Our cluster randomised trial began in the autumn of 2001. Forty public high schools—about 75% of such schools in the state of Morelos—were randomised to three arms. We estimated sample size according to four factors: expected intra-cluster (school) correlation, expected difference in the outcome variables at follow-up, average of expected observations by school, and feasible number of schools per arm (total). We used an intra-cluster correlation factor of 0.001, based on previous studies. As all outcome variables were proportions, we estimated the sample size by using the difference from the most restrictive proportion (50%), assuming that we wanted to be able to detect an improvement of 10% (five percentage points). From the available data, we knew that the average cluster (school) size was 220 first year students. We estimated the minimum number of schools at 12 per arm. We selected schools and asked them to participate, on the basis of stratified random sampling (stratified by degree of urbanisation), with sampling proportional to school size. All invited schools participated. We used data from the Mexican National Population Council to assign the category of “marginalisation” (see box) and degree of urbanisation for the communities where the schools are located. None of the schools had offered such specific and detailed education about emergency contraception or HIV prevention before. All participating 10th grade students were asked to respond to the questionnaires (in Mexico, high school comprises the 10th to 12th grades, ages 15 to 18). The overall response rate for each round of data collection was more than 95% of students attending school on the day the questionnaire was administered.
Ten of the 40 schools were randomised as control schools and continued with the biology based sex education course implemented by the Ministry of Education. We randomised 15 schools to receive the HIV education course with condom promotion and 15 schools to the same course plus a module on emergency contraception and improved access to such contraception. Two of these 30 schools initially randomised as intervention schools did not teach the intervention course, even though their teachers had been trained successfully. The primary analysis was an intention to treat analysis. We also report an actually treated analysis, in which these two schools that did not receive the intervention were included as control schools (see tables on bmj.com
Marginalisation is the term used in Mexico for the government's multidimensional assessment of poverty in a community
- Housing (per cent of households without piped water, without sewage, without electricity, with a dirt floor, and with more than two people per room)
- Income (per cent ≤ 2× minimum wage)
- Education (per cent aged > 15 who are illiterate, per cent aged > 15 who did not complete primary school)
- Urbanisation (per cent who live in towns with more than 5000 occupants)
Participating schools chose 106 teachers to take part in a week long (40 hour) training session between November 2001 and January 2002. Teachers from the 15 schools randomised to teach emergency contraception had an extra two hour training module. Training covered the content and goals of each class. Teachers participated directly in the activities they would teach. Questionnaires given before and after training showed improved basic knowledge of HIV and AIDS, attitudes towards people with HIV, perception of ability to teach the material, and confidence in dealing with sensitive subjects in class.31
The curriculum was based on teaching life skills and followed the guidelines of the UN programme on HIV/AIDS for effective school based programmes.3,4
Almost half of the time in class focused on the consequences of unprotected sex and how to avoid it. Other classes dealt with the social pressures that influence sexual behaviour (peer pressure, cultural values) and provided practice in communication, negotiation, and refusal skills.
In February 2002, the teachers began to teach the 15 week, 30 hour course (16 weeks, 32 hours for the promotion with contraception arm). Students completed a 93 item anonymous questionnaire during class on three occasions: baseline in February 2002, immediately after intervention in June 2002, and one year later in June 2003. The questionnaires covered knowledge and attitudes about HIV, AIDS, and emergency contraception; sexual experience; and the use of condoms at first and most recent intercourse. We also asked about tobacco, alcohol, and drug use, compensated sex (exchange of sex for money, goods, or favours), social networks, socioeconomic status, and intention to continue in school. Data were entered twice to minimise errors in data entry.
The research team monitored the progress of the intervention in each school throughout the programme. Monitoring consisted of 424 telephone calls to teachers, 212 visits to schools to speak with teachers and head teachers, and 25 direct class observations.
We obtained informed signed consent from each student before each questionnaire was completed. All questionnaires were anonymous.
All analyses took the cluster sample design into account (fixed effects were used for regression and standard errors were adjusted for number of primary sampling units for the descriptive statistics). We used Stata 8.0 for all analyses.
We used nine questions on biology, transmission, and prevention to assess knowledge of HIV. Responses were graded on a five point scale from certain that the statement is true to certain that it is false. We summed the points for each question (0 for completely incorrect, 4 for completely correct) to produce a single score for each student. The resulting score varied from 0 to 36.
We considered students to be knowledgeable about emergency contraception if they identified such contraception as pills, taken orally, that prevent pregnancy if taken after sexual intercourse. To investigate attitudes about condoms, we asked about the acceptability of condoms; we assessed whether the student would use a condom, and whether they would prefer sex with a condom at their next sexual intercourse. We also asked whether they would interrupt sex to ask their partner to put on a condom, and whether they would tell their partner that they would only have sex with a condom (sex was conditional on condom use).
Because the questionnaire was self administered and anonymous, we could not ensure that all questions were answered or that answers were consistent. Thus, sample size is not homogeneous across different variables or analyses.
For the behavioural variables included in the baseline questionnaire, we estimated the difference in differences at school level by using fixed effects logistic regression models to correct for intra-school correlation and to take potential trends into account.
We included a dichotomous variable for each intervention and another to distinguish the baseline survey from the follow-up survey. Thus, the follow-up dummy variable intends to capture the time trend, and the intervention variables intend to capture the impact of each intervention on the outcome variable.
We used age at baseline instead of age at the time of the survey to avoid confusion with the time trend. For the follow-up survey, we used the age the respondent would have been at baseline to make the variables comparable.