Study design and population

This paper is a secondary analysis of the 2018 Ethiopia Population-based HIV Impact Assessment (EPHIA). The EPHIA was a cross-sectional, household-based sero-survey conducted in urban Ethiopia from October 2017 to March 2018. The survey aimed to estimate the prevalence of HIV and the coverage of HIV prevention and treatment services among adults and adolescents between 15–64 years. Our analysis includes eligible household members between the ages of 15–24 years old, who slept in the household the night before the survey, spoke one of six survey languages and were willing and able to provide written informed consent or assent. Non-de facto household members, those who did not sleep in the household the night before the survey, were not eligible for participation. We assessed the HIV testing behaviors of persons aged 15–24 years and focused on those who were HIV-negative or unaware of their HIV-positive status. We aim to identify the factors that might facilitate youth’s access to HIV testing, particularly among those at risk of acquiring HIV and those at risk of unknowingly transmitting the virus.

Sampling design

The EPHIA used a two-stage cluster sampling design [14] to select a representative sample of enumeration areas (EAs) and households across urban areas of Ethiopia. Urban was defined as any region, district, or capital where residents do not engage in agricultural practices as the primary source of income. The 2007 National Ethiopian Census served as the sampling frame for the selection of enumeration areas. The first stage of sampling included 393 EAs selected randomly from all eleven regions of the country based on a probability proportional to regional population size. The second stage of sampling involved an average number of 30 households randomly selected per cluster, using an equal probability method, for a total of 11,810 households.

Participant eligibility and data collection

A household questionnaire was administered to selected households. The questionnaire assessed for the biobehavioral factors related to HIV infection, including sociodemographic factors, reproductive history, sexual activity, HIV testing and the utilization of HIV care and treatment services, and attitudes towards HIV disclosure. Data from household and individual interviews were collected using mobile tablets and were stored on a central server [15].

All eligible participants received home-based HIV testing and counseling (HBTC) following Ethiopia’s rapid HIV diagnostic algorithm. Although participants self-reported their HIV status in the individual questionnaire, HIV status for the survey was defined by the result of the HIV rapid test. HIV rapid test results were returned to consenting participants at the household. Laboratory confirmation of seropositive samples was conducted using the Geenius HIV-1/2 supplemental assay. Self-reported awareness of HIV-positive status was confirmed by HIV-1 RNA and antiretroviral (ARV) drug testing [16].

Variables

We assessed the demographic, behavioral, and structural factors associated with HIV testing in the year preceding the survey. We generated a three-level outcome variable comparing: 1) those who self-report testing in the 12 months preceding the survey, 2) those self-reporting testing more than 12 months since the survey, and 3) those who never tested. Those who reported ever testing for HIV and reported testing in the 12 months prior to the survey, regardless of whether they received their test results, were categorized as having tested for HIV in the last 12 months. We categorized those reporting a last HIV test date as more than 12 months preceding the survey separately. Those who did not report ever testing for HIV were categorized as having never tested.

Age, marital status, education, employment, and urban area size are covariates in this analysis. Youth were disaggregated into two five-year age groups of 15–19 years and 20–24 years. For this analysis, the term ‘adolescent’ refers to persons aged 15–19 years, the term ‘young adult’ are persons aged 20–24 years, and the term ‘youth’ describes all those aged 15–24 years. Small and large urban areas were determined by their population size. Small urban areas were those with less than 50,000 people and large urban areas were areas with 50,000 people or more. Individuals missing data due to a response of “don’t know” or refusal to answer the questions on marital status, education, and employment were categorized with those who answered negatively to the question (never married, no education, and not employed, respectively). Those missing a response were included in the sample to maintain the overall sample size and the power within the regression models. Of this sample, 0.7% of respondents are missing or did not disclose their marital status, 0.1% of respondents did not disclose their employment status, and 0.2% of respondents are missing or did not disclose their highest level of education.

The sexual risk behaviors included in this analysis were age at first sex, the presence of a non-spousal or non-live-in sex partner in the past 12 months, condom use at last sex in the past 12 months, and engagement in transactional sex. To evaluate the accessibility of HIV testing within the Ethiopian healthcare system, we included the EPHIA variables for antenatal care (ANC) and attendance at tuberculosis clinics in our analysis.

Data analysis

Survey weights were applied to the responses to the household interview, the individual participant interview, and the HIV biomarker survey data. Sampling weights allow for two-cluster survey data to be generalizable beyond the sample by accounting for sample selection probability and for the nonresponse of the households and individuals included in the original sample. We analyzed responses from survey participants ages 15–24 years who had complete questionnaire and HIV testing data, and thus used the biomarker weights. Additionally, data analysis was only conducted among HIV-negative and unaware HIV-positive individuals, with awareness of HIV status confirmed through self-reporting or confirmatory ARV drug testing. Individuals that were found to have ARVs in their blood were considered aware of their HIV status and excluded from the analysis.

We generated bivariate weighted percentages and prevalence ratios (PRs) of testing for HIV in the last 12 months by select demographic, behavioral, and healthcare covariates. We conducted significance testing (p<0.05) to determine if prevalence ratios for testing for HIV in the last 12 months significantly differed from 1, and generated p-values (Table 2). We generated two multinomial logistic regression models [17], separating male and female respondents to eliminate the bias of confounding by gender. In the multinomial model, we compared the likelihood for testing in the past 12 months and testing more than 12 months before the survey to the likelihood of never testing. The models were adjusted for age, marital status, and employment in the past 12 months. To assess sexual behavioral risk for HIV infection, the variable for ‘non-spousal sex partner in the past 12 months’ was included in the final model. Similarly, medical male circumcision and antenatal care attendance served as indicators for healthcare access within their respective male and female multivariate models. We conducted significance testing to determine if the adjusted relative risk ratios significantly differed from 1, using a p-value of less than 0.05. Stata statistical software version SE 15.0 was used to perform all statistical analysis. All weighted estimates and 95% confidence intervals (CI) were calculated using jackknife replicate weights that account for sampling, nonresponse, and under coverage in the variance estimation.

Ethics for human subjects research

The study was approved by the Institutional Review Boards at the Ethiopia Public Health Institute, Columbia University Medical Center, Westat, and the Institutional Review Boards at The Centers for Disease Control and Prevention in Atlanta and Ethiopia. The survey was conducted by the Ethiopia Public Health Institute (EPHI), the Central Statistics Agency of Ethiopia, ICAP at Columbia University, Westat, and with technical assistance provided by the CDC. Participant consent was obtained through written informed consent or assent. Adult participants 18–64 years and emancipated minors 13–17 years provided informed consent prior to completing the individual interview and prior to survey blood collection. Adolescents 12–17 years were required to obtain parental permission and provide written assent to participate in the interview and blood collection.

Sample characteristics

Youth participants resided, almost evenly, between small and large urban areas (Table 1). Twenty-two percent (22.3%, 95% CI: 20.5–24.1) of all youth have ever married or lived with a partner as if married (Table 1). However, the proportion of female youth who have married was three times that of male youth. Among urban youth 15–24 years, 64% (95% CI: 61.8–66.2) report achieving secondary or post-secondary education as their highest level of education. More than two-thirds (69.4%, 95% CI: 67.5–71.3) of all youth were not employed or did not receive compensation for work in the 12 months preceding the survey (Table 1).

Nearly forty percent (39.6%, 95% CI: 37.8–41.4) of all youth 15–24 years reported ever having sex. Of these, 18.7% (95% CI: 16.5–21.0) of females and 13.4% (95% CI: 11.1–16.1) of males report sexual debut before the age of 15. Thirty-seven percent (36.8%, 95% CI: 33.5–40.3; n = 682) of youth reported having at least one non-spousal or non-live-in sexual partner in the last year; but the rate was higher in males (58.1%, 95% CI: 52.4–63.5) versus females (23.0%, 95% CI: 20.2–25.9). For youth who reported having sex in the 12 months preceding the survey, 83.4% (95% CI: 80.9–85.7; n = 1,751) reported that they did not use a condom at last sex. Proportionally more female than male youth reported not using a condom at last sex (92.2% (95% CI: 90.6–93.6) vs. 69.3% (95% CI: 63.4–74.7), respectively).

Thirty percent (30.1%, 95% CI: 26.5–33.9) of male youth had a medical circumcision (circumcision performed by a physician or clinical officer) (Table 1). Among female youth who gave birth in the three years preceding the survey, 94.5% (95% CI: 89.8–97.1) report to have received antenatal care during their most recent pregnancy (Table 1).

Prevalence of HIV testing in the past 12 months

Among HIV-negative and HIV-positive Ethiopian youth unaware of their HIV status, 51.4% (n = 3,971) report ever being tested for HIV (Fig 2). Twenty-two percent (21.8%) of these youth reported to having an HIV test in the 12 months preceding the survey and 27.5% tested for HIV more than a year before the survey (Fig 2). Proportionally more youth aged 20–24 than youth 15–19 years reported having tested in the last 12 months (31.4% (95% CI: 29.2–33.7) vs. 12.3% (95% CI: 10.9–13.8), respectively (Table 2). Twenty-seven percent (26.8%, 95% CI: 25.2–28.6) of female youth reported testing for HIV in the last 12 months, while only 16.8% (95% CI: 14.8–19.0) of male youth have tested within this same timeframe. Forty-seven (47.6%, 95% CI: 42.5–52.8) percent of female youth who attended an antenatal clinic in the last three years also received an HIV test in that year (Table 2). Forty percent, (40.5%, 95% CI: 32.8–48.8) of those who have ever been to a TB clinic also received HIV testing in the last year (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Percentage^† and prevalence ratio of testing for HIV ≤12 months among HIV-negative and unaware HIV-positive youth (15–24 years), by demographic characteristics, EPHIA 2017–2018.

https://doi.org/10.1371/journal.pone.0265710.t002

Predictors of HIV testing in youth

For all youth, age, gender, marital status, and recent employment were significant predictors of HIV testing in the 12 months preceding the survey. Young adults (20–24 years) were twice as likely to have tested for HIV in the last 12 months than those 15–19 years of age (PR = 2.6, 95% CI: 2.3–2.9) (Table 2). Regardless of age, female youth were almost twice as likely to have tested in the last year in comparison to male youth (PR = 1.6, 95% CI: 1.4–1.8). Those ever married or have lived with a partner were three times as likely to have tested in the last 12 months (PR = 2.8, 95% CI: 2.5–3.1). Employed youth were nearly twice as likely to have HIV tested in the past year than unemployed youth (PR = 1.6, 95% CI: 1.4–1.8).

Youth who have been sexually active (PR = 3.2, 95% CI:2.8–3.6), and those who received an STD diagnosis in the last year (PR = 1.5, 95% CI: 1.0–2.1) were significantly more likely to have tested for HIV in the last year than to have never tested (Table 2). Those reporting at least one non-spousal, non-live in sexual partner in the past 12 months were significantly less likely to have tested for HIV in the past year, in comparison to those with all marital or live-in partners (PR = 0.74, 95% CI: 0.6–0.9). Youth who report using a condom at last sex were significantly less likely to have tested for HIV in the last year (PR = 0.80, 95% CI: 0.6–1.0). Past visits to a TB clinic significantly increased the likelihood of testing HIV in the past 12 months among all youth (Table 2).

Table 3 presents the multinomial logistic regression models for HIV testing by gender. Among males, controlling for variables of sexual behavioral risk and healthcare access, being older (20–24 years) was the only significant predictor for recent HIV testing likelihood. Young adult males (20–24 years), in comparison to adolescent males (15–19 years), had a three-fold likelihood of testing for HIV in the past year (RR = 2.8, 95% CI:1.4–5.7) than never testing. Marital status, education, and history of non-spousal sex partner in the past 12 months were not associated with HIV testing at any time, either in the past 12 months or more than 12 months ago for males (Table 3). In contrast, for female youth, HIV testing in the past 12 months was associated with being older (20–24 years old), being married, having at least a secondary education, and ANC attendance at the last pregnancy (Table 3). This suggests that female youth may receive more effective HIV messages than male youth, and/or that existing HIV services may be more targeted to female youth.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Among youth ages 15–24 (HIV-negative and unaware HIV-positive), multinomial regression of HIV testing by select demographic characteristics, EPHIA 2017–2018.

https://doi.org/10.1371/journal.pone.0265710.t003

Age and gender differences in HIV testing

The age and sex differences in recent HIV testing among youth in urban Ethiopia are consistent with the results of previous population-based surveys conducted in other Sub-Saharan African countries. A pooled analysis of data from the 2007–2018 Zambia Demographic Health Surveys (DHS) examined age and sex differences in the uptake of HIV testing services among men and women 15–49 years. Women 25–49 years had the greatest increase in HIV testing overtime, HIV testing uptake for older men (25–49 years) and female youth (15–24 years) increased at the same rate, while the uptake of HIV testing increased the least for male youth 15–19 years [18]. In eastern and southern Africa, population-based surveys conducted from 2015–2021 revealed that a quarter of adolescent girls aged 15–19 years tested for HIV in the last year, while only 17% of adolescent boys 15–19 years tested within the same timeframe [19]. These figures were even lower in central and western African countries, with only 13% adolescent girls and 3% of adolescent boys testing within the last year [19]. Our study findings contribute evidence that in sub-Saharan African adolescent boys consistently test for HIV less often than their older and female counterparts. In Ethiopia, female youth can engage in HIV testing as they enter childbearing age, as federal policies promote the integration of HIV testing and counseling into antenatal and prevention of mother-to-child transmission of HIV (PMTCT) services [13]. Male youth do not have a similar entry point. As adolescent boys reach young adulthood, undiagnosed HIV infection may lead to advanced HIV disease or to the unknown spread of HIV to their partners

Partner testing and non-marital sex partnership

The 2007 Ethiopian Federal Ministry of Health HIV Counseling and Testing (HCT) Guidelines emphasize the promotion of HIV counseling and testing to couples and women [13]. Couples are encouraged to test at ‘pre-engagement, pre-marital, or pre-conception’ stages, with partner notification services offered [13]. While partner testing should be encouraged, our findings show that being in a marital union does not significantly increase HIV testing among males in the 15–24 age group (Table 3). Previous studies have shown that partner-testing services do not necessarily capture men [20, 21]. This was seen in Kenya, where 88% of pregnant women receiving ANC care tested for HIV in the previous year, but only 5% of their male partners had tested within that same timeframe [22]. In Ethiopia, male youth are likely to debut sexually 2.5 years before entering marriage [12]; with males in urban areas marrying even later [12]. Given the proportion of male youth in urban Ethiopia delaying marriage and reporting sex with a non-marital, non-live in partner (Table 1), the Ethiopia national HIV program may consider implementing contract tracing and offering HIV testing for all sexual partners of HIV cases regardless of marital status.

Effective strategies to increase HIV testing among youth

Limitations

The implications of our analysis are limited by the nature in which the data were collected. As a household survey, HIV testing history was captured through self-report rather than from an electronic medical record. Our data is subject to recall bias as respondents may have forgotten if they have ever tested or misreported the date of their last HIV test. The data is also subject to social desirability bias. The age of sexual consent in Ethiopia is eighteen [15]. Youth younger than 18 years of age may not disclose their prior sexual activity or engagement in HIV testing services to avoid any cultural stigma associated with engaging in sex before the age of consent. Estimates for testing for HIV and sexual risk behavior are likely underreported as a result. Lastly, the inclusion of unemancipated minors (15–17 years) in our sample required that they be related to an eligible and consenting adult. Our findings may better characterize the HIV testing behaviors of emancipated minors, 15–17 years old. Emancipation status was not included as a possible predictor of testing due to our sample including youth who are and are not of legal age.