Effects of a Sexual Risk-Reduction Intervention for Teenagers: A Cluster-Randomized Control Trial

In the present cluster-randomized trial, 2047 teenagers aged 12–19 years initially participated. The participants were from 11 cities in central and coastal Colombia, corresponding to the urban and rural areas of the country. All the participants were authorized by their parents and/or guardians, who signed an informed consent form, to participate in the study. Only students between the 8th and 11th grades of high school participated. The participants were required not to have previously participated in formal sex education programs. The study excluded the participants who did not provide consent and those whose legal guardians did not provide consent. The exclusion also applied to those who, despite attending the evaluation phases, did not participate in the group sessions. The sociodemographic characteristics of the participants in each group are described in Table 1, and Fig. 1 shows data on the sample recruitment process and the percentage of dropouts at each stage of the project.

The project was submitted to the ethics and scientific committee of the participating universities, and upon obtaining approval, the search began for educational institutions willing to participate voluntarily. Subsequently, informed consent was requested from the legal guardians of the underage participants, as well as written consent from the teenagers. The schools were selected to meet similar conditions in terms of in-person attendance, mixed-gender education, public or private institution, and the socioeconomic status of the institution. The schools were randomly assigned to either the control or the experimental group.

Both the experimental and control groups were administered a list of evaluative questionnaires as a group in the first week. To preserve confidentiality, each participant was assigned an identification code to link the data of each participant in the different phases of the process. Only the person responsible for the study had correspondence information between the code and each participant’s data. The experimental group participants received the program, which involved one group session per week, with 25–30 participants. The experimental group participants received the program in person at their educational institution. The sessions were held during class hours, varying the schedule so that the students did not experience conflicts with other classes. Each group had a single researcher throughout the program who facilitated five intervention sessions of one hour once per week.

Participation in each of the sessions was voluntary; however, 85.4% of the experimental group participants attended all five sessions. The control group participants received no intervention during the 5 weeks of intervention in the experimental group. Once this time period was completed, the participants completed a posttest and a follow-up test. Once the research process was completed, a five-session intervention was carried out for those who wanted to participate, adhering to the ethical principles of research with human participants.

The program implementers had to (a) have a minimum degree in psychology; and (b) have received training in sexual and reproductive health, specifically in the application of the COMPAS program. This training was complemented by discussion after each session to resolve possible doubts and discuss them with the project coordinator. As the program is highly structured, the implementers had to strictly follow the application manual in all sessions.

After the initial evaluation week, the control group participants had a period of 5 weeks in which they did not receive any treatment. In week 7, the posttest was applied; the follow-up test was applied 6 months later. Subsequently, the COMPAS program was applied to the control group (following the principle of beneficence). The participants did not receive any reward for their participation in this study.

This study proposes a comprehensive view to evaluate the effectiveness of the COMPAS program. To this effect, we have included socio-demographic and behavioural measures, as well as safe sex predictors.

The characteristics of the sample were analyzed using descriptive statistics. Baseline differences between the experimental and control groups were analyzed using the t-test for quantitative variables (numerical) and cross-tables (χ²) for categorical variables. An intent-to-treat approach was used; therefore, data from all participants were analyzed, regardless of the level of intervention received [37]. The attrition rate in the posttest and the 6-month follow-up was calculated. Differences in sociodemographic variables and main outcomes (e.g., condom use) between participants who dropped out and those who were retained were also investigated using t-tests and cross-tables. When differences were statistically significant, Cohen’s d effect size was estimated. The coefficients were interpreted as follows: 0.20 = small, 0.50 = medium, and 0.80 = large (Cohen, 1988). Following previous studies (e.g., [21]), intervention effects compared to the control group were analyzed using generalized estimating equations (GEE). Values were adjusted for baseline differences, participants’ age, gender, and socioeconomic level. The GEE models are valuable for evaluating interventions in cluster-randomized control trials because they control for correlations among responses when the sample is clustered within schools [38]. The unit of randomization was the school, while the individual was the unit of analysis. Consequently, the school variable was controlled for in all the analyses. All analyses were performed using IBM SPSS Statistics for Windows, version 26.

Figure 1 shows the flow of participants throughout the study. In the posttest, there was no statistically significant difference in the loss of participants between the two experimental conditions (χ² = 1.78, p = 0.18). However, teenagers who dropped out were older (p < 0.001; d = 0.44), and a higher proportion were sexually experienced (χ² = 13.84, p < 0.001; OR = 0.65[.52, 0.81]) and women (χ² = 6.35, p < 0.01; OR = 0.75 (0.60, 0.94) than those who remained in it. Consistent condom use and the percentage of condom use were unrelated to the loss of participants.

At the 6-month follow-up, there were significant differences between the experimental conditions (χ² = 36.29, p < 0.001; OR = 0.58 [0.48, 0.69]). A higher proportion of participants in the COMPAS program condition dropped out compared to the control group. Teenagers who dropped out during follow-up were older (p < 0.001; d = 0.51), and a higher proportion were sexually experienced (χ² = 19.48, p < 0.001; OR = 0.66 (.55, 0.79)) than those who remained in it. The participants’ gender, consistent condom use, and percentage of condom use were unrelated to the loss of participants in the follow-up.

The mean age was 15.24 years (SD= 1.37; range = 12–19 years), with a significant difference in this variable between the groups, although with a small effect size (p < 0.001; d=0.32); the experimental group (EG) had a higher mean age. A total of 52.1% of the participants were women, and their parents were married in 34.4% of the cases. Socioeconomic level was also a contrasting variable between the groups, with significant differences (p<0.001; d=0.12); most of the participants belonged to stratum 2 and 3, indicating a prevalence of medium-small purchasing power in both groups.

Regarding the psychosexual characteristics of the adolescents, 89% of the sample identified themselves as exclusively heterosexual, and only 0.27% of the participants had children. Regarding sexual characteristics, 54% of the adolescents were sexually active and used condoms 31% of the time, with only 19% of the sample consistently using condoms. The age of initiation of sexual intercourse was evaluated, finding that the mean age was 14.43 years old (SD=1.56) for vaginal penetration, 14.36 years old (SD=1.73) for anal penetration, 14.35 years old (SD=1.89) for masturbation, 14.34 years old (SD=1.79) for oral sex, and 13.28 years old (SD=1.8) for dry humping (defined as consensual genital stimulation through clothing). Table 1 summarizes the characteristics of the participants by group at the beginning of the study.

In this phase, the effect of the intervention continued to show a change in 13 of the 17 variables analyzed, two fewer than in the posttest. A decrease was found in behavioral intention to use condoms and in attitudes toward HIV, specifically those related to HIV-protective behaviors when faced with obstacles. However, most psychological and behavioral variables showed significantly positive changes from baseline measures.

Despite the good indicators shown by this research, this study is not free of limitations, which must be considered when interpreting the results. First, although the program was applied in 11 different cities, Colombia is a country with extensive multiculturalism and high rates of economic, educational [51], and educational inequalities [52], rendering some populations more vulnerable to events such as STDs, sexual abuse [53, 54], and unplanned pregnancies [55]. Therefore, the results of this study cannot be generalized to all teenagers in the country. Another limitation was the differences in some sociodemographic variables between the EG and CG, specifically the mean age, which was higher in the EG, and the socioeconomic level (stratum), which was higher in the CG. Furthermore, the teenagers who participated in the study were mostly heterosexual, which leaves the effectiveness of the program among LGBT populations, who have shown greater vulnerability to sexual risk, unaddressed [56]. Finally, we also recognize that there were high attrition rates, a phenomenon usually reported in longitudinal studies [57]. However, these attrition rates did not differ significantly between the groups in the initial phases. Although at the 6-month follow-up phase there were higher attrition rates in the EG, the variables of participant gender, consistent condom use, and percentage of condom use were not related to participant attrition. Future research should evaluate the long-term effectiveness of the program in Colombia, as well as identify the mediating variables of program effectiveness, such as fidelity of implementation, the participants’ individual characteristics [47], and the program facilitators’ individual characteristics [11, 19]. Finally, it is important that the effectiveness of the program be tested in vulnerable populations in different regions of Colombia, perhaps using technological tools to increase program coverage in remote areas. It is hoped that research in these avenues could lead to a decrease in the rate of STIs and unplanned pregnancies among Colombian adolescents.