The target population was never-married men and women aged 15–24 years living on the island of Mauritius. This population consisted of approximately 173,200 people, which was about 15% of the total population of Mauritius in 2000 13.

A two-stage cluster sampling method was used to select respondents. The enumeration areas (EAs) of the Central Statistics Office were used as the sampling frame for the first stage. Each EA consisted of approximately 300–600 people comprising 75–150 households, and those in the capital city of Port Louis and Plaines Wilhems District were defined as being from urban areas (43% of the total population). In the first stage, 50 EAs from a total of 3472 were selected using the probability proportional to size (PPS) method 1415. All never-married persons aged 15–24 years in selected areas were listed during the preparatory fieldwork for a total of 2592 persons. In the second stage, 12 males and 12 females from each EA were selected using a computer-generated random numbers algorithm. Using this multistage procedure, 600 males and 600 females were chosen to participate in the survey. The sample size was chosen to allow us to detect a 10% difference in sexual experience rate between in-school and out-of-school students with sufficient statistical significance (α = 0.05, two-tailed) and a power of 0.8, as well as to adjust for the confounding effect of multiple factors.

Of the young people selected, 59 respondents (4.9%) did not participate. The primary reasons for nonparticipation were "moved away," "gone abroad," "refused (by himself/herself or parents)," and "married." Consequently, 1141 young people (575 men and 566 women) completed the survey interviews, yielding a response rate of 95.1%. Comparisons of participants and nonparticipants showed no marked difference in sociodemographic characteristics with respect to gender, religion, or age.

The research questionnaire was developed based on the Family Health International (FHI) HIV/AIDS/STD Behavioral Surveillance Surveys (BSS) Questionnaire for youth 15. This instrument includes questions that can assess internationally agreed-upon indicators for young people's behaviors in relation to HIV/AIDS 16.

A multiphase process was used to develop the research instrument to ensure that it was developmentally, culturally, and linguistically appropriate. First, seven focus group discussions were conducted in youth centers to tailor the instrument culturally and linguistically for Mauritian youth. Second, persons working in nongovernmental organizations (NGOs) that deal with HIV/AIDS and drug use were invited to review the instrument and make recommendations for its modification. Third, the English instrument was translated into Creole and back-translated into English to confirm the accuracy of the Creole translation. Finally, the instrument was pilot-tested with 50 young people during preparatory fieldwork. Fieldworkers reported the items for which participants had difficulty in understanding the phrase or terms of the question after the pretest interview. The questionnaire was modified based on the fieldworkers' comments and finalized in January 2003.

The final version of the instrument included the following domains of interest: sociodemographic characteristics (19 items), social life (15 items), knowledge and attitude about condoms (10 items), sexual experience (22 items), and knowledge of HIV/AIDS and exposure to an HIV/AIDS program (30 items). A complete questionnaire may be obtained from the first author.

The sociodemographic variables included sex, age, religion, educational background, and work experience. Religion represents racial/ethnic background of the participants. Usually, Hindus represent people whose ancestors came from India, Muslims, from India and the Middle East, and Christians, from Africa, Europe, and China. Concerning social life, questions were asked regarding respondents' mobile phone use, nightclub visits, pornographic film viewing, Internet use, and alcohol and other drug use. Items concerning knowledge and attitude about condoms included questions related to obtaining and the ability to use condoms.

In the section on sexual behavior, young people were first asked about lifetime sexual experiences. Sexual intercourse was defined as vaginal or anal penetration. Those who reported a history of sexual intercourse responded to further questions, including the age of first sexual intercourse, condom use at the first sexual encounter, age of the partner, number of lifetime sexual partners, and sexual activity in the previous 12 months. Those who were sexually active in the previous 12 months were also asked questions regarding the use of condoms at the last sexual encounter and the frequency of condom use with both commercial and noncommercial partners. A commercial partner was defined as a partner with whom a respondent had sex in exchange for money; any partner other than this was defined as a noncommercial partner.

The section on HIV/AIDS knowledge contained 11 items, including modes of HIV transmission and measures for preventing sexual transmission. Items on the exposure to HIV/AIDS prevention programs had questions on exposure to posters on HIV/AIDS and knowledge of government programs and NGOs.

To examine the instrument's reliability, 45 items extracted from the original survey were retested 1 week later for 48 randomly selected respondents. Overall, a high test-retest correlation was observed between items (Pearson r = 0.85–1.00 [median = 0.96] for continuous variables, kappa = 0.26–1.00 [median = 0.71] for categorical variables) 17. The internal consistency was satisfactory (α = 0.68) for 11 items of knowledge. In addition, a random response technique was administered to all subjects to assess the validity of the interview-based response on sexual experience in a subsequent procedure 181920; at the end of interview, respondents were asked to answer one of two "yes/no" questions, (1) "Is your birthday on the 1st to 10th of the month?" or (2) "Have you ever had sexual intercourse?" by randomly picking up a piece of paper that contained one of these questions. The interviewer did not know which question they answered. Under this less intimidating circumstance, 28.4% male and 7.4% female were estimated to have answered "yes" to question 2. These prevalence rates were comparable to those obtained in the face-to-face interviews (30.9% for men and 9.7% for women).

A team of 50 trained fieldworkers under the direct supervision of five superiors conducted the fieldwork. Data were collected over a 3-week period in January and February 2003. Fieldworkers visited the homes of the respondents and interviewed them in their native language, Creole, using the structured questionnaire. Due to the sensitive nature of some questions, the gender of the interviewers was matched to that of the respondents. The interviews took 20–30 min to complete on average. The decision to use face-to-face interviews was made by taking into account the local language environment of Mauritius where the first language (Creole) is a highly conversation-based language, and French or English is not understood by people with a low educational attainment.

Statistical analyses were performed using SPSS Complex Samples 12.0J (SPSS Japan, Tokyo, Japan) and SUDAAN 8.0.2 (Research Triangle Institute, Research Triangle Park, NC, USA) to account for the clustered sampling design. First, we assessed the prevalence of HIV-related sexual behaviors. Subsequently, using "being sexually experienced" and "condom nonuse at last sex" as the outcome variables, bivariate associations with dichotomous correlates were analyzed using cross-tabulation. Finally, variables that achieved a statistically significant association (P < 0.10) in the bivariate analyses and variables that were epidemiologically important were entered into logistic regression models to obtain an adjusted odds ratio for each association.

The study protocol was reviewed and approved by the Kyoto University Graduate School and Faculty of Medicine Ethics Committee and Mauritius Ministry of Health and Quality of Life. Before each interview, the interviewer explained the objectives and methodology of the study to each participant; the interviewer signed the survey document when verbal informed consent was obtained from the participant.

The mean age for men was 19.2 years (SE = 0.14) and that for women was 18.6 years (SE = 0.11). The median ages were 19 years for men and 18 years for women, respectively. Religious affiliations comprised approximately 55% Hindu, 20% Muslim, and 25% Christian. Nearly half of the participants were still attending school and lived in urban areas at the time of the survey. For men, 45.8% were employed full time, whereas 57.4% of women had not worked in the previous 12 months (Table 1).

The proportion of people who had ever had penetrative sexual intercourse differed significantly between men (30.9%; 95% CI, 25.8–36.5) and women (9.7%; 95% CI, 6.5–14.0; P = 0.001). The mean age of first sex among those sexually experienced was 17.4 years for males (median = 17) and 17.8 years for females (median = 18). The mean age of the first partner was 21.4 years for females (median = 21), indicating that women tend to have an older first partner. Significantly more men (65.0%) than women (32.0%) reported having had more than one sexual partner (P = 0.001). Condom use during sexual intercourse was low for both men and women, and only 35.7% of men and 24.5% of women reported having used a condom consistently with their noncommercial sex partner during the 12 months before completing the survey (Table 2). Only 23 male respondents reported sex with a commercial partner.

The correlates of sexual experience in terms of sociodemographic characteristics and social life indicators such as exposure to nightclubs and drug use were analyzed by gender (Table 3). Logistic regression results indicate that for both sexes, non-Christians (i.e., Hindus and Muslims) had a significantly lower adjusted odds ratio (AOR) of being sexually experienced (AOR = 0.27 for men and 0.15 for women). Men who had ever worked, ever visited nightclubs, or had a history of using marijuana had a significantly higher likelihood of being sexually experienced (AOR = 3.14, 3.07, and 3.09, respectively). Although not statistically significant in the logistic regressions, most social life factors among men, i.e., having a mobile phone, a history of watching pornographic films, and drinking alcohol, were associated with being sexually experienced in bivariate analyses. Women who were not in school, had ever visited nightclubs, or reported a history of alcohol use in the 12 months before the survey had a significantly higher likelihood of being sexually experienced in logistic regression analyses (AOR = 4.28, 3.14, and 2.81, respectively).

Of those who had sex with noncommercial partners in the last 12 months (n = 123 for men and n = 40 for women), more than 50% of men and 70% of women did not use a condom at the last sexual encounter. The correlates of condom nonuse at the last sexual encounter were analyzed in terms of sociodemographic characteristics, condom attitudes, sexual behaviors, HIV/AIDS knowledge, and exposure to an HIV/AIDS program (Table 4). Male and female data were combined for condom nonuse analyses because prior bivariate analyses stratified by gender indicated that the direction and magnitude of all independent variables were similar for men and women. In the logistic regression analyses, condom nonuse at the first sexual encounter was found to be the most significant correlate of nonuse at the last sexual encounter (AOR = 12.92). Moreover, those who had not been exposed to HIV prevention programs by NGOs that promoted HIV prevention among young people had a significantly higher AOR (3.27). Those who were female and older had higher likelihoods (AOR = 3.09 and 2.68, respectively) of not using a condom at the last sexual encounter, although these factors were not statistically significant in logistic regression analyses.