The Korea Youth Risk Behavior Web-Based Survey (KYRBS) is a nationally representative population-based repeated cross-sectional survey that investigates health-related behaviors such as tobacco and alcohol use, mental health, and sexual behaviors among adolescents across all 17 provinces in South Korea (aged 12–18, grades 7–12). The participants were selected using nationwide stratified multistage random cluster sampling to represent middle and high school students across the country. The target population of the KYRBS consists of all middle and high school students in South Korea for a given year. Initially, the population was divided into 117 strata using regions and school types as stratification variables. The number of sample schools was allocated using a proportional distribution method based on the population composition. Stratified cluster sampling was employed, with the school as the primary sampling unit and the class as the secondary sampling unit. All students in the selected classes were surveyed, excluding those with long-term absenteeism, special needs, or dyslexia. Staff responsible for the sample schools received related training prior to the survey. On the survey day, students were escorted to the school’s computer lab and randomly assigned to a computer. As the KYRBS is an anonymous, self-administered online survey, the supervising staff informed the participants before the survey began that the results would not be shared with anyone. The complex survey weight used in this study, provided by the Korea Centers for Disease Control and Prevention, is the product of the inverse of the sampling rate and the inverse of the response rate, multiplied by the post-adjustment rate. The KYRBS response rates were The KYRBS response rates were 90.9%, 94.8%, 95.1%, 97.6%, 97.7%, 95.5%, and 96.4% from 2006 to 2012 [25].

Among the 519,473 participants, we excluded 3 who had no data on suicide attempts (n = 519,470). We also excluded 4,223 participants who denied to provide their demographic information including age (n = 515,247). A detailed description of the KYRBS and its measurements can be found elsewhere [25, 26].

The primary outcome was the occurrence of suicide attempts over the past year. Suicide attempts were assessed using the following questions: “During the past 12 months, did you ever actually attempt suicide?” [27] Participants provided binary responses of either “yes” or “no” in response to those questions. We also used depressive episodes and suicidal ideation as secondary outcome variables. Depressive episodes were assessed using the following question: “During the last 12 months, did you ever have feelings of sadness or despair that have interrupted your daily life for at least 2 weeks?” Suicidal ideation was assessed using the following question: “During the past 12 months, have you ever seriously considered attempting suicide?” The KYRBS has used these measurement methods for 16 years (2005–2020) to monitor mental health in Korean adolescents [26]. Other national epidemiological investigations such as the Korea National Health and Nutrition Examination Survey and the Korean Community Health Survey have adopted the same questionnaires to monitor depression and suicidality in the Korean population [28, 29].

Participants self-reported sexual violence perpetration and victimization in the survey. Participants who answered “yes” to the single question, “Have you ever sexually perpetrated another” were classified as sexual violence perpetrators; those who did not were classified as the reference group. Sexual violence victimization was measured using the question, “Have you ever been sexually victimized by another?” for the yes and no responses, respectively. A similar single-question assessment of sexual violence was used in the Youth Risk Behavior Survey in the United States [30]. In South Korea, where adolescent sexuality remains a taboo subject, discussions and recognition of sexual violence have been limited. From 2006 to 2012, perceptions of sexual violence in South Korea were likely confined to illegal coercive intercourse by non-partners, potentially overlooking “less violent” offenses such as sexual harassment or stalking, which may not have been widely recognized as sexual violence perpetration.

We used pre-specified potential confounders for the multivariate regression model. We confirmed the empirical associations between sexual violence perpetration, victimization, and adolescent suicide attempts using univariate regression model. The covariates were age, urbanicity, study year, household socioeconomic status (SES), residential status, body mass index (BMI), cigarette smoking, alcohol consumption, drug use, physical activity, insufficient sleep, subjective health, subjective happiness, premature puberty, early sexual debut, sexual orientation, and sexually transmitted diseases. The age ranged from 12 to 18 and was binarized into 12–15 and 16–18 years to reflect developmental stages [31]. The urbanicity of participants’ residence was determined by their location. The study year ranged from 2006 to 2012 and was automatically encoded by the survey system. All other variables were measured by self-report. The household socioeconomic status was assessed on a five-point scale (very low, low, middle, high, and very high) and the categorized into low, middle and high group to maintain a sufficient sample size. The residential status was an indicator of where and who the participant lives with (living with family, relatives, friends or alone, and facility such as an orphanage). The BMI was calculated based on self-reported height and weight, and categorization was made on the basis of gender- and age-specific percentiles according to Korean growth standards [32]. The cigarette smoking was defined as one or more days of cigarette smoking during the past 30 days. The alcohol consumption was defined as having had one or more drinks in the last 30 days. The drug use was defined as ever using habitual and intentional drug in lifetime. The physical activity was defined as at least 30 min of moderate-intensity exercise on three or more days per week, or strength training on three or more days per week in the past week. The insufficient sleep was assessed on a five-point (very sufficient, sufficient, average, insufficient, and very insufficient) and was defined as having insufficient or very insufficient sleep in the past week. The poor subjective health and happiness were assessed on a five-point (strongly agree, agree, neutral, disagree, and strongly disagree) and were defined as disagreeing or strongly disagreeing with feeling healthy/happy normally. The sexual orientation was defined based on whether there was sexual contact with a same-sex partner (same-sex orientation), an opposite-sex partner (opposite-sex orientation), or unknown if there was no sexual contact. The early sexual debut was defined as having first sexual intercourse before age 12. The premature puberty was defined as having experienced nocturnal emissions (in the case of boys) or menarche (in the case of girls) before the age of 12 years old. The history of sexual transmitted disease was defined as ever having sexual transmitted disease in lifetime.

All analyses were conducted separately for boys and girls. The 1-year prevalence and standard errors of suicide attempts and sexual violence were calculated using survey weights to represent the entire population of Korean adolescents. Based on the KYRBS data, the prevalence of suicide attempts was 4.8%, thereby satisfying the rarity assumption that enables the approximation of odds ratios (ORs) to relative risk [33]. While applying the survey weights, we calculated the ORs for suicide attempts between adolescents with and without experiences of sexual violence perpetration using multivariate logistic regression models. The associations between sexual violence victimization and suicide attempts were also analyzed using the multivariate logistic regression models. The study year was excluded from the pre-specified covariate set as it has no empirical relationship to either exposures or outcomes. We sequentially introduced covariates in the models to check for confounding effects on the estimated associations, and we presented the results in the Supplementary Results.

Interaction analyses were conducted on both an additive and multiplicative scale. We used An additive interaction occurs when the combined effect of two exposures is larger (or smaller) than the sum of its individual effects; an interaction on a multiplicative scale refers to a combined effect that is greater (or smaller) than the product of the individual effects [34]. Additive interactions were assessed using the relative excess risk due to interaction (RERI). The positive additive interaction was indicated by RERI values > 0, while the opposite indicated a negative additive interaction [35]. Multiplicative interactions were estimated using the product of perpetration and victimization in the outcome analysis. The positive multiplicative interaction was indicated by the ratio of ORs values > 1, while the opposite indicated a negative multiplicative interaction.

Using the same methodology, we conducted an additional analysis using depressive episodes and suicidal ideation as secondary outcome variables. We also calculated the E-value, which can be used to assess the robustness of the identified associations to the potential unmeasured confounders [36]. We applied Firth’s penalized logistic regression models as a sensitivity analysis to minimize potential bias from the rare exposure and outcome, despite not encountering any non-convergence issues in the main analysis [37]. Statistical analyses were conducted using SAS software version 9.4.