The study team analyzed a subset of the Kenya Population-based HIV Impact Assessment (KENPHIA) 2018–2019 focused on boys aged 11–14 years who reported knowing what sex is and who had heard of HIV; the final analytic sample included 495 participants. The authors constructed a five-item index to measure HIV-related knowledge, grouping participants into categories with 0–2, 3–4, or 5 correct responses. After adjusting for age, education level, and residence, boys in the intermediate (3–4 correct) and highest (5 correct) knowledge categories had a higher prevalence of reporting they could refuse unwanted sex compared with the lowest knowledge group. The adjusted prevalence ratios reported were aPR = 1.32 (95% CI: 1.11–1.57) for the 3–4 correct group and aPR = 1.29 (95% CI: 1.06–1.57) for the five-correct group. The authors also ran sensitivity analyses treating the knowledge score as a continuous variable; each additional correct response was associated with increased prevalence of perceived refusal ability (aPR = 1.08, 95% CI: 1.02–1.14), although the relationship departed from linearity according to their report.
Short sentences state the central point. Longer sentences explain that investigators reported a consistent association between higher HIV-related knowledge and perceived refusal skills, but that the effect size did not clearly increase from the 3–4 to the 5-correct category, suggesting a plateau in the categorical specification.
The outcome measuring intention to have sex before marriage was available for 408 boys in the analytic sample. The study team modeled this outcome with survey-weighted Poisson regression, adjusted for the same covariates. After adjustment for age, residence, and education, categorical knowledge levels were not associated with intention to have premarital sex. Sensitivity analyses that modeled the knowledge score continuously likewise showed no association with intention to have sex before marriage. In short: knowledge level did not predict declared intention to engage in premarital sex within this cross-sectional dataset.
This finding was consistent across adjusted models, according to the authors, and it emphasizes that knowledge alone—at least as captured by the five-item index used here—may not translate into differences in declared sexual intentions among early adolescent boys.
Investigators assessed whether boys had communicated with a parent or guardian about sex (n = 495). Prevalence of parent/guardian communication rose across knowledge categories in descriptive analyses, but the category-specific adjusted associations did not achieve statistical significance in the main categorical models. When the knowledge score was modeled as a continuous variable in sensitivity analyses, each additional correct response was associated with higher prevalence of parent/guardian communication (aPR = 1.28, 95% CI: 1.03–1.60). The authors advise caution in interpreting these mixed results because the categorical and continuous specifications yielded different inferences.
Short declarative sentences note the pattern. Longer ones explain that the data hint at a link between knowledge and caregiver communication, but the signal depends on how knowledge is operationalized.
The analytic approach used survey-weighted Poisson regression with a log link and jackknife variance estimation to estimate adjusted prevalence ratios and 95% confidence intervals. Models were adjusted for age, education level, and residence. The primary categorical exposure was the three-level knowledge index (0–2, 3–4, 5 correct); sensitivity analyses treated the score as continuous to test linear associations. The outcomes were measured in subsamples depending on response availability: intention to have sex before marriage (n = 408), perceived ability to refuse unwanted sex (n = 479), and parent/guardian communication about sex (n = 495).
Readers should note that the authors explicitly reported the modeling choices and the covariates included. They also acknowledged nonlinearity for the perceived refusal ability association in the continuous specification, which suggests the relationship may not be strictly proportional across the score range.
The authors flagged key limitations inherent to this secondary cross-sectional analysis of survey data. Causality cannot be inferred from the observed associations because exposure and outcomes were measured at the same time point. The five-item knowledge index and the binary or categorical outcome measures capture self-reported constructs; the investigators did not report objective behavioral validation in this dataset. Subsample sizes varied by outcome (for example, intention to have sex before marriage was available for 408 boys), which affects precision and interpretation. The authors also noted discrepant results between categorical and continuous specifications for parent/guardian communication and therefore urged cautious interpretation of those findings.
Finally, the study team recommended that age-specific HIV prevention strategies should combine knowledge dissemination with caregiver communication, life-skills training, safeguarding measures, and social support, and that such multi-component approaches require prospective evaluation to determine whether improved HIV-related knowledge leads to measurable changes in behaviors or in empowerment to refuse unwanted sex among early adolescents in Kenya.
The article is a preprint and has not undergone peer review; the authors declared no competing interests and reported appropriate ethical approvals for the parent survey and for the secondary analysis. Data access is available through the PHIA Data Manager per the data-use terms; the authors stated code can be made available on request.