Ethical approval was obtained from the Institutional Review Board (IRB) of Jimma University with reference No: JUHI/IRB 329/23, date 20/03/2023, which states that “the research protocol meets the ethical and scientific standards outlined in national and international guidelines” in accordance with the Declaration of Helsinki. A formal letter of support was obtained from Mattu University to their respective colleges requesting to cooperate with researchers during the study. Informed consent was obtained from all the participants based on “written consent form” prepared for this purpose and to be signed by all the participants before participating. As all university-level students are older than 18 years, they are believed to be capable of providing informed consent. In addition, the law in Ethiopia does not require such a group of young people to be accompanied by parents or guardians to provide consent to their behaviors, stating that “Assent will be sought from a study participant under the age of 18 years old” (NRERG 7th ed: Art 6.15). Before start, the study's objective was explained to the participants and voluntary participation was allowed by explaining their full rights to participate or not, and to withdraw their consent at any stage if they wish without giving any reason or no penalty. Furthermore, the participants were also assured that the questionnaire was anonymous and that their responses were fully confidential.

Various theoretical models have been used to understand HIV risk behaviors and promote safer sexual behaviors across at-risk populations. Of these, the information-motivation-behavioral skills (IMB) model (10–12) is one that has been widely used to understand and promote HIV prevention behaviors. The IMB model asserts that HIV prevention information, motivation, and behavioral skills are fundamental determinants of HIV preventive behaviors (10–12). The model specifies that information and motivation work through behavioral skills to influence HIV-preventive behaviors. The model also assumes that information and motivation may have a direct effect on HIV-preventive behavior when complicated behavioral skills are not required for practice, such as abstinence, as opposed to using condoms by young women (8–10). According to the IMB model, specification of information, motivation, and behavioral skills content most relevant to a specific behavior (abstinence and condom use) and identification of the IMB model constructs that powerfully influence the practice of a specific behavior is crucial for designing empirically targeted interventions to promote a specific sexual health behavior (10, 11, 13, 14).

The IMB model's constructs are also regarded as highly generalizable approach to understanding and promoting sexual health behavior across populations and behaviors of interest (9–11). However, despite the recommendation for its application to a range of sexual health behaviors, there is limited research using the IMB model for understanding predictors of safer sexual dual protection behaviors, such as dual protection use among young women at risk of STI/HIV and unwanted pregnancy.Thus, this study aimed to utilize the IMB model for examining predictors of dual protection use among female university students in Ethiopia for designing targeted interventions to promote dual protection behaviors for this population. The IMB model was chosen because it specifies measurement and statistical procedures for eliciting information, motivation, and behavioral skills factors that are relevant to a particular behaviors (15–18).

As illustrated in Figure 1, the IMB model for predicting dual protection use posits that an individual's levels of information, motivation, and behavioral skills are determinants of dual protection use. Specifically, possessing specific information about the risks of HIV/pregnancy and the benefits of dual protection, motivation to engage in dual protection (attitudes toward dual protection use and social norms regarding it) and behavioral skills for dual protection (the ability to negotiate with partners and self-efficacy in condom use) will directly predict dual protection use. Furthermore, information and motivation also influence behavioral skills, thereby indirectly affecting dual protection use (see Figure 1).

This study was conducted at Mattu University in Ethiopia. Mattu University is a newly established third-generation public university in the country. It has two campuses: the Mattu main campus, located at 551 km southwest of Addis Ababa, and Beddele campus, located at 431 km southwest of Addis Ababa.This study was conducted between April and December 2023.

A cross-sectional survey was used with undergraduate female university students randomly selected from the two campuses of Mattu University. The inclusion criterion includes being unmarried female students and willing to participate and provide informed consent. Whereas married female students, and those who were critically ill were excluded.

To obtain a sufficient number of participants for numerical analysis, the sample size for the study was calculated for different objectives, and the largest size was taken as the final sample size.n=2(zα/2+z1−β)2(μ1−μ2)2δ=2(1.96+1.28)2(Δδ)2=2(10.51)δ2=21(0.20)2=210.04=525×2=1050Thus, since 1,050 was the largest sample size calculated, it was taken as the final sample size to be used for all objectives, including this study.

A multistage sampling technique was used to recruit participants from the two campuses of Mattu University. First, three non-health colleges from each campus were selected using stratified sampling. Then, two departments from each college and two classrooms from each department were selected using cluster sampling in which all female students who fulfilled the inclusion criteria were invited to participate in the study.

To assess the levels of HIV risk reduction/dual protection information, motivation, behavioral skills, and behaviors, a validated measure of the IMB model of HIV prevention behaviors was was adopted from previous research (8, 10, 13, 14, 19, 20).

Levels of HIV risk reduction and dual protection information were measured with 12 items, using T/F responses. a) Knowledge about HIV risk and prevention was assessed with 8 items, such as (1) “HIV can be transmitted through sexual intercourse”; 2) “Abstinence is the most effective way of avoiding risk of STI/HIV”; or 3) “Consistent condom use is an effective method of preventing STI/HIV,”. b) Knowledge about safer sexual dual protection behaviors was assessed with four items, such as (1) “Abstinence is the best way to avoid the risk of HIV/STI and pregnancy for unmarried youth”; (2) “Consistent condom use prevents HIV/STI and pregnancy, providing dual protection against both”; and (3) “Dual methods use is options for dual protection against both HIV/STI and pregnancy simultaneously” (8, 10, 13, 14, 19, 20).

Motivation to practice dual protection use was assessed based on attitudes and social norms regarding dual protection, using 5 pionts Likert scale: a) Attitudes toward condom use were assessed with 4 items: (e.g., 1) “condoms do not interfere with sexual enjoyment,” 2) “consistent condom use is effective for preventing HIV/STI and pregnancy”, and 3) “condoms should always be used, even if a woman is using hormonal contraception.” b) Social norms about condom use were assessed with two items: (1) “The belief that condom use may show my partner that I do not love/trust him” and 2) “If I ask my partner to use condoms, he would suspect me that I have another partner and that he would become angry at me” (8, 10, 13, 14, 19, 20).

Behavioral skills for dual protection use were assessed based on perceived ability and Self-efficacy for dual protection, using 5 pionts Likert scale: a) Perceived ability to negotiate with one's partner was assessed with two items: (e.g., 1) “How confident are you in your ability to negotiate with partners to use condoms consistently?” and 2) “How confident are you in using condoms consistently even if you are using hormonal contraception?”). b) Self-efficacy in using condoms was assessed with two items: (e.g., 1) “If your steady partner doesn't want to use a condom, how certain are you that you could convince him to use condoms?”; 2) “If you are already using another method of contraception for pregnancy, how certain are you that you could always use condoms for reducing the risk of HIV?” (8, 10, 13, 14, 19, 20).

Sexual dual-protection behaviors were assessed using self-reported protection method used. 1)

Primary sexual abstinence was assessed by asking “Have you ever had sex to date?” with a response (yes or no).

Data collection was carried out in their classrooms, where students were asked to fill out an anonymous questionnaire prepared in English and then translated into Afaan-Oromoo and Amharic languages, for better understanding of the concepts of each question. The questionnaires were administered by 4 research teams who had a B.Sc degree, but not staff of the same university, and they were trained as supervisors to help students complete the questionnaire, to explain the purpose of the study, clarify questions whenever necessary, and collect the completed questionnaires at the end. A pre-test of the questionnaire was administered to 5% of the sample, with a total of 51 students, but not taking part in the study.

The collected data was cleaned, edited, and analyzed using SPSS version 23. Descriptive statistics were used to summarize the distribution of participants based on their demographic characteristics and the level of HIV risk reduction/dual protection information, motivation, behavioral skills, and behaviors. Bivariate and multivariate analysiswere performed using structural equation modeling (SEM) with AMOS 24 (17, 18) to examine correlates and predictors of dual-protection use. Spearman's correlations were used to examine correlations among the indicators of the IMB model and dual protection use. Estimating the parameters of the measurement model (confirmatory factor analysis) and the structural model (path analysis) was accomplished with a maximum likelihood estimate in AMOS. Confirmatory factor analysis (CFA) was used to examine the relationships between the observed variables and latent constructs in the model (i.e., to test the reliability and validity of the measurement model). Path analysis was used to examine the relationships between the latent constructs in the structural model (i.e., the information, motivation and behavioral skills constructs) and identify the predictors of dual protection use.

As shown in Table 3, results of Spearman's correlations analysis showed that the informational indicators of knowledge about HIV risk reduction and safer sexual dual protection behaviors were strongly correlated(r = .722, p < .001). In addition, the informational indicators are also correlated with the motivational indicators of attitudes towards dual protection(r = .597, p < .001) and social norms about dual protection use (r = .547, p < .001), and also with the behavioral skill indicators of self-efficacy in using dual protection (r = .508, p < .001).Similarly, the motivational indicators of attitudes toward dual protection use and social norms about dual protection were strongly correlated (r = .910, p < .001), and both are also correlated with the behavioral skill indicators of self-efficacy for dual protection use (r = .637, p < .001) and (r = .565, p < .001) respectively.

Overall, the bivariate analysis showed that dual protection use was strongly correlated with knowledge about HIV risk reduction (r = .332, p < 0.01), knowledge of safer sexual dual protection behaviors (r = .337, p < 0.01), attitude towards dual protection use (r = .477, p < 0.01), social norms about dual protection use (r = .372, p < 0.01) and perceived self-efficacy for using dual protection (r = .501, p < 0.001) (see Table 3).

The extent to which the data conformed to the theoretical model was assessed using the model's fit indices that revealed the Chi-squared test value (X2 = 11.399, P = .220), comparative fit index (CFI = .99), and root mean square error of approximation (RMSEA = .048), all indicating the model fit was acceptable.

As presented in Table 4, the maximum likelihood estimates of the measurement model showed that all the factor loadings for the relationships between the observed variables and latent constructs were strong and statistically significant, suggesting reliability of the model. Specifically, the standardized factor loadings (β = 0.89) for knowledge of HIV risk and prevention and (β = 0.82) for knowledge of safer sexual dual protection behaviors on the information construct suggest their strong and significant relationships. In addition, the standardized factor loadings (β = 1.1) for attitudes toward dual protection use and (β = 0.89) for social norms about dual protection on the motivation construct also suggest strong and significant relationships. Finally, the standardized factor loadings (β = 1.1) for perceived ability to negotiate safer sex and (β = 0.86) for self-efficacy in using dual protection on the behavioral skill construct indicated strong and significant reliability (P < 0.001) (see Table 4).

As shown in Table 5 and Figure 2, the results of path analysis indicated that information (β = 0.27, P < .001) and motivation (β = 0.40, P < .001) were significantly associated with behavioral skills, and that behavioral skills, in turn, were reliably associated with dual protection use (β = 0.27, P < 0.001). In addition, the direct effect of motivation on dual protection use was strong (β = 0.29, P < 0.001), whereas the direct effect of information on dual protection use was moderate (β = 0.04, P < 0.05). Overall, the full model's constructs accounted for approximately 27.4% of the variation in dual-protection use, while the information and motivation constructs explained about 37% of the variation in behavioral skills (see Table 5 and Figure 2).

This study has several limitations. First, the study focused only on female university students for the purpose of examining patterns and predictors of dual protection use to simultaneously prevent both STI/HIV and unwanted pregnancy, excluding male students for priority concern. Second, the data based on self-reported sexual behaviors may lead to bias arising from socially desirable responses, because sex remains a sensitive topic in Ethiopian culture. However, the use of self adminstared data collection methods and anounmous questioniiares is intended to reduce such bias.Third, the data from the cross-sectional survey can reveal only the associations between variables at a single point in time; thus, it is recommended that future research should focus on longitudinal studies to test the model.