Our findings show that Khat and alcohol use, sex, age, and school-status were independently associated with both self-report of unprotected sexual intercourse and initiation of sexual activity. The results are unlikely to be biased. Selection of the study population used a probabilistic sampling method and was not based on any of the factors under study. Thus, selection bias is unlikely. Although information bias due to a differential and systematic under-reporting of sexual behaviour among the various variables is a possibility, it is more likely that this under-reporting is randomly distributed than otherwise. If this is the case, the resulting random misclassification will tend to bias the odds ratio towards the null value . The implication for our study is that the odds ratios reported are actually underestimates of the true effect sizes in the population. We have adjusted for potential confounding by using a multivariate logistic model.
Use of alcohol, Khat and other substances were measured over the four weeks prior to the interview, while that of risky sexual behaviour over the 12 months prior to the interview. It is assumed that the four-week window assessment period is representative of long-term individual pattern of use of these substances. The correctness of this assumption can not be verified from the present study.
In this study, the association between male sex and unprotected sex, although significant, was small when other factors such as age, school status, and substance use behavior were adjusted for in a logistic model. Although hormonal factors might be expected to increase impulsiveness and risk taking behavior in males, the association with male sex may also be mediated through other intermediate behaviors such as alcohol use. In male attendees at sexually transmitted disease clinics in southern Vietnam, being aged under 20, not married, not having a current girlfriend, using alcohol before sex and substance use were all factors independently associated with visiting a female sex worker (FSW) . Similarly in Colombia, a longitudinal survey revealed that adolescents with increased drug use were more likely to engage in unprotected sex as well as multiple partnerships.
It is more difficult to speculate on the reasons for the fairly strong two-fold association between increasing age and unprotected sex. This finding goes against previous studies and our expectation. Although we have adjusted for the possible effects of alcohol and khat use, we can not exclude the effect of other potential confounding factors.
We have also shown that being out of school was strongly associated with unprotected sex. A similar finding was reported in a previously conducted study among Addis Ababa youths, where being out of school showed a strong association with sexual risk behaviour. In New South Wales, 16-year old out-of-school adolescents had consistently higher rate of reported substance use compared to their age matches in school . Although underlying behavioural problems or mental disorders could be linked to the reasons for youths being out of school, and although we have previously shown that mental and behavioural disorders are prevalent in adolescents in Ethiopia[28, 29], we can not confirm the association from the present study. In Bonomo et al.'s  study of Australian 16–17-year-olds, alcohol-related sexual risk-taking, psychiatric morbidity and high frequency of alcohol consumption had strong independent associations. In a ten-year prospective study, others reported that adolescents who have experienced drinking alcohol even once or twice during the past 12 months were more likely to exhibit more substance use, face academic problems and become involved in delinquent behaviors during the latter stages of school (middle and high school) compared to non-drinkers . Early drinking was also found to be associated with early sexual initiation .
Khat is primarily used for its stimulant effect. Users report that Khat intake results in increased energy levels and alertness, improves self-esteem, creates a sensation of elation, enhances imaginative ability and the capacity to associate ideas, and improves the ability to communicate. It has not yet been associated with alteration of rational decision-making and has not been shown to increase risk-taking behaviour. Although some users also take alcohol after Khat to counteract the stimulant properties and facilitate sleep, this should not confound the observed association as alcohol use was adjusted for in the logistic model.
Our finding of a linear and strong association between unprotected sex and alcohol is to be expected because of the nature of alcohol in decreasing inhibitions, altering rational decision making, and increasing risk-taking behaviour. A qualitative study among military conscripts in Northern Thailand drew the following conclusions regarding alcohol consumption and inconsistent condom use: Alcohol is (1) consciously used by men to reduce inhibitions that constrain their interpersonal interaction with women and with each other; (2) reduces inhibitions of individuals to sexual risk taking; (3) provides a socially acceptable excuse for non-use of condoms; (4) is associated by conscripts with brothel attendance; and (5) is seen to enhance male sexual pleasure, in contrast to condoms, which are said to reduce pleasure. In a meta-analysis of the association between alcohol intake and condom use, drinking at first intercourse was associated with decreased condom use, but alcohol was not related to condom use in recent sexual encounters and in recent encounters with new partners. On the other hand drinking was related to non condom use among adolescents .
Initiation of sexual activity was not associated with being male when other factors were adjusted in the logistic model. Others have reported that males become sexually active at a younger age than females . The reason why we did not detect any difference between males and females in initiation of sex is probably due to the influence of factors other than individual behavior. For example, although in most cultures females are expected not to initiate sex before they are married, they may be forced to have sex as has been shown by recent studies of sexual abuse in Ethiopia [35, 36]. Sex initiation was also associated with being out of school. As described above for unprotected sex, the reason for the association could be underlying behavioural problems that predispose youth to leave school and also engage in risky sexual behaviour. Thus, being out of school is likely to be just a marker of such underlying behavioural and mental problems.
Khat use was also strongly associated with sex initiation. However, Khat use was assessed over the four weeks prior to the interview, while initiation of sex was assessed for a longer period spanning from puberty to 19 years of age. It is thus possible that sex was initiated well before the initiation of Khat use or vice versa. As we do not have the pattern of use of Khat beyond the four-week period prior to the survey, it is difficult to speculate on the significance of this strong association. As is the case with Khat use, the strong and linear association of alcohol use with initiation of sex, and the association of use of substances other than Khat with sex initiation, are difficult to interpret because of the uncertain temporal relationships between the various variables.