Abstract

Early coital debut is a risk factor for HIV. In this paper we investigate the predictors of young adolescents' transition to first intercourse using a social cognition theoretical framework. The analyses reported here were based on a longitudinal study of 2360 students in the schools allocated to the control arm of a cluster-randomized controlled trial to investigate the effect of a school-based HIV prevention programme among Grade 8 students in Cape Town. Structural equation modelling was performed with Mplus version 3.11. Of the 1440 students who were virgins at baseline, 1144 remained virgins 15 months later and 296 (20.6%) reported having had their first sexual intercourse. Transition to first sexual intercourse was more likely among males than females, among older students and among students with a lower socio-economic status. Transition to first sexual intercourse was significantly associated with intentions to have sexual intercourse, poor self-efficacy to negotiate delayed sex and intimate partner violence. The model predicted 35% of the variance in intentions and 16% of the variance in transition. These findings indicate some of the factors that influence young adolescent's transition to first intercourse and that need to be addressed when designing effective interventions.

Introduction

Early coital debut, defined as 15 years of age or younger, has been shown to be a risk factor for multiple sexual partnerships [1] and HIV infection [2, 3]. In South Africa in 2002, 7.8% of females and 17.5% of males aged 15–24 years reported an early coital debut [4]. Once they become sexually active at an early age, few young people report consistent condom use, and inconsistent condom users are more likely to be infected with HIV [4]. The delay of sexual debut has been advocated as an important HIV prevention strategy for young people [5], but to design effective strategies, we need to identify the key constructs, processes and mechanisms that influence young adolescent's transition to first intercourse.

The theory of planned behaviour (TPB) [6, 7] is a social cognition theoretical approach which has been extensively and successfully applied to the prediction of condom use in developed countries [8, 9] and has also been shown to be applicable in a range of other domains, including substance use, and engaging in physical activities and sports [10, 11]. TPB proposes that one's behavioural intentions predict one's behaviour. Intentions are in turn influenced by three conceptually independent influences. ‘Personal attitude’ refers to a tendency to respond favourably or unfavourably to an object or behaviour. This tendency is moulded by perceptions about the probability that the behaviour will lead to particular outcomes and whether the outcomes are evaluated positively or negatively. ‘Subjective norms’ encompass the perception of whether significant others think one should or should not perform the behaviour and the extent to which one is motivated to comply. ‘Perceived behavioural control’ refers to the perceived ease or difficulty of performing the behaviour [7]. A concept which comes close to perceived behavioural control is self-efficacy [12], which is the subjective probability that one is capable of executing a certain course of action. This theory's relevance to the understanding of HIV risk behaviour among adolescents in sub-Saharan African countries has been suggested from cross-sectional studies [13–18]. However, there is a paucity of longitudinal studies among young adolescents in the region and anywhere in the world [1]. Our choice of constructs to measure and our construction of scales were guided by the TPB and related conceptual models.

Violence is a conspicuous feature of adolescent sexual relationships in South Africa [19, 20] and a potential barrier between ‘healthy’ intentions and healthy sexual behaviour. For example, in 2002, 14% of male and 13% of female South African high school students reported having been assaulted by their girlfriend or boyfriend (respectively) in the 6 months preceding the survey [21]. Several qualitative and cross-sectional studies suggest that, for adolescent girls, violence in intimate relationships plays an important role in the transition to first sexual intercourse [19, 22–24]. It is plausible that intimate partner violence plays a role in adolescent boys' transition to first intercourse too, especially considering that they report high levels of assault by intimate partners and coerced sex [21, 25, 26].

We conducted a longitudinal study investigating the factors predicting transition to first sexual intercourse among Grade 8 high school students in Cape Town. We used a conceptual model closely related to the TPB [27] to conceptualize the influences on early sexual debut, and we investigated the extent to which the experience of being a victim of physical violence predicted early sexual debut.

Methods

This study was part of a larger multi-site prospective panel study, the South Africa Tanzania project, which has been described elsewhere [28]. In each of three sites, Cape Town, Mankweng and Dar es Salaam, we conducted a cluster-randomized controlled trial to investigate the effect of school-based HIV prevention programmes among Grade 8 students (12–14 years at baseline). The research took place during March 2004 to May 2005.

Sample

All high schools in Cape Town were stratified by postal (zip) code groupings, an indicator of demographic factors such as socio-economic level, race and culture. Fifteen schools were randomly selected from a set of 39 randomly selected schools that have participated in a number of previous projects carried out by members of the research team [29]. These schools were then matched in pairs to 15 other schools on a range of demographic characteristics (student population size, ‘race’ of students, language of instruction, geographical area and amount of school fees paid as a proxy for the socio-economic status). For each pair, one school was randomly assigned to the intervention arm of the trial and the other to the control arm. All Grade 8 students in the included schools were invited to participate in the study. The analyses reported here are based on the cohort of students in the schools allocated to the control arm of the trial in Cape Town and who participated in the baseline data collection (N = 2360). We left out the schools in the intervention arm to exclude the effects of the HIV intervention on the dependent variable.

Measurement and variables

The baseline survey was conducted in February to March 2004 and the second follow-up survey in March to April 2005. At each time, a self-completed anonymous electronic questionnaire was administered to students in a choice of two of the three languages spoken in Cape Town, using handheld computers. The instrument was subjected to a test–retest reliability study during 2003 [30] and the internal consistency and test–retest correlations of the scales were adequate. In addition, we compared the reliability of the electronic questionnaires with paper questionnaires and found that the intra-scale reliability and the test–retest correlations of the scales for each data collection method were adequate and similar [30].

The scales used in the measurement of social cognition constructs are described in Table I. The items measuring exposure to physical violence in intimate relationships were derived from the Gender Violence Conflict Tactics Scale [31] and included questions about being beaten, punched with something that could hurt, threatened with a knife or other weapon or a knife or other weapon being used against the respondent by a boyfriend or girlfriend. These four variables combined into one dichotomy (0 = not exposed to such violence; 1 = exposed). Socio-economic status was composed into a standardized sum score based on six standardized variables: number of assets (TV, electricity, bicycle, tap water and car), number of people sleeping in the same room (reversed), a subjective assessment of the material situation of the family (five categories from scarcity of food to luxury), father's and mother's level of education and living in a ‘shack’ (dichotomy). Intention was measured by the extent of agreement or disagreement to the statement: ‘I plan to have sex in the next six months.’ The dependent variable, ‘transition to first intercourse’ was based on questions about sexual intercourse (vaginal or anal) at baseline and at the second follow-up data collection. All variables except transition to non-virginity were based on students' baseline reports.

Table I.

Overview of the scales used to measure the theoretical constructs/latent independent variables

Latent variable Number of items Sample item Response format and scoring αa 
HIV/AIDS knowledge 14 Can a person who looks healthy, but has HIV pass the virus on to other people through unprotected sexual intercourse? 3 response categories (yes, no, don't know). Scored 1 for each correct answer 0.66 
Negative social outcome expectancies related to not having sexual intercourseb If I had a girl/boyfriend and I did not have sexual intercourse with her/him, she/he will think that I do not love her 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.74 
Severity of threats from consequences of unprotected sex (HIV/AIDS, STDs and pregnancy)b HIV is a big threat against my personal health 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.69 
Susceptibility to consequences of unprotected sex (HIV/AIDS, STDs and pregnancy)b I am likely to get HIV infected if I have sex without using condom 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.64 
Subjective norm favouring delayed sexual transition One item on descriptive norms: ‘Most of my friends do not plan to have sex until they are older.’ Four items on injunctive norms such as: ‘Most of my friends think that one has to be older before having sex.’ 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.65 
Self-efficacy as regards saying no to sex I would be able to refuse to have sex if I didn't feel like it. 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.79 
Intentions to remain a virgin I plan to have sexual intercourse within the next 6 months. 5 response categories (1 = strongly disagree to 5 = strongly agree) NA 
Latent variable Number of items Sample item Response format and scoring αa 
HIV/AIDS knowledge 14 Can a person who looks healthy, but has HIV pass the virus on to other people through unprotected sexual intercourse? 3 response categories (yes, no, don't know). Scored 1 for each correct answer 0.66 
Negative social outcome expectancies related to not having sexual intercourseb If I had a girl/boyfriend and I did not have sexual intercourse with her/him, she/he will think that I do not love her 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.74 
Severity of threats from consequences of unprotected sex (HIV/AIDS, STDs and pregnancy)b HIV is a big threat against my personal health 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.69 
Susceptibility to consequences of unprotected sex (HIV/AIDS, STDs and pregnancy)b I am likely to get HIV infected if I have sex without using condom 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.64 
Subjective norm favouring delayed sexual transition One item on descriptive norms: ‘Most of my friends do not plan to have sex until they are older.’ Four items on injunctive norms such as: ‘Most of my friends think that one has to be older before having sex.’ 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.65 
Self-efficacy as regards saying no to sex I would be able to refuse to have sex if I didn't feel like it. 5 response categories (1 = strongly disagree to 5 = strongly agree) 0.79 
Intentions to remain a virgin I plan to have sexual intercourse within the next 6 months. 5 response categories (1 = strongly disagree to 5 = strongly agree) NA 

NA, not available; STD, sexually transmitted disease.

a

Alpha statistic derived from the baseline survey.

b

One of the three latent variables used to measure the ‘attitude’ construct.

Procedure

A letter was sent to parents to describe the study and they could complete a declination form to prevent the participation of their child. Participating students signed assent forms. We used an anonymous system of linking individual student's data over time, described elsewhere [28]. Ethical clearance was provided by the Western Norway Regional Committee for Medical Research Ethics and the Research Ethics Committee of the Faculty of Health Sciences, University of Cape Town.

Measurement models

Reflective measurement models were assumed for the following scales: severity of threats, susceptibility, social outcome expectancies and self-efficacy. For all other multi-item scales, additive mean scores or formative indices were constructed. Measurement models were tested out initially. For some scales, and obviously due to similar wording of items, measurement models could be slightly improved by allowing correlated error terms. Since the fit of the overall measurement model was adequate even without allowing error terms to correlate [Comparative Fit Index (CFI) = .937, Tucker-Lewis index (TLI) = .926, Root Mean Square Error of Approximation (RMSEA) = .045 and Standardized Root Mean Square Residual (SRMR) = .044], it was decided to run all analyses without any correlated error terms.

Statistical analysis

We needed statistical techniques which could handle control for the cluster design and analysis of latent variables. This combination of requirements is met by the Mplus statistical software [32]. Frequency distributions and cross-tabulations were done with SPSS version 13.0. For statistical testing of cross-tabulations, the complex design analysis of two-way frequency tables was used in order to control for the cluster effect. Structural equation modelling including testing of measurement models was carried out with Mplus version 3.11. In Mplus, maximum likelihood parameter estimates (MLR) were applied. It was assumed that the outcome variables (transition intentions and transition) were metric. The Satorra–Bentler test [33] was used for testing of differences in fit between nested models. In order to ease interpretation of regression coefficients involving dichotomous predictors (gender and exposure to violence), the standardized regression coefficients were divided by the standard deviation of the predictors. After this adjustment, one unit change on dichotomous predictors can be interpreted as the magnitude of change associated with moving from one category to the other.

Role of the funding source

The study sponsors had no role in the study design; collection, analysis and interpretation of data; in the writing of the report or in the decision to submit the paper for publication.

Results

Two of the 15 intervention schools declined to participate in the randomized controlled trial, and we therefore excluded the two control schools matched to these schools. This left us with 13 control schools. The total number of eligible students in the 13 participating schools at baseline was 3118. The number of students who actually participated in the baseline survey was 2360. The participation rate was 76%. The attrition was caused by students not permitted by their parents to participate in the study and by students not present in class during the data collection.

Of the 2360 participating students, 396 students were excluded from these analyses because they reported at baseline they had already had their first vaginal or anal intercourse. Another 524 students were excluded because there was no valid information about vaginal or anal intercourse at either baseline or second follow-up data collection.

Among the 1440 remaining students, 836 were girls and 604 were boys. While 1144 remained virgins at the second follow-up data collection, 296 (20.6%) reported to have had their first intercourse (vaginal or anal) by the second follow-up. Among girls, 108 (12.9%) made the transition to first intercourse, while among boys, 188 (31.1%) made the transition. The transition to vaginal intercourse was reported by 87 (10.4%) girls and 147 (24.3%) boys. The transition to anal intercourse was reported by 38 (4.5%) girls and 95 (15.7%) boys.

Table II shows the baseline descriptive data for girls and boys, stratified by transition to first intercourse. At baseline, 76 (9.1%) girls and 142 (23.5%) boys reported being the victim of physical violence in intimate relationships. Among those who made the transition to first intercourse, 52 (27.8%) boys and 19 (17.8%) girls reported that they had ever been forced to have sexual intercourse against their will. Of those boys and girls who reported at baseline being victims of physical violence from intimate partners, 32 (35.6%) reported at second follow-up ever being forced to have intercourse, compared with 38 (18.9%) of those who were not victims of physical violence at baseline.

Table II.

Baseline descriptive data for girls and boys, stratified by whether they made the transition to first intercourse during the 15 months subsequent to the baseline survey

 Girls
 
Boys
 
 Transition to non-virginity?
 
Transition to non-virginity?
 
 Yes (n = 108) No (n = 728) Yes (n = 188) No (n = 416) 
Covariates 
    Age, mean (SD) 13.2 (3.4) 13.0 (0.9) 13.6 (2.2) 13.0 (2.4) 
    SES, mean (SD) −0.7 (1.1) 0.1 (1.0) −0.4 (1.1) 0.2 (0.9) 
Social cognition constructs 
    Knowledge, mean (SD) 8.1 (2.3) 8.5 (3.0) 7.9 (2.9) 8.8 (2.8) 
    Social (negative) outcome expectancies related to not having intercourse, mean (SD) 2.7 (1.0) 2.4 (1.0) 2.9 (1.0) 2.5 (1.0) 
    Severity of threats from consequences of unprotected sex, mean (SD) 4.1 (0.7) 4.3 (0.6) 4.2 (0.8) 4.3 (0.7) 
    Susceptibility to consequences of unprotected sex, mean (SD) 4.1 (0.8) 4.2 (0.8) 4.1 (0.8) 4.2 (0.8) 
    Subjective norm favouring delayed sexual transition, mean (SD) 3.9 (0.7) 4.3 (0,6) 3.8 (0.8) 3.9 (0.7) 
    Self-efficacy as regards saying no to sex, mean (SD) 3.9 (0.6) 4.2 (0.7) 3.8 (0.8) 4.0 (0.7) 
Intentions to remain a virgin for next 6 months, n (%) 
    Intend to remain a virgin 52 (49.5) 534 (74.1) 84 (45.6) 265 (64.5) 
    Uncertain 18 (17.1) 107 (14.8) 28 (15.2) 60 (14.6) 
    Intend to have sexual debut 35 (33.3) 80 (11.1) 72 (39.1) 86 (21) 
Victim of sexual violence, n (%) 28 (26.2) 48 (6.7) 63 (34.1) 79 (19.3) 
 Girls
 
Boys
 
 Transition to non-virginity?
 
Transition to non-virginity?
 
 Yes (n = 108) No (n = 728) Yes (n = 188) No (n = 416) 
Covariates 
    Age, mean (SD) 13.2 (3.4) 13.0 (0.9) 13.6 (2.2) 13.0 (2.4) 
    SES, mean (SD) −0.7 (1.1) 0.1 (1.0) −0.4 (1.1) 0.2 (0.9) 
Social cognition constructs 
    Knowledge, mean (SD) 8.1 (2.3) 8.5 (3.0) 7.9 (2.9) 8.8 (2.8) 
    Social (negative) outcome expectancies related to not having intercourse, mean (SD) 2.7 (1.0) 2.4 (1.0) 2.9 (1.0) 2.5 (1.0) 
    Severity of threats from consequences of unprotected sex, mean (SD) 4.1 (0.7) 4.3 (0.6) 4.2 (0.8) 4.3 (0.7) 
    Susceptibility to consequences of unprotected sex, mean (SD) 4.1 (0.8) 4.2 (0.8) 4.1 (0.8) 4.2 (0.8) 
    Subjective norm favouring delayed sexual transition, mean (SD) 3.9 (0.7) 4.3 (0,6) 3.8 (0.8) 3.9 (0.7) 
    Self-efficacy as regards saying no to sex, mean (SD) 3.9 (0.6) 4.2 (0.7) 3.8 (0.8) 4.0 (0.7) 
Intentions to remain a virgin for next 6 months, n (%) 
    Intend to remain a virgin 52 (49.5) 534 (74.1) 84 (45.6) 265 (64.5) 
    Uncertain 18 (17.1) 107 (14.8) 28 (15.2) 60 (14.6) 
    Intend to have sexual debut 35 (33.3) 80 (11.1) 72 (39.1) 86 (21) 
Victim of sexual violence, n (%) 28 (26.2) 48 (6.7) 63 (34.1) 79 (19.3) 

SD, standard deviation.

Initial testing of gender differences of associations gave no reasons for developing separate structural equation models for males and females. All associations included in the equation model shown (see Fig. 1 and Table III) were statistically significant (P < 0.05) and all excluded associations were not significant. A number of social cognition predictors were neither significantly associated with intentions nor with transition (Fig. 1). This was the case for perceived susceptibility to sexually transmitted infections (STIs), perceived threat from STIs and subjective norms. The strongest predictor of intentions was social outcome expectancies with a standardized coefficient of 0.43. Those who expected negative social consequences of delayed transition had stronger intentions to have sexual intercourse. Knowledge was significantly, but weakly, associated with intentions (coefficient = −0.06). Intentions to have intercourse increased with age (0.09), decreased with higher socio-economic status (−0.15) and were lower for females than for males (−0.18).

Table III.

Testing of final Mplus model with the Satorra–Bentler test; included in the model are associations significant at the P < 0.05 level only

Constrained association Model with one association constrained to zero
 
Final model
 
Differences between models
 
d.f. Correlated error term Chi square d.f. Correlated error term Chi square d.f. Chi square P
Constraining associations included in the model 
    Intentions with 
        Gender 128 1.277 433.599 127 1.276 421.429 8.668 0.01 
        Age 128 1.268 433.475 127 1.276 421.429 47.802 0.001 
        Socio-economic status 128 1.253 450.264 127 1.276 421.429 −17.287 0.001 
        Knowledge 128 1.272 426.004 127 1.276 421.429 5.988 0.05 
        Social outcome expectancies 128 1.266 549.012 127 1.276 421.429 −31895.8 0.001 
    Transition with 
        Intention 128 1.266 435.287 127 1.276 421.429 −3464.5 0.001 
        Gender 128 1.283 445.586 127 1.276 421.429 11.122 0.001 
        Age 128 1.266 434.693 127 1.276 421.429 −3316.0 0.001 
        Socio-economic status 128 1.255 443.345 127 1.276 421.429 −15.521 0.001 
        Violence 128 1.279 438.185 127 1.276 421.429 10.094 0.01 
        Self-efficacy 128 1.276 426.811 127 1.276 421.429 4,218 0.05 
Adding and constraining associations not included in the model 
    Intentions with          
        Violence 127 1.276 421.429 126 1.273 420.582 0.512 NS 
        Subjective norms 139 1.272 440.691 138 1.271 441.257 −0.401 NS 
        Self-efficacy 127 1.276 421.429 126 1.258 423.228 −0.508 NS 
        Perceived threat 198 1.238 626.390 197 1.227 631.433 −1.481 NS 
        Susceptibility 177 1.237 542.293 176 1.230 545.470 −1.287 NS 
    Transition with 
        Knowledge 127 1.276 421.429 126 1.277 418.291 2.729 NS 
        Subjective norms 139 1.272 440.691 138 1.277 438.915 3.052 NS 
        Social outcome expectancies 127 1.276 421.429 126 1.280 420.075 1.754 NS 
        Threat 198 1.238 626.390 197 1.237 625.047 −0.458 NS 
        Susceptibility 177 1.237 542.293 176 1.235 542.063 0.115 NS 
Constrained association Model with one association constrained to zero
 
Final model
 
Differences between models
 
d.f. Correlated error term Chi square d.f. Correlated error term Chi square d.f. Chi square P
Constraining associations included in the model 
    Intentions with 
        Gender 128 1.277 433.599 127 1.276 421.429 8.668 0.01 
        Age 128 1.268 433.475 127 1.276 421.429 47.802 0.001 
        Socio-economic status 128 1.253 450.264 127 1.276 421.429 −17.287 0.001 
        Knowledge 128 1.272 426.004 127 1.276 421.429 5.988 0.05 
        Social outcome expectancies 128 1.266 549.012 127 1.276 421.429 −31895.8 0.001 
    Transition with 
        Intention 128 1.266 435.287 127 1.276 421.429 −3464.5 0.001 
        Gender 128 1.283 445.586 127 1.276 421.429 11.122 0.001 
        Age 128 1.266 434.693 127 1.276 421.429 −3316.0 0.001 
        Socio-economic status 128 1.255 443.345 127 1.276 421.429 −15.521 0.001 
        Violence 128 1.279 438.185 127 1.276 421.429 10.094 0.01 
        Self-efficacy 128 1.276 426.811 127 1.276 421.429 4,218 0.05 
Adding and constraining associations not included in the model 
    Intentions with          
        Violence 127 1.276 421.429 126 1.273 420.582 0.512 NS 
        Subjective norms 139 1.272 440.691 138 1.271 441.257 −0.401 NS 
        Self-efficacy 127 1.276 421.429 126 1.258 423.228 −0.508 NS 
        Perceived threat 198 1.238 626.390 197 1.227 631.433 −1.481 NS 
        Susceptibility 177 1.237 542.293 176 1.230 545.470 −1.287 NS 
    Transition with 
        Knowledge 127 1.276 421.429 126 1.277 418.291 2.729 NS 
        Subjective norms 139 1.272 440.691 138 1.277 438.915 3.052 NS 
        Social outcome expectancies 127 1.276 421.429 126 1.280 420.075 1.754 NS 
        Threat 198 1.238 626.390 197 1.237 625.047 −0.458 NS 
        Susceptibility 177 1.237 542.293 176 1.235 542.063 0.115 NS 

NS, not significant.

Fig. 1.

Transition to vaginal or anal intercourse and intentions by predictors. Structural equation model (n = 1333).

Fig. 1.

Transition to vaginal or anal intercourse and intentions by predictors. Structural equation model (n = 1333).

Transition for first sexual intercourse increased with intentions (0.10), violence (0.37) and age (0.10) and decreased with socio-economic status (−0.13) and self-efficacy with regard to negotiating delayed sex (−0.08). Transition was also lower for females than for males (−0.32). The model predicted 35% of the variance in intentions and 16% of the variance in transition. The model which was based on the assumption that all variables were continuous showed the following fit indices: CFI = 0.969, TLI = .960, RMSEA = .042 and SRMR = .036. They all indicate adequate model fit.

Discussion

This study found a high 15-month incidence of first sexual intercourse among 13 year olds: 13% among girls and 31% among boys. A substantial proportion (14% of girls and 26% of boys) at the average age of 13 years reported that they intended to have sexual intercourse in the next 6 months. We also found high levels of partner violence, with more boys than girls reporting at baseline that they had been a victim of physical violence in intimate relationships. This is consistent with other research which reported partner violence experienced by up to 40% of adolescents in a relationship, with a similar gender distribution [25, 26].

Socio-economic status was inversely associated with intention to have an early sexual debut and inversely associated with having an early sexual debut. Adolescents from poorer homes are likely to feel they have fewer opportunities in life, and they might lack the educational, career and recreational aspirations characterizing adolescents from wealthier homes [34]. Also they might not have the same exposure to interventions that aim to postpone first intercourse and might not have the same levels of parental supervision at home. This finding concurs with a study of older adolescents (15–19 years old), where those from homes with insufficient money for food and clothes were almost twice as likely to have had their sexual debut than those from homes in which some luxury goods were affordable [35].

Adolescents who expected negative social consequences related to abstinence (for example being less accepted by friends or losing friends) had stronger intentions to have sexual intercourse. Also, adolescents with better knowledge about the transmission of HIV, methods to prevent acquiring HIV, the relationship between other STIs and HIV and the difference between HIV and AIDS were less likely to intend to make the transition to first sexual intercourse.

The transition to first sexual intercourse was predicted by adolescents' intentions to have first intercourse, by low self-efficacy to negotiate delayed sexual debut and by being a victim of physical violence in intimate relationships. Self-efficacy can be improved through vicarious experience (modelling interventions) or verbal persuasion. Also, the idea that there is reciprocal determinism between individual-level factors and environmental factors [36] suggests that self-efficacy is likely to be maximized in a social environment that is supportive to the behaviour and that community-level interventions focusing on social norms around early sexual debut are important. Addressing other individual factors such as improving self-esteem and treating depression is also likely to have a positive effect on an adolescent's self-efficacy [34].

Not only were adolescents in physically violent relationships more likely to make the transition to first sex but also once they had made the transition, they were also more likely to report coerced sex. Other research has provided insights into the correlates of adolescent partner violence and suggests that peer opinion leaders might be particularly suited to influence the attitudes and subjective norms correlated with partner violence [26]. Recently, two community-based HIV prevention programmes have demonstrated reductions in intimate partner violence [37–39], and their adaption for young adolescents is advocated.

Our finding that subjective norms did not influence intentions or sexual debut is unexpected given that other research has revealed the strong and consistent influence of social norms on adolescent sexual behaviour [40] and given the common assertion that African cultures are more collectivist than individualist and that in such cultures subjective norms will play a strong role [13, 14]. The social influence aspect, however, is perhaps even more adequately captured by the social outcome expectancy scale. Without introducing this scale in the model, the social norm predictor may have obtained significance.

A strength of our study is its longitudinal design, allowing us to ascertain temporal sequences. Our use of structural equation modelling enabled the modelling of latent variables and indirect pathways of influence (as opposed to direct, independent pathways) and with testing of a complex model. Our study has several weaknesses. The non-response at baseline and the loss to follow-up might have resulted in biases. Our analyses did not include factors such as family connectedness, religiosity, self-esteem and academic aspirations, factors which have been shown to be associated with the time of sexual debut in other settings [41]. If we had measured other aspects of adolescents' social contexts, personal coping skills and resources and mental health, we might have accounted for more of the variance in transition to first sexual intercourse. We did not collect information on adolescent's sexual partners and the perpetrators of physical and sexual violence. This limits our understanding of the circumstances of the adolescent's first sexual intercourse. There is already qualitative research explaining gender power inequities resulting in young women being the victims of violence in their intimate partnerships and explaining the context of early first sex for young women [19, 22]. There is little research explaining the context of early first sex among boys. Our findings show that anal intercourse is a significant feature of early sexual intercourse especially for boys. We were not able to know whether or not it was receptive intercourse, nor were we able to speculate how a boy's experience of physical violence affects his transition to first intercourse.

Conclusion

We have shown that for young adolescents in South Africa, individual choices about sexual debut are constrained by socio-economic conditions. This implies that poverty reduction and initiatives to reduce the obstacles to economic and educational opportunities need to be central in HIV prevention programmes. We have also shown that early experiences of intimate partner violence are a risk factor for early sexual debut. Peer-led or opinion leader programmes might be particularly suited to influence attitudes and subjective norms relating to partner violence and gender inequities, as well as to influence beliefs about the social acceptability of delaying sexual debut. Our findings also indicate adolescents' need to be exposed to programmes to promote communication as an alternative to violence and to help them develop their skills and confidence to assert their choices about when they will make the transition to first sexual intercourse.

Funding

European Commission—International Cooperation Activities Research Programme (Fifth Framework Programme—ICA4-CT-2002-10038).

Conflict of interest statement

None declared.

Thanks to the participating students, the schools in which the research was undertaken and the Western Cape Education Department for making the research possible. The partners and principal investigators of the study include University of Cape Town (A.J.F.), Muhimbili University College of Health Sciences (Sylvia Kaaya), University of the North (Hans Onya), Karolinska Institute (Minou Fuglesang), University of Maastricht (H.S.), University of Oslo (Knut-Inge Klepp), World Population Foundation (Jo Reinders) and University of Bergen (L.E. A.—coordinator). This study was carried out as part of the European Union-funded research project. The full title of the project is ‘Promoting sexual and reproductive health. School-based HIV/AIDS prevention in sub-Saharan Africa'. The acronym for the project is ‘SATZ', which stands for South Africa Tanzania.

References

1.
World Health Organisation
Risk and Protective Factors Affecting Adolescent Reproductive Health in Developing Countries
2006
 
2.
Pettifor
AE
Rees
HV
Steffenson
A
, et al.  . 
HIV and Sexual Behaviour among Young South Africans: A National Survey of 15–24 Year Olds
 , 
2004
Johannesburg
Reproductive Health Research Unit, University of Witwatersrand
3.
Drain
PK
Smith
JS
Hughes
JP
, et al.  . 
Correlates of national HIV seroprevalence: an ecologic analysis of 12 countries
J Acquir Immune Defic Syndr
 , 
2004
, vol. 
35
 (pg. 
407
-
20
)
4.
Pettifor
AE
Rees
HV
Kleinschmidt
I
, et al.  . 
Young people's sexual health in South Africa: HIV prevalence and sexual behaviours from a nationally representative household survey
AIDS
 , 
2005
, vol. 
19
 (pg. 
1525
-
34
)
5.
Laga
M
Schwartlander
B
Pisani
E
, et al.  . 
To stem HIV in Africa, prevent transmission to young women
AIDS
 , 
2001
, vol. 
18
 (pg. 
931
-
4
)
6.
Ajzen
I
The theory of planned behaviour
Organ Behav Hum Decis Process
 , 
1991
, vol. 
50
 (pg. 
79
-
211
)
7.
Ajzen
I
Perceived behavioural control, self-efficacy, locus of control, and the theory of planned behaviour
J Appl Soc Psychol
 , 
2002
, vol. 
32
 (pg. 
1
-
20
)
8.
Sheeran
P
Taylor
S
Predicting intentions to use condoms: a meta-analysis and comparison of the theories of reasoned action and planned behaviour
J Appl Soc Psychol
 , 
1999
, vol. 
29
 (pg. 
1624
-
75
)
9.
Albarracin
D
Johnson
BT
Fishbein
M
, et al.  . 
Theories of reasoned action and planned behaviour as models of condom use: a meta-analysis
Psychol Bull
 , 
2001
, vol. 
127
 (pg. 
142
-
61
)
10.
Ajzen
I
Nature and operation of attitudes
Annu Rev Psychol
 , 
2001
, vol. 
52
 (pg. 
27
-
58
)
11.
Armitage
CJ
Conner
M
Efficacy of the theory of planned behaviour: a meta-analytic review
Br J Soc Psychol
 , 
2001
, vol. 
40
 (pg. 
471
-
99
)
12.
Bandura
A
Self-Efficacy: The Exercise of Control
 , 
1997
New York
Freeman
13.
Aarø
LE
Schaalma
H
Astrøm
AN
Klepp
K-I
Flisher
AJ
Kaaya
S
Predicting health behaviour: the applicability of social cognition models in an African context
Promoting Adolescent Sexual and Reproductive Health in Eastern and Southern Africa
 , 
2007
Uppsala, Sweden
Nordic Africa Institute
14.
Giles
M
Liddell
C
Bydawell
M
Condom use in African adolescents: the role of individual and group factors
AIDS Care
 , 
2005
, vol. 
17
 (pg. 
729
-
39
)
15.
Vergnani
T
Flisher
AJ
Blignaut
R
Factors affecting condom use by South African adolescents
 
14th International AIDS Conference, Barcelona, Spain, 7–12 July 2002
16.
Bosompra
K
Determinants of condom use intentions of university students in Ghana: application of the theory of reasoned action
Soc Sci Med
 , 
2001
, vol. 
52
 (pg. 
1057
-
69
)
17.
Lugoe
W
Rise
J
Predicting intended condom use among Tanzanian students using the theory of planned behaviour
J Health Psychol
 , 
1999
, vol. 
5
 (pg. 
25
-
43
)
18.
Klepp
K-I
Ndeki
SS
Thuen
F
, et al.  . 
Predictors of intentions to be sexually active among Tanzanian school children
East Afr Med J
 , 
1994
, vol. 
73
 (pg. 
218
-
24
)
19.
Wood
K
Maforah
F
Jewkes
R
‘He forced me to love him’: putting violence on the adolescent sexual health agenda
Soc Sci Med
 , 
1998
, vol. 
47
 (pg. 
233
-
42
)
20.
Jewkes
R
Vundule
C
Maforah
F
, et al.  . 
Relationship dynamics and teenage pregnancy in South Africa
Soc Sci Med
 , 
2001
, vol. 
5
 (pg. 
733
-
44
)
21.
Reddy
SP
Panday
S
Swart
D
, et al.  . 
Umthenthe Uhlaba Usamile—The South African Youth Risk Behaviour Survey 2002
 , 
2003
Cape Town
South African Medical Research Council
22.
Wood
K
Jewkes
R
“Love Is a Dangerous Thing”: Micro-Dynamics of Violence in Sexual Relationships of Young People in Umtata
  
Medical Research Council Technical Report. Pretoria, South Africa: Medical Research Council, 1998
23.
Buga
G
Amoko
D
Ncayiyana
D
Sexual behaviour, contraceptive practice and reproductive health among school adolescents in rural Transkei
S Afr Med J
 , 
1996
, vol. 
86
 (pg. 
523
-
7
)
24.
Jewkes
R
Jejeebhoy
SJ
Shah
I
Thapa
S
Non-consensual sex among South African youth: prevalence of coerced sex and discourses of control and desire
Sex Without Consent: Young People in Developing Countries
 , 
2005
London
Zed Books
25.
Swart
LA
Seedat
M
Stevens
G
, et al.  . 
Violence in adolescents' romantic relationships: findings from a survey amongst school-going youth in a South African community
J Adolesc
 , 
2002
, vol. 
25
 (pg. 
385
-
95
)
26.
Flisher
AJ
Myer
L
Marais
A
, et al.  . 
Prevalence and correlates of partner violence among South African adolescents
J Child Psychol Psychiatry
 , 
2007
, vol. 
48
 (pg. 
619
-
27
)
27.
Fishbein
M
The role of theory in HIV prevention
AIDS Care
 , 
2000
, vol. 
12
 (pg. 
273
-
8
)
28.
Aarø
LE
Flisher
AJ
Kaaya
S
, et al.  . 
Promoting sexual and reproductive health in early adolescence in South Africa and Tanzania: development of a theory- and evidence-based intervention programme
Scand J Public Health
 , 
2005
, vol. 
30
 (pg. 
148
-
60
)
29.
Flisher
AJ
Parry
CDH
Evans
J
, et al.  . 
Substance use in Cape Town, South Africa: prevalence rates and correlates
J Adolesc Health
 , 
2003
, vol. 
32
 (pg. 
58
-
65
)
30.
Mukoma
W
Mathews
C
Flisher
AJ
, et al.  . 
Use of electronic questionnaires on handheld devices to evaluate the effects of a school-based HIV prevention programme on adolescent sexual behaviour
 
XV International AIDS Conference, Bangkok, 11–16 July 2004
31.
Straus
MA
Hamby
SL
Boney-McCoy
S
, et al.  . 
The revised Conflict Tactics Scales (CTS2): development and preliminary psychometric data
J Fam Issues
 , 
1996
, vol. 
17
 (pg. 
283
-
316
)
32.
Muthén
LK
Muthén
BO
Mplus User's Guide
 , 
2004
3rd edn
Los Angeles, CA
Muthén and Muthén
33.
Satorra
A
Heijmans
RDH
Pollock
DSG
Satorra
A
Scaled and adjusted restricted tests in multi-sample analysis of moment structures
Innovations in Multivariate Statistical Analysis. A Festschrift for Heinz Neudecker
 , 
2000
London
Kluwer Academic Publishers
(pg. 
233
-
47
)
34.
Eaton
L
Flisher
AJ
Aarø
LE
Unsafe sexual behaviour in South African youth
Soc Sci Med
 , 
2003
, vol. 
56
 (pg. 
149
-
65
)
35.
Kelly
K
Parker
W
Communities of Practice. Contextual Mediators of Youth Response to HIV/AIDS. Sentinel Site Monitoring and Evaluation Project
 , 
2000
Grahamstown, South Africa
Centre for AIDS Development, Research and Evaluation
 
http://www.cadre.org.za/. Accessed: 14 October 2006
36.
Bandura
A
Social Foundations of Thought and Action: A Social Cognitive Theory
 , 
1986
Englewood Cliffs, NJ
Prentice Hall
37.
Pronyk
PM
Hargreaves
JR
Kim
JC
, et al.  . 
Effect of a structural intervention for the prevention of intimate-partner violence in rural South Africa: a cluster randomised trial
Lancet
 , 
2006
, vol. 
368
 (pg. 
1973
-
83
)
38.
Jewkes
R
Nduna
M
Levin
J
, et al.  . 
A cluster randomised controlled trial to determine the effectiveness of Stepping Stones in preventing HIV infections and promoting safer sexual behaviour amongst youth in the rural Eastern Cape, South Africa: trial design, methods and baseline findings
Trop Med Int Health
 , 
2006
, vol. 
11
 (pg. 
3
-
16
)
39.
Jewkes
R
Nduna
M
Levin
J
, et al.  . 
Evaluation of Stepping Stones. A Gender Transformative HIV Prevention Intervention
  
South African Medical Research Council Policy Brief. Pretoria, South Africa: South African Medical Research Council, 2007
40.
Kirby
D
Understanding what works and what doesn't in reducing adolescent sexual risk-taking
Fam Plann Perspect
 , 
2001
, vol. 
33
 pg. 
276
  
[viewpoint]
41.
Carvajal
SC
Parcel
GS
Banspach
SW
, et al.  . 
Psychosocial predictors of delay of first sexual intercourse by adolescents
Health Psychol
 , 
1999
, vol. 
18
 (pg. 
443
-
52
)

Comments

0 Comments