The communication scales (subscales as well as total scale) had high internal consistency (alpha coefficients .83 and higher), thus indicating that we were able to measure communication about HIV/AIDS, abstinence and condoms constructs across three communication partners and on three occasions with an acceptable degree of accuracy. The correlations among communication sumscores across measurement occasions are sufficiently high to demonstrate some stability in such communication. The patterns are also consistent with the assumption that longer time spans produce lower autocorrelations. These findings give reasons for being confident about the quality of the communication scale.
In line with findings from some previous studies [10–14, 19, 20, 22], the present study shows significant cross-sectional associations between communication variables and condom use. The associations are quite similar across communication partners (parents, other adult family members, and teachers) and topics discussed (HIV/AIDS, abstinence and condom – tables not shown). Communication sumscore means are generally highest among those who use condoms consistently, lower among the occasional condom users, and lowest among the never-users. To the best of our knowledge, this is the first time such a broad and consistent pattern of cross-sectional associations between communication variables and condom use has been demonstrated in a study from sub-Saharan Africa.
We have only identified one previous study which examines the association between communication on sexuality issues and condom use prospectively . They found some prospective association between degree of comfort in talking with parents about sex and condom use. In our study similar associations proved significant in three different statistical analyses with control for a number of other predictors. This is consistent with the idea that communication between adolescents and responsible adults leads to safer sexual practices. However, there are other possible explanations. There could be third variables, not measured in our study, which accounted for the association. Prospective longitudinal designs cannot completely rule out such alternative explanations. However, since we controlled for condom use at the previous data collection occasion as well as a number of possible confounders in each model, we have at least moved one step towards confirming causality. An important next step in this line of research would be to demonstrate that experimentally-induced increases in communication would be followed by increased condom use.
We have shown that communication predicts condom use prospectively over six and even twelve months. Odds ratio values ranging from 1.20 to 1.22 amount to odds ratio values ranging from 4.29 to 4.91 if we compare the highest with the lowest value on the communication scale and if we compare consistent use with never-use. If we had measured the extent of high quality communication on sexuality issues, the prediction would certainly have been stronger. High quality communication in this context could include openness, dialogue rather than monologue, contextualisation and specificity of the messages, timing of communication, and a context characterized by authoritative rather than authoritarian parenting [32–34].
Our review of previous studies, as well as the one by Markham and associates  did not provide much consistent evidence for a positive association between adolescent-adult communication and condom use. Still, the findings from the present study are highly consistent. The more adolescents communicate on sexuality issues with parents, other adult family members and teachers, the higher are the odds of using condoms and using condoms consistently. One reason why we can draw such a consistent picture is that the present study is based on a high number of observations. Our estimates are therefore fairly accurate. Furthermore we may hypothesize that the communication skills may be higher and the context of communication taking place may be more favourable in families where there is at least some communication on sexuality issues. Our simple and straightforward measures of communication may reflect not only amount of communication, but also other aspects such as quality and context.
Gender differences in sexuality communication have been focussed in a number of studies among adolescents in sub-Saharan Africa [9, 31, 35]. No study has, however, examined gender differences in the prospective association between sexuality communication and condom use. If male dominance in sexual relationships was pronounced, one might hypothesize that the association was stronger for males than for females. In the present study the interaction with gender was, however not close to being significant. Perhaps adolescent girls in the sub-Saharan African contexts covered by this study after all do have some influence on young couple’s decisions regarding use of condoms.
It is well established that young males more than females tend to take risk, and this has also been shown to be the case for condom use . Consistent with this, we found condom use and consistent condom use to be more common among females than among males. In some South African studies condom use has been shown to be more prevalent among males than among females . Factors that may contribute to explaining these apparent inconsistencies deserve to be examined in future studies.
Previous studies from South Africa have shown that condom use is less common among disadvantaged groups [38, 39]. This pattern is confirmed in the present study. High scores on our socioeconomic status index are associated with condom use. Interventions specifically targeting disadvantaged groups should be developed and tested. Action to reduce socioeconomic inequalities and improve the situation of disadvantaged groups may, however, prove to be the most effective remedy.
Strengths and limitations of the present study
Local personnel and experts, familiar with the local languages and cultures, were involved in the development of instruments for the present study. Most questions and scales used for data collections were carefully piloted and tested. The internal consistency of scales for the measurement of interpersonal communication is high. The number of observations in this study is also quite high. Cluster effects have been systematically adjusted for in all statistical analyses of data. The findings are highly consistent. These are obvious strengths of this study. However, one clear limitation is in the measurement of communication between significant adults and adolescents on sexuality related issues. Besides the frequency of communication, which was measured in this study, additional information on the characteristics of the communication (dialogue, monologue, structured or structured, perceived quality of the communication, and timing among others) could have improved our prediction of condom use.
Another possible limitation of this study is also related to the way communication with adults is measured. The very first item was phrased like this: ‘How often do your parents or guardians talk with you about HIV/AIDS?’ The phrase ‘talk with you’ is used consistently for all nine items in the communication scale, and could imply that we refer to communication initiated by adults. In the light of Kerr & Stattin’s view that spontaneous disclosure of information is more important than parental monitoring in predicting adolescents’ adjustment , this way of phrasing the items may not be the best one. In future studies we recommend that other ways of wording such items are tested, for instance: ‘How often do you and your parents or guardians talk about HIV/AIDS?’ This is more neutral and does not imply that the parents initiate the communication. There could also be separate questions specifically related to the issue of who initiates such communication.
In spite of the rather simple and straightforward way communication is measured in this study, we have been able to show that communication predicts condom use prospectively. It is likely that high quality communication and communication containing specific messages and advice from adults would show stronger prediction of condom use than what has been shown in the present study. Separate measurement of communication with father and mother (or male and female caregivers) would also have added to the richness of the data. Communication on sexuality with fathers and mothers may function differently, especially for boys vs. girls. In future studies communication with father and mother should be measured separately.