to child transmission prevention (PMTCT) [28, 29]. This study found that media exposure was significantly associated with women not tested for HIV. This finding agrees with previous studies on influence of mass media on HIV testing [30, 31]. The odds of not being tested for HIV were higher among women with no media exposure when compared to women with media exposure. Studies show that media exposure influences HIV testing by acting on the perception of risk and motivating the individual to protect themselves [30, 31]. The reduction of stigma by mass media was also reported to have a positive effect on the probability of being tested for HIV [30, 32]. According to this study, the odds of not being tested for HIV during ANC follow-up were higher among women having a stigmatizing attitude towards people living with HIV as compared to women who didn’t have. This is supported by a study conducted on HIV related stigma and HIV testing [23, 33]. HIV/AIDS is the most stigmatized medical condition compared to all others, and stigma toward people infected with the virus or the social groups they are a part of has been present since the beginning of the epidemic [23]. Individuals living with HIV/AIDS have social issues related to the illness in addition to personal medical issues [34–36]. Consequently, women would not be inclined to employ voluntary counseling and testing services out of fear of stigmatization [37–39]. Though studies show that routine HIV testing is a prerequisite for biological HIV prevention efforts [23]. Anticipated stigma, however, continues to be a strong predictor of HIV testing behaviors, which inhibits many people, particularly women, from getting tested and disclosing their status. Regarding household headship, the odds of not being tested for HIV were higher among women from households with male household heads compared to women from households with female household heads. This is in agreement with the findings of the previous study [40]. It would make sense given that women who live in households headed by men are more likely to have a single sexual partner, which lowers their intention to get tested for HIV due to lower risk and a fear of the virus. Another possible explanation could be that women are not free to choose whether to get HIV testing or other health treatments when the head of the household is male. Our study also found that women from rural areas had higher odds of not being tested for HIV compared with those from urban areas, and this is supported by previous studies [11, 12]. It is generally accepted that urban populations tend to use healthcare more frequently than their rural counterparts because of improved financial resources, health knowledge, and access to transportation. This may also be justified because rural populations in most poor nations lack access to health facilities, which reduces their chances of receiving HIV testing and counseling [41]. The current study also found that women in countries with high fertility rates have higher odds of not being tested for HIV. This trend can be attributed to several factors. High fertility rates are often associated with limited access to healthcare services, including HIV testing, as healthcare resources in these countries are typically stretched thin [42]. Additionally, high fertility rates are frequently linked to lower socioeconomic status and limited education, which can further hinder access to HIV testing and other preventive services [43]. The increased odds of not being tested for HIV in high fertility countries can also be explained by cultural norms that prioritize childbirth over preventive health measures and potential stigma associated with HIV testing [44]. To improve HIV testing rates in these countries, it is essential to integrate HIV testing with maternal and child health services, increase community awareness, and reduce stigma. Furthermore, the findings revealed that women in countries with low literacy rates face particularly high odds, being more likely to miss an HIV test compared to those in countries with moderate literacy rates. Low literacy rates are strongly associated with poor health outcomes, including lower rates of HIV testing [43–46]. Individuals with low literacy levels often have limited understanding of health information, which can impede their ability to seek and utilize healthcare services [44, 45]. The significantly higher odds of missing an HIV test among women in low literacy countries can be explained by the barriers posed by low literacy, such as difficulty in understanding the importance of HIV testing, navigating healthcare systems, and communicating with healthcare providers [45]. Interventions to improve HIV testing rates in these settings should focus on enhancing health literacy through education programs, using culturally appropriate communication strategies, and providing support services to help individuals navigate the healthcare system [43]. Though the study used multilevel mixed-effect logistic regression model to determine two-level factors, which was not possible by using conventional logistic regression, only sub-Saharan African countries that have had standard DHS with data on HIV testing as part of ANC since 2015 were considered in the analysis, which might have affected the generalizability of the study findings to all sub-Saharan African countries. Conclusion This study highlights the multifaceted nature of factors influencing HIV testing during antenatal care among women in sub-Saharan Africa. The study revealed that individual level factors such as young age, low or no education, unemployment, not being in a union, low wealth index, male household head, and limited media exposure significantly impact the likelihood of women not being tested for HIV. At the community level, rural residence, low community media exposure, and high community illiteracy were found to be significant barriers to HIV testing. Moreover, broader country-level factors such as high fertility rates and low literacy rates also play a crucial role. The presence of stigmatizing attitudes towards people living with HIV exacerbates these challenges, highlighting the need for comprehensive anti-stigma campaigns. Our findings underscore the importance of addressing these diverse factors through targeted interventions. Healthcare workers providing HIV testing services as part of antenatal care should consider these individual, household, community, and country-level determinants. Policies aimed at preventing mother-to-child transmission of HIV must be tailored to address the specific needs of different clusters, ensuring that interventions are both effective and contextually relevant. By adopting a holistic approach that considers the complex interplay of these factors, we can improve HIV testing rates during antenatal care and ultimately enhance maternal and child health outcomes across sub-Saharan Africa.