The data used in this paper comes from a cross-sectional survey on attitude and behaviour towards premarital sex among college students of Kathmandu, the capital of Nepal, carried out in 2006. The survey involved a total of 1137 students (573 male and 564 female) studying in 12 colleges affiliated to Tribhuvan University in Kathmandu. Analyses were confined to 428 students who reported that they have ever had sexual intercourse. The scientific committee, which included the Ethical Review Board of the University Grant Commission in Nepal, approved the proposal and provided funding for this study.

A two-stage random sampling technique was applied. The first stage included a random selection of 12 colleges in Kathmandu. In order to select these colleges, a list of all the private and public colleges affiliated with Tribhuvan University and located in the Kathmandu Valley (which includes three districts: Bhaktapur, Lalitpur and Kathmandu) was obtained from the office of the Vice Chancellor in Kathmandu. This list included colleges that provide intermediate (commonly known as Grades 11 and 12), undergraduate and graduate degrees. In the second stage, two classes were randomly selected from each sampled college. The number of students in a class ranged from 40 to 60. All the colleges are co-educational, and all male and female students present on the day of the interview in the sampled classes were requested to participate in the study.

Due to the sensitive nature of the study and the educational background of the respondents, a self-administrated, structured questionnaire in the Nepali language was used to obtain information. The questionnaires were first developed in English and then translated into Nepali. Almost all sections pertaining to the behavioural aspect were based on Behavioral Surveillance Survey questionnaire developed by Family Health International/Impact [⁴³]. However, necessary modifications were made to suit the sample population. The questionnaires were pre-tested among students in a college that had not been selected as part of the study, and later refined as required. The pre-test was conducted to determine whether the questionnaires were in sequential order and the wording in Nepali was understandable or not. Most of the questions were close ended; a few open-ended questions were also included.

Female and male students filled in the questionnaire separately in different classrooms. Each student was allocated a separate bench, as in an exam setting, before the questionnaire was distributed to them. A male researcher supervised the male students' class while a female researcher supervised the female students' class. Students were then requested to place the filled-in questionnaire on a table in the corner of each class.

Before starting the study in a sampled college, approval from the campus administrative authority was obtained. All the participants involved in the study were fully informed about the nature of the study, the research objectives, and the confidentiality of the data. After this, verbal consent was obtained from the participants before they were enrolled in the study. The consent form was also written in the local language, stating the study's objectives, nature of the participant's involvement, risk and benefits, and confidentiality of the data. Students were requested to read the consent form carefully. They were given clear options on voluntary participation. It was also made clear that they could refuse to answer any of the questions and terminate the interview if and when they desired.

All of the approached students agreed to participate in the study. Confidentiality of information was ensured by removing personal identifiers from the completed questionnaires. The names of sampled colleges were not made public and thus, it is not possible for anyone outside the research team to trace reported incidents of sexual behaviour to respondents. Respondents were thus protected from any possible adverse repercussions of participating in the study.

The measurable outcome of the study is condom use at first sexual intercourse, a dichotomous variable indicating whether or not the respondent had used a condom during the first sexual encounter.

The independent variables used in the study were: sex of the respondent; age; caste and/or ethnicity; level of education; age at first sexual intercourse; marital status; types of first sexual partner; permanent place of residence; alcohol consumption; mass media exposure; and living arrangement. All these variables were organized into two or three categories, based on those used in other literature, as well as on the frequency distribution of the variables.

The indicators of exposure to mass media include an exposure to radio, television and newspapers. The majority of the students rent rooms in the Kathmandu Valley so it is assumed that not all of them have TVs in their rooms. Thus, radio and TV were combined and treated as electronic media, and newspaper as print media. Almost all students were exposed to at least one type of media (either electronic or print). Because of this, the mass media variable was organized into two categories. If a student was exposed to only one type of media, it was considered as low exposure; if a student was exposed to both print and electronic media, it was considered as high exposure.

Similarly, some ethnic groups in Nepal offer alcohol to the gods in religious ceremonies, and people have to consume alcohol in these ceremonies. Therefore, this variable was categorized into two: (1) never/rarely consume alcohol and (2) frequently consume alcohol (two or three times a week).

The variable, "living arrangement", is also organized into two categories: those students who live with their family members were considered as "with biological family"; those who live away from family members were considered as "without biological family". The other independent variables were categorized in the same way.

All completed survey questionnaires were entered into a database after manual coding and validation. Data entry and validity checks were performed for all the questionnaires by using the computer software programme, dBase IV. The cleaned and validated data was transferred into the SPSS software programme for further processing and analysis.

Both bivariate and multivariate techniques were applied in the analysis. The Chi-square test was used to test the association between the variables. Those variables that were significant in the bivariate analysis were further reexamined in the multivariate analysis (binary logistic regression) in order to identify the significant predictors after controlling for other variables. Before the multivariate analysis, multicollinearity among variables was assessed, and the least important variable, which was highly correlated to other variables, was removed from the logistic model.

Among the students, about two in five (39%) (47% boys; 28% girls) had sexual intercourse irrespective of their marital status. Among these sexually active students (n = 428), around 27% were aged 15 to 19. Eleven percent reported that they had had sex before the age of 16. A large majority of the sexually active students were from outside of the Kathmandu Valley. More than a third (34%) of the students consumed alcohol frequently (two or three times a week). Almost half of the students resided with their biological families. Almost all students were exposed to at least one type of mass media (either electronic or print). Furthermore, both male and female college students were generally aware of HIV/AIDS and knew of at least one mode of transmission of HIV/AIDS (data not shown).

Among the sexually active students, just less than half (48%) had used condoms during their first sexual contacts. Table 1 shows the clear association between condom use at first sexual intercourse and different socio-demographic characteristics. Of those who used condoms at their first sex, a significantly higher proportion than their comparison group: were males; were aged 15 to 19; were from the Brahmin and Chhetri communities; had first sex at age 16 or older; were unmarried; had sex with a boyfriend/girlfriend; were from outside the Kathmandu Valley; had high exposure to the mass media; and lived with their biological families (Table 1).

Binary logistic regression analysis was used to measure the strength of the association between various independent variables and the probabilities of using condoms at first sex. Only those variables that had significant association in bivariate analysis were reassessed in the logistic model. Before the multivariate analysis, multicollinearity among the variables was assessed. It was found that the variables, "marital status" and "types of first sex partner", were highly correlated (r = 0.7). Therefore, the variable, "marital status", was not included in the logistic model.

Analysis from logistic regression showed that age group, caste and/or ethnicity, age at first sexual intercourse, types of first sex partner, alcohol consumption, and mass media exposure were significant predictors for condom use at first sexual intercourse. Students aged 20 or older were less likely to use condoms (OR = 0.40) than students aged 15 to 19. Similarly, students from castes other than Brahmin and Chhetri were 40% (OR = 0.60) less likely to use condoms during their first sexual intercourses than those from other castes and ethnic groups. Furthermore, students who had sexual intercourse for the first time in the age groups of 16 to 19 and 20 or older were almost four times (OR = 3.5) and nine times (OR = 8.9) more likely, respectively, than students who had first sex before 16 years to use condoms at first sexual intercourse (Table 2).

Those students who had first sex with a commercial sex worker were about five times more likely than those who had first sex with their spouse to use condoms during first sexual intercourse. An inverse relationship was observed between alcohol consumption and condom use. Those students who had frequently consumed alcohol were 54% (OR = 0.46) less likely to use condoms at first sexual intercourse than those who did not consume alcohol. On the other hand, students who had higher exposure to both print and electronic media were about twice (OR = 1.75) as likely as who had lower media exposure to use condoms during their first sexual intercourses (Table 2).

Note: *** = p < 0.001 ** = p < 0.01 * = p < 0.5

Note: *** = p < 0.001 ** = p < 0.01 * = p < 0.5