The study area falls within a health and demographic surveillance system (HDSS) covering approximately 260,000 Siaya County residents across three primarily rural, contiguous administrative areas: Gem, Karemo, and Asembo. HIV prevalence among adults 15–49 years in Siaya is 15.3%, three-fold higher than the national estimate of 4.9% ³³. The Siaya HDSS is described in detail elsewhere³⁴ Siaya county, including areas within the HDSS, has been the focus of a multi-country evaluation of the DREAMS program ^32,35, and a cluster randomized control trial for interventions aimed to improve sexual and reproductive health among adolescent girls and young women ^16,36.

In 2010, the Kenya Medical Research Institute (KEMRI) initiated home-based HIV testing and longitudinal behavioral surveillance approximately every two years (4 total rounds between 2010–2018) among a simple random sample of 6,809 compounds in the Gem area of the HDSS. A compound consists of physically co-located households in which immediate and extended family members reside. After the initial survey in 2010, subsequent HIV bio-behavioral survey (HBBS) rounds included compounds within the Gem HDSS to which an HBBS participant had moved. Consenting or assenting individuals aged 13 years and older, who were either HDSS residents (continuously resided in the HDSS for four calendar months) or who spent the night preceding the survey in a study compound, were eligible to participate. In 2018, HBBS was expanded to include a simple random sample of 5,930 additional compounds in the Gem HDSS to increase the representation of AGYW 13–24 years of age. The 2018 HBBS was conducted during January–December and included a total of 18,074 compounds, including 3,995 initially selected in 2010, 8,149 to which an HBBS participant had subsequently moved, and 5,930 for targeted AGYW enrollment. All interviews and HIV testing and counseling (HTC) services were conducted in-person, privately in the home using a pre-coded structured questionnaire. Participants were asked about their demographics, sexual behaviors, HIV testing history, perception of HIV risk, and access to HIV prevention, and care and treatment services. After completion of the survey, participants who did not have documentation of a previous positive HIV test (e.g. HIV clinic or outreach program card, antiretroviral prescription) were offered rapid HIV testing. Testing was done according to the Kenyan Ministry of Health testing guidelines³⁷, incuding a screening test using Determine (Alere, Orlando, FL, USA), which if positive, was followed by a confirmatory test using First Response (Premier Medical Corporation, Nadi Daman, India). Clients with a positive confirmatory test were referred to a health facility for retesting per the national guidelines. HIV testing and counseling was done by trained counselors. Participants newly identified as HIV-positive, or known HIV-positive but not receiving care, were linked to care at their preferred health clinic. Data were entered into the CommCare mobile app in the field using tablets and auto-synchronized to a secure cloud-based study server with end-to-end encryption.

This analysis included adolescent girls and young women 15–24 years of age who participated in the 2018 HBBS. This age group was chosen to align with previous national surveys and routinely collected HIV program data in Kenya^26,27. Outcomes of interest were HIV status and perception of risk for HIV acquisition. HIV status was dichotomized coded as “negative” for participants with a negative HIV test result, and “positive” for participants who had a positive test result, or documentation of known HIV positive status. Participants who reported having a previous positive or negative HIV test without documentation, or a previous indeterminate result, and declined testing, were classified as “unknown” and excluded from the analysis.

Participants were asked about their perceived chance of acquiring HIV, with possible responses: “no risk at all”,“small”, “moderate”, “great”, “already have HIV” or “don’t know”. For the analysis, responses were classified into either “high risk” for those who responded “great” or “moderate”; “low risk” for those who responded “no risk at all” or “small risk”. Analysis of HIV risk perception was restricted to participants who had ever had sex, and who tested HIV-negative or newly diagnosed HIV-positive at the time of the survey. Exposure variables included socio-demographic characteristics (age, education level attained, marital status, socioeconomic status, migration status, orphanhood); sexual behaviors (age of sexual debut, condom use, forced sex, intimate partner violence, transactional sex, number and age of sexual partner(s), partner circumcision status); risk perception; reproductive health history; and use of HIV prevention services (i.e. HTC, DREAMS).

Frequencies and percentages were calculated to describe participant characteristics. We calculated weighted prevalence estimates and assessed factors associated with HIV infection and high HIV risk perception using unadjusted and adjusted prevalence ratios. Individual-level weights were applied based on the inverse product of the probability of a compound being selected out of all the compounds in the study area, and the number of people interviewed per compound out of the total people residing in the same compound at the time of the visit. Weighted prevalence estimates of HIV infection, and high and low perception of HIV risk were calculated overall and for each of the covariates assessed. Generalized Linear Models were fitted to estimate unadjusted and adjusted (aPR) prevalence ratios and associated 95% confidence intervals (CI) using backward stepwise elimination. Variables associated with each outcome in bivariable analysis (p<0.20) were included in the variable selection for the multivariable model. Finally, prevalence of HIV-infection and high- and low-risk perception by the number of factors were estimated among sexually active AGYW. These factors were independently associated with HIV infection in the multivariate model. All analyses were performed using Stata statistical software (release 16, College Station, TX).

The study was reviewed and approved by Kenya Medical Research Institute Scientific and Ethics Review Unit (#1801). This study was also reviewed and approved by the United States Centres for Disease Control and Prevention (CDC) Institutional Review Board (#3308). Eligible adults aged 18 years and above and emancipated minors (persons below 18 years of age who are either pregnant, a parent, married, or heading a household) provided written informed consent before participation in the survey and HTC. Minors aged 13–17 years provided assent after parental consent was granted.

A total of 3,828 participants with known HIV status (i.e. known HIV-positive or tested during the survey) were included in the analysis. The majority, 63.0% (n=2,395), were aged 15–19 years; the median age was 18 [interquartile ranges (IQR) 16–21] years. Overall, 45.6% (1,742) of AGYW had completed a primary level of education or below; 26.5% (1,008) had ever been married; 2.9% (105) had newly (in past 4 months) in-migrated to the HDSS, and 11.4% (414) had returned to the HDSS after out-migrating. Nearly one quarter were orphaned, either paternally or maternally (18.6%), or both (5.1%). Overall, 95.9% (n=3670) had ever had an HIV test. Among the participants who reported having a previous negative test, 36.0% (n=1292) had been tested less than 3 months ago, 35.9% (n=1301) in the past 3 to 12 months, 13.5% (n=374) more than 12 months ago, and 14.6% (n=524) could not recall the date of the previous test. Nearly three quarters (71.9%) of AGYW had accessed HTC in the past 12 months, and approximately one-half, 52.3% (n=1,610), had accessed a DREAMS program service in the past 12 months (Supplemental table 1).

A minority,13.0% (244), of sexually active AGYW had a high/moderate perceived risk of acquiring HIV infection. Over half, 53.0% (2056), of AGYW had ever had sex. Sexually active participants reported the following behavioral risks: 15.0% (311) had their sexual debut at <15 years of age, and 44.5% (876) did not use a condom at their first sexual intercourse. A substantial proportion of AGYW had either been forced to have sex, 8.5% (161); or experienced partner violence, 13.4% (261); and, a small proportion, 1.9% (34), reported having an STI diagnosis in the past 12 months. In the past 12 months, the minority of sexually active participants, 8.3% (156), had multiple sexual partners; 10.2% (201) had a casual sexual partner; 13.2% (221) had an older (by ≥10 years) sexual partner, and 5.1% (80) had sex with a known HIV-positive partner . The majority, 70.7% (1241), reported that their primary male sexual partner was circumcised and 39.6% (697) reported having been given money/gifts/favors for sex in the past 12 months. The majority of participants, 73.0% (1,465), had ever been pregnant, and 61.0% (1121) had used contraception in the past 12 months. Overall prevalence of HIV infection was 4.5% (174).

Sociodemographic and behavioral risk factors associated with HIV infection among the 2,056 sexually active participants in the bivariable analysis are shown in Table 1 and Supplemental table 2. In the multivariable analysis, being aged 20–24 years (aPR 2.04, 95% CI 1.33–3.12); having been diagnosed with an STI (aPR 3.01, 95% CI 1.74–5.19); having an older sexual partner (aPR 1.45, 95% CI 1.02–2.08); and, having an HIV-positive partner in the past 12 months (aPR 9.09, 95% CI 6.18–13.37), or not knowing a partner’s HIV status (aPR 1.70, 95% CI 1.11–2.61), were significantly associated with HIV infection. Having secondary education ongoing/incomplete (aPR 0.57, 95% CI 0.37–0.86) or complete secondary/tertiary education (aPR 0.39, 95% CI 0.22–0.67) was associated with lower HIV prevalence (Table 1). There was no difference in HIV prevalence among AGYW who had, versus had not, participated in DREAMS program in the past 12 months (weighted prevalence 5.8% [95% CI 4.1 – 8.2] vs. 6.1% (4.3 – 8.4), respectively) (Supplemental table 2).

Overall, 16.7% (343) of sexually active AGYW had no other known risk factors, 38.2% (786) had 1 risk factor, and 29.5% (607) had 2 risk factors, 12.8% (264) had 3 risk factors, and 2.7% (56) had 4+ risk factors associated with HIV infection in the multivariable analysis. HIV prevalence increased with the number of risk factors; sexually active AGYW who reported having 0, 1, 2, 3, and 4+ factors had an HIV prevalence of 0.9% (95% CI 0.2–2.5), 3.3% (95% CI 2.2–4.8), 9.1% (95% CI 6.9–11.6), 17.4%(95% CI 13.0–22.5) and 35.7% (95% CI 23.4–49.6), respectively (Figure 2).

The majority (74%) of the 1658 sexually active AGYW who, at the time of the survey, did not have a known HIV-positive status, reported having low HIV risk perception; 13% (241) reported high-risk perception; and, 14% (267) did not know their perceived level of risk. Factors associated with high HIV risk perception in the bivariable analysis are shown in Table 1 and Supplemental table 3. In multivariable analysis, high perception of HIV risk was associated with not using a condom at first sex (aPR 1.61, 95% CI; 1.26–2.05), having been diagnosed with an STI (aPR 2.44, 95% CI; 1.30–4.58), having had anal sex (aPR 2.35, 95% CI; 1.33–4.14), having had multiple sexual partners (aPR 1.79, 95% CI; 1.31–2.47), an HIV-positive partner (aPR 4.38, 95% CI; 2.96–6.47), or a partner with unknown HIV status (aPR 1.73, 95% CI; 1.33–2.25) (Table 1).

Of the sexually active AGYW who did not have a previous HIV diagnosis, and who reported knowing their perceived risk for HIV, 12.6% (209) had no risk factors for HIV infection, 32.7% (542) had 1 risk factor, 28.2% (468) had 2 risk factors, 19.9 % (330) had 3 risk factors and 6.6% (109) had 4+ risk factors associated with HIV in the multivariable analysis. HIV risk perception increased with the number of HIV-infection risk factors; prevalence of high risk perception among sexually active AGYW who reported having 0, 1, 2, 3, and 4+ factors was 7.7% (95% CI 4.4–12.1), 9.2% (95% CI 6.9–12.0), 14.5% (95% CI 11.5–18.1), 22.1% (95% CI 17.8–27.8) and 31.2% (95% CI 22.7–40.8), respectively. Prevalence of low risk perception among sexually active AGYW who reported having 3 and 4+ factors was 77.9% (95% CI 73.0–82.2) and 68.8% (95% CI 59.2–77.3), respectively (Figure 3). Of the 267 sexually active AGYW who reported not knowing their perceived level of risk, 6.4% (17) had 0 risk factors for HIV infection, 24.7% (66) had 1 risk factor, 30.7% (82) had 2 risk factors, 26.2% (70) had 3 risk factors and 11.9% (32) had 4+ risk factors. Of the 21 AGYW newly diagnosed with HIV infection, 57.1% (12) had 3 or more risk factors for HIV infection; however, only 12.0% (3) reported having a high perceived risk of HIV infection; and 12.0% (3) didn’t know their perceived risk for HIV.