Uganda was severely affected by the spread of HIV/AIDS early in the pandemic. Through political commitment from the highest level, a comprehensive multisectoral HIV/AIDS control program was put in place.1 Since the early 1990s, there is evidence (from the Ministry of Health sentinel surveillance system2 and other sources3,4) that HIV prevalence rates in pregnant women aged 15 to 24 years have been significantly declining. In addition, some large-scale cohort studies (eg, the Medical Research Council (UK) supported population-based cohort study in Masaka district and the Rakai Project) have documented declining HIV prevalence rates in the general population.5,6 In addition, a decline in HIV incidence has been documented in the Masaka cohort.7,8 Although there is currently an intensive debate going on as to whether increased sexual abstinence or increased condom use is mostly responsible for these declines in the younger population, there is definite evidence of sexual behavior changes5,9,10 and that these changes are linked to the reduction of HIV incidence.11
Evidence on HIV prevalence has often been derived from cross-sectional studies with a maximum of 1 follow-up. The United Nations Program on HIV/AIDS (UNAIDS) recommends that in addition to regular monitoring of HIV prevalence rates in sentinel groups, there should be an assessment of HIV-related risk behaviors in suitable population groups on a regular basis.12 Here, we report on the results over 6 years from a sentinel surveillance system for sexual behavior among secondary school students in an area of western Uganda for which good data exist on declines in HIV prevalence rates among young pregnant women4 but for which information on sexual behavior was largely missing.
The objectives of the study were as follows:
1. To establish time trends in sexual behavior of secondary school students from western Uganda
2. To examine time trends in variables of sexual behavior related to the Ugandan “ABC” approach (abstinence, number of sexual partners, sexual activity in past 3 months, and condom use) over a period of 6 years
3. To investigate and explain gender-specific differences in these trends over time
The study was carried out in the 3 districts of Kabarole, Kamwenge, and Kyenjojo (formerly Kabarole district) with a total population of 1034,855 in 2002. Since 1991, a comprehensive AIDS control program has been implemented through the District Health Services in the context of a Ugandan-German Technical Cooperation (Deutsche Gesellschaft fuer Technische Zusammenarbeit [GTZ]).13 District Health Management Teams (DHMTs) have the mandate for implementing comprehensive HIV/AIDS prevention and care services in their jurisdictions in a decentralized health care system as it exists in Uganda. The district-based AIDS control activities are part of the Ugandan National HIV/AIDS Control Program and follow the national guidelines of the Ugandan Ministry of Health. Interventions at the time of the study included various approaches to health education and behavioral change communication, syndromic treatment of sexually transmitted diseases (STDs), social marketing of condoms, and home-based care of patients with AIDS.14,15 In-school and out-of-school youth have been a focus of the program since the beginning in 1991. A specific school outreach program for reproductive health started in 1997, in which all secondary schools in the 3 districts were contacted at least once per year. During this period, more than 27,000 students and 1250 teachers received health education sessions lasting approximately 1 hour. Also, more than 220 individual counseling sessions were held with students' groups and teachers. In addition to the school-based education, students had various other opportunities outside of school to learn more about HIV/AIDS, such as participating in general community meetings, attending drama performances on HIV/AIDS issues, and reading printed leaflets that were developed and distributed by the District AIDS Program.
The surveillance of sexual behavior of school students was part of a national and district government program to control HIV/AIDS in the 3 selected districts. Implementation of this program was done by DHMTs in the 3 districts from 1995 through 2001. Specific responsibilities were delegated to the District Health Educator (DHE) and his 2 assistants, who visited the schools. The research component for this program consisted of input into program design, validation of the results from the self-administered questionnaires, analysis of data, and interpretation and dissemination of study results.
This study consisted of repeated cross-sectional surveys (time series study) without a control group.
All 35 secondary schools in the study area were stratified into urban and rural schools. The urban stratum comprised the town of Fort Portal (approximately 45,000 inhabitants in 2001) and schools at larger settlements along main roads, whereas the rural stratum consisted of schools that were situated along minor roads. From these lists, 7 urban and 8 rural schools were randomly selected to serve as sample schools throughout the study period. The number of selected schools was based on the assumption that, on average, 80 students would be attending senior (S2-S4) classes per school. This would provide a sample of at least 500 students in each stratum per year, which, in turn, would give 95% confidence intervals no greater than ±4.5% points around a response of 50%. Taking into account the cluster sampling method, the 95% confidence intervals around a response of 50% would be ±5.0% points.
A self-administered questionnaire consisting of 7 questions relating to knowledge of HIV preventive behaviors and practice of these behaviors, some of which had subquestions, was used. The questionnaire was in English, which is the language of instruction in all secondary schools in Uganda. The questionnaire was pretested in 2 schools before the first survey. Comprehension of the questions was good, and only minor changes were made. The key questions in the questionnaire remained the same throughout the study. The cooperation of headmasters was sought for the survey by a joint letter from the District Health and Education Departments. The surveys were carried out annually in March through April between 1995 and 2001, with the exception of 1999, when no survey could be organized because of logistic constraints. Students were not informed before the exact date of the surveys. Two experienced field assistants explained the questionnaire to the students present on the day of the survey, emphasizing that any answers would be strictly voluntary and anonymous. Teachers were not present in the class when students filled out the questionnaire.
Validation of Results From Self-Applied Questionnaires
The validation of the data from the self-administered questionnaire was done in 3 ways.
A reliability test was conducted with a random sample of 272 respondents from 3 urban schools by repeating the self-administered questionnaire after 7 days and comparing the responses for each key question for each student (test-retest substudy). The test-retest results were analyzed using the Kappa test for percent agreement between first and second responses for categoric data and calculation of the proportion of matched versus discrepant responses for continuous data. The mean Kappa value for all questions was 0.69 (range: 0.53-0.86), indicating good overall agreement between the responses.
Agreements (Kappa values) for each of 5 main key questions of the questionnaire were as follows:
1. Did you ever had sexual intercourse: 0.86
2. How old were you when you had sexual intercourse for the first time: 0.74
3. How many sexual partners did you have since becoming sexually active: 0.75
4. Did you have sexual intercourse in the past 3 months: 0.62
5. Did you ever use a condom: 0.73
Responses of students from a particular class were compared over time. For example, collective responses of S2 classes in each school in 1995 were compared with collective responses of the same classes who were filling out the questionnaire as S3 students in 1996. We would expect that the responses to the question “When did you have sex for the first time” would be the same for the S2 students in 1995 compared with the (same) S3 students in 1996. Other variables were examined in a similar way (Table 1). The responses of the same students in different years were quite consistent, especially for age at first sexual contact. Slight inconsistencies in the collective responses by class could have been caused by changes to the classes (eg, students not attending school on the day of the survey, students dropping out of school, students joining a school as new students within a particular period under observation).
A qualitative substudy using focus group discussions (FGDs) was conducted in 1998, and the findings from the FGDs were compared with those from the self-administered questionnaires. For this purpose, 2 sets of FGDs were conducted: the first set involved 12 focus groups with either 6 boys or 6 girls, respectively, from the groups that had done the test-retest experiment. The second set of 16 FGDs were carried out with 6 participants each (either boys or girls) at 5 randomly selected schools who did not participate in the test-retest study. Detailed results from this substudy are published elsewhere.16
Data were entered with EpiInfo 6.4 (World Health Organization [WHO]/Centers for Disease Control and Prevention [CDC]), and open-ended questions were coded using predefined codes. After data cleaning using logical and consistency checks, the data were transferred to STATA 6.0 (Stata Corporation, College Station, TX) for statistical analysis. The variables “ever had sex” and “age at first sex” were analyzed using survival analysis. The input data for survival analyses were age of the participant, whether or not he or she had ever had sex, and, if applicable, recalled age at first sex. Reported age at first sex was the “failure” event, and those participants who never reported having had sex were censored at their current age at the date of the survey. This procedure was recommended by Zaba et al.17 Cox hazard ratios (HRs) were reported with 95% confidence intervals for female and male students separately. The variables “ever use of condoms,” “number of sexual partners,” and “sexual intercourse in the past 3 months” were analyzed using logistic regression, with time as the important covariate of interest. We compared data from 3 periods: 1995 and 1996, 1997 and 1998, and 2000 and 2001. We fitted the models with all possible combinations of variables. For example, we included 1 variable, 2 variables, and all 3 variables in the models. From these combinations, we chose the models with the best fit. We examined all possible interactions and found that time and gender had the only significant interaction for the “ever had sex” variable.
The results in univariate analysis of all 6231 eligible questionnaires showed, as expected, significantly different sexual behavior between boys and girls. For example, boys had a significantly higher odds ratio (OR) for “ever had sex” than girls, had used condoms less often, and were more likely to have had more sexual partners. Therefore, we decided to carry out our statistical analysis for male and female students separately. Because we used a cluster sampling method, with the sampling unit being a school, we also analyzed the data taking this into account. Results from this analysis were similar to the original results. To reduce possible bias in analysis of time trends, the age range of students was limited to 12 to 20 years. The internal consistency of the data was good. For example, none of those who reported that they had never had sex reported using a condom. For the FGDs, content analysis was carried out using standardized analytic procedures for qualitative studies.16
The study was approved by the District Medical Directors and the District Education Officers of the 3 districts. All headmasters of the sampled schools agreed to participate in the study. The headmasters informed parents about the study and gave consent on their behalf. This followed the established procedures for research in schools as established by the Kabarole District Government and the Ministry for Education. Because this was a surveillance program of the Ugandan Ministry of Health, ethical clearance was not thought to be required by the research team (who mainly handled the data analysis and interpretation). Ethical clearance for the test-retest study and for the qualitative substudy was given by the Health Ethics Research Board of the University of Alberta and by the Uganda Council for Science and Technology.
Between 1995 and 2001, a total of 7098 questionnaires were collected. Of these, the following were excluded: 709 (10.0%) because of age older than 20 years (69.7% of male students, maximum age of 33 years), 4 (0.1%) because of age younger than 12 years, and 154 (2.2%) because of incomplete data on age or gender, leaving 6231 questionnaires for analysis. The response rate for the questions was between 97.8% and 94.6%. There was only a 77% response rate to the question asking about age at the first sexual encounter.
Of the 6231 respondents, 3271 (52.5%) were male and 2960 (44.5%) were female. The mean age of all participants was 16.7 years (SD = 1.6 years, range: 12-20 years). The mean age of male students was slightly higher than that of female students (17.1 years, SD = 1.6 years vs. 16.2. years, SD = 1.5 years). A total of 4941 (79.3%) of all participants were in the age group between 14 and 18 years (76.8% of male students and 82.1% of female students). A total of 477 (7.7%) participants were younger than 14 years of age (4.4% of male students and 11.3% of female students), and 813 (13.0%) were older than 19 years of age (18.8% of male students and 6.7% of female students). Over time, the mean age increased in male students from 16.6 to 17.6 years and for female students from 15.7 to 16.7 years. A total of 12.1% of all respondents said that they had had an HIV test, and 68.3% were interested in getting an HIV test. A total of 5.6% of girls said that they had had a child (Table 2).
Results of Qualitative Substudy
Most students said that they generally liked the questionnaire and that the questions were easy to understand. For example, 2 students stated:
▪ I really don't think there is any complicated term here and what maybe we know as slangs is understood here (16-year-old girl).
▪ Some of us may not want to explain by mouth why we did not use a condom, but it is easy to write it (17-year-old girl).
Many students in all FGDs said that there is a high level of sexual activity in their age group. In comparison, the 1998 questionnaire indicated that 53.5% of the students reported having had sex. Most students also thought that their peers start sexual activity early. In the 1998 questionnaire, mean age at first sexual intercourse was 13.8 years for male students and 14.0 years for female students. Most respondents said that they believed students in their age group have several partners (in the 1998 questionnaire, the median number of lifetime sexual partners was reported to be 4). Condom use was considered widespread. The suggested percentage of students having ever used a condom mentioned in the FGDs ranged between 60% and 80% (in the 1998 questionnaire, it was 65.1%). Students also said that they believed condom use has increased substantially over the past few years.
Knowledge of Sexually Transmitted Infection Prevention
Knowledge about the prevention of sexually transmitted infections (STIs) increased significantly over the period under observation. Specifically, knowledge that condom use is an effective way to prevent transmission of HIV and STIs increased significantly during the period. In contrast, the level of knowledge that being faithful to 1 partner is an effective way to prevent HIV and STI transmission decreased slightly in the same period. Descriptive data on general information about the students and how they tried to prevent HIV and STIs are shown in Table 3.
Results of ABC (Abstinence, Be Faithful and Condom use)
Compared with 1995 to 1996, male students were slightly more likely to report ever having had sex in 1997 to 1998 and less likely in 2000 to 2001. This was different for female students, in whom the HR for ever having had sex increased substantially in 1995 to 1996 and 1997 to 1998 and then decreased only slightly in 2000 to 2001. Overall, urban students were more likely to have had sex (HR = 1.31 and 1.24 for male and female students, respectively; see Table 4).
Reported Sexual Activity
In univariate and multivariate analysis, there was no significant change in the trend over the years in sexual activity in the previous 3 months. Increasing age was the only statistically significant predictor of reporting having had sexual contacts in the past 3 months (see Table 5).
Be Faithful (B): Number of Sexual Partners
The number of sexually active students who reported 2 or more lifetime sexual partners decreased slightly over the years. Male students did not show a trend in reducing the number of partners (2 or more) over the period, whereas fewer female students reported having multiple partners between 1995 to 1996 (OR = 1.0) and 2000 to 2001 (OR = 0.73), which was borderline significant (P = 0.69). Conversely, fewer male urban students reported having multiple sexual partners (OR = 0.84; P < 0.01) compared with rural students, although only a small nonsignificant difference (OR = 0.95) in urban female students compared with their rural peers was observed (see Table 5).
Ever Condom Use (C)
Reported ever use of condoms by sexually active students increased significantly for all participants from 78.4% in 1995 to 1996 to 94.8% in 2000 to 2001. The OR for reported condom use also increased substantially for male and female students. The OR for reported ever condom use increased substantially for male and female students during the period of the study (2.51 and 2.59, respectively). This was the most highly significant trend in all sexual variables. Urban female students reported being more likely to use condoms than rural female students (OR = 1.29; P = 0.078). There was no difference between the condom use of urban and rural male students (see Table 5).
We measured trends over time in sexual knowledge and behavior of secondary school students in 3 districts in western Uganda between 1995 and 2001 as an evaluation of the Ugandan Government/GTZ HIV/AIDS education program using the ABC approach. Male students reported being more abstinent over time, whereas, in contrast, female students reported being less abstinent over time. Although male students did not report changes in having multiple partners over time, female students did (but only borderline significant). Sexually active male and female students did not report a significant change in their sexual activity in the past 3 months. The most striking finding was the reported increased condom use by male and female students. Therefore, we conclude, overall, that the ABC approach has worked in secondary school students.
Self- and interview-administered questionnaires are the most commonly used methods of data collection on sexual behavior.18 Self-administered questionnaires are also often used in school classes for several reasons: they provide considerable privacy for the respondent, they can be administered by 1 person to a large group, and they are relatively inexpensive to use in group situations.19 The reliance of sexual behavior research on self-reported data, combined with the sensitive nature of the topic, renders it susceptible to problems of validity and reliability, however.20 We addressed both of these issues by testing the questionnaire for its reliability with a test-retest experiment and by assessing its validity through comparison with a comprehensive qualitative substudy. Based on the test results outlined earlier, we are fairly confident that the questionnaire was a valid and reliable tool to measure knowledge and sexual behavior in our study population with a reasonable degree of accuracy. The increased condom use reported by our study participants was indirectly validated by the numbers of condoms distributed or sold in the study area, which steadily increased over the years and totaled 3.7 million condoms distributed between 1992 and 2001.15
Other studies have compared trends of sexual behavior over time. In Masaka district, Uganda, an adult cohort of 10,000 persons was followed for 7 years, and declining HIV prevalence was reported.8 Two studies on sexual behavior trends were also conducted in adults in Zambia. Both studies reported declining HIV-1 prevalence, which was explained by increased condom use, an increase in age at the first sexual contact, and a decrease in the number of sexual partners.21,22 A systematic review of HIV/AIDS trends in Uganda by Singh et al23 acknowledges increased condom use and an increasing number of single women having less than 2 partners between 1995 and 2000 as causes for a declining HIV prevalence. This generally agrees with our findings. Singh et al23 also reported that the number of sexual partners of unmarried men has substantially increased over the same period, however. This is different from our study, where we found that our male participants (all were unmarried students) did not report an increased number of sexual partners over time. This difference may be explained by the inclusion of unmarried men who were not in school in the analysis of Singh et al.23 The main difference between these cited studies and our study is the age of participants. Although our participants were mainly adolescents, other studies reported results mainly in adults. This makes our study findings especially important, because knowledge about the sexual knowledge and behavior trends in young people is necessary to address their needs in HIV and STD prevention programs better.
Our study findings that sexual behavior trends include increased condom use, reduction of the number of sexual partners by female students, and increased abstinence in male students may explain the decline in HIV-1 prevalence that has been observed in Kabarole district, especially in the younger age groups.4 The HIV prevalence in girls and women younger than the age of 20 years declined from 19.9% in 1995 to 10.0% in 2001 in our study area (Basic Health Services Project annual report, 2004 [unpublished]). Recent findings from the Rakai district in Uganda indicate that death was a major explanation for a declining HIV prevalence and increased condom use among adults in this district.24 Death is not a likely explanation for the declining HIV prevalence in our young participants. Death attributable to AIDS in our sample was probably quite rare because of the long time span between initial HIV infection and death attributable to AIDS. Therefore, the most probable explanation for the declining HIV prevalence in this population is behavior change. Can we conclude that condom use, increased abstinence by male students, and partner reduction by female students are the explanation for this significant decline in the HIV prevalence? Our data cannot answer this question. There may also be other sexual behaviors involved that we did not measure and could have affected HIV transmission (eg, coitus interruptus, local herbs or substances used intravaginally before intercourse). Our data demonstrate that condom use showed the greatest change over the period under observation compared with the changes in the other sexual behavior variables measured. This means that condom use was a major mechanism for improving safe sex among our participants. The other positive changes of more abstinence and reduction of the number of sexual partners also have to be acknowledged and appreciated as contributing to safe sex behavior, however. The multimethod approach to HIV prevention programs has been reflected by other investigators who maintain that the decline in HIV prevalence in Uganda is best explained by a combination of several interventions implemented in Uganda using the ABC approach.23,25,26 Our study results confirm this.
Our study has the following limitations:
1. Even with careful validation of the self-reported responses in the questionnaire, we cannot completely exclude a reporting/selection/social desirability bias. As Table 1 shows, however, the responses of students in the first, second, and third interviews corresponded quite well. The results from the test-retest exercise also revealed good agreement for the key questions. This, together with the results from the qualitative evaluation, gives us some confidence that the responses of our students were valid.
2. We did not assess attendance of the students in the schools. Therefore, we cannot report the magnitude of no participation. If absenteeism rates varied over time, this might have introduced a selection bias, because absent students may have been different from the general population of secondary students. Also, girls who became pregnant during the time this program was implemented were likely to have dropped out of school. Similarly, we did not record students who dropped out of school permanently or new students who joined a school during the period of the study.
3. Because this study questioned only school students and did not include youth who were not enrolled in school, the results from this study can only be generalized to the school population in this age group. We estimate that the average enrollment in secondary school during the time of the study was around 30%, being higher in S2 classes and lower in S4 classes.
The positive sexual behavioral changes that we observed in secondary school students between 1995 and 2001 is likely one explanation for the declining HIV prevalence observed in this population in the same period. The observation that this change in sexual behavior is largely attributable to an increase in condom use and, to a lesser extent, to increased abstinence and a reduction in the number of sexual partners contradicts the assertions that an “abstinence-only” message is what is required. Our evidence points to the importance of condom use and its need to be an essential approach for changing sexual behavior. One other important result of our study is the different ways in which male and female students changed their sexual behavior. This suggests that different approaches are needed for male and female students when they are educated about HIV/AIDS and other STDs. Most school classes are mixed, and education sessions are held for the entire class. Because urban students were more likely to change their sexual behavior compared with rural students, there is also an urgent need to focus HIV/AIDS prevention efforts primarily on rural communities.
The findings of our study do not advocate for any particular method of achieving sexual behavioral change in young people. Condom use cannot do the job alone, nor can it be left out of the options people are given for their choices about safe sexual behavior. One important lesson from decades of family planning programming is applicable here: the more contraceptive methods that were offered to give people more choices, the more likely they were to select a method that was most appropriate for their family planning needs. Any HIV/AIDS control program that wants to be successful should promote as many methods as are known to be effective to reduce HIV transmission.
We would like to thank the Ministry of Health Uganda for it's support in this work and permission for publication. We would also like to acknowledge the valuable contributions of a number of people over time, particularly Nazarius M. Tumwesigye, Professor Frank von Sonnenburg, University of Munich, Germany, and Drs. Peter Weis, Uli Wagener, and Ulrich Vogel from GTZ headquarters, Eschborn, Germany, who have assisted in the development and improvement of the questionnaire and Ms. Gudrun Sahlmüller, GTZ AIDS project Fort Portal, Uganda, who was instrumental in the initialization of the surveys. We also thank Narmatha Thanigasalam and Arif Alibhai for help with the statistical analysis.
1. Okware S, Opio A, Musinguzi J, et al. Fighting HIV/AIDS: is success possible? Bull World Health Organ
2. STD/AIDS Control Programme. HIV/AIDS Surveillance Report
. Kampala, Uganda: Ministry of Health; 2001.
3. Fabiani M, Accorsi S, Lukwiya M, et al. Trend in HIV-1 prevalence in an antenatal clinic in North Uganda and adjusted rates for the general female population. AIDS
4. Kilian A, Gregson S, Ndyanabangi B, et al. Reductions in risk behaviour provide the most consistent explanation for declining HIV-1 prevalence in Uganda. AIDS
5. Kamali A, Carpenter LM, Whitworth JA, et al. A. Seven-year trends in HIV-1 infection rates, and changes in sexual behaviour, among adults in rural Uganda. AIDS
6. Wawer M, Serwadda D, Gray RH, et al. Trends in HIV-1 prevalence may not reflect trends in incidence in mature epidemics: data from the Rakai population-based cohort in Uganda. AIDS
7. Whitworth JA, Mahe C, Mbulaiteye SM, et al. HIV-1 epidemic trends in rural south-west Uganda over a ten year period. Trop Med Int Health
8. Mbulaiteye SM, Mahe C, Whitworth JA. Declining HIV-1 incidence and associated prevalence over ten years in a rural population in south-west Uganda. Lancet
9. Asiimwe-Okiror G, Opio AA, Musinguzi J, et al. Change in sexual behaviour and decline in HIV infection among young pregnant women in urban Uganda. AIDS
10. Konde-Lule JK, Wawer MJ, Sewankambo NK, et al. Adolescents, sexual behaviour and HIV-1 in rural Rakai district, Uganda. AIDS
11. Ahmed S, Lutalo T, Wawer M, et al. HIV incidence and sexually transmitted disease prevalence associated with condom use: a population study in Rakai, Uganda. AIDS
12. Joint United Nations Program on HIV/AIDS and World Health Organization Working Group on Global HIV/AIDS and STI Surveillance. Guidelines for Second Generation Surveillance for HIV
. WHO/CDS/CRS/EDC/2000.5, UNAIDS/00.03E. Geneva, Switzerland: World Health Organization; 2000.
13. Deutsche Gesellschaft für Technische Zusammenarbeit. Responding to AIDS in the Developing World: the GTZ Contribution
. Eschborn, Germany: GTZ; 1995:38-39.
14. Kamugisha J, Kipp W. District AIDS Control Programme: the Kabarole District Health Team Experience
. Entebbe, Uganda: Ministry of Health; 1994.
15. Kilian A. Building a Healthy Population: Seventeen Years of Basic Health Services in Western Uganda
. Eschborn, Germany: Deutsche Gesellschaft fuer Technische Zusammenarbeit (GTZ); 2005.
16. O'Connor H. Surveillance of sexual behavior in secondary school students in western Uganda. MSc thesis, University of Alberta, Edmonton, Canada, Fall 1999.
17. Zaba B, Pisani E, Slaymaker E, et al. Age at first sex: understanding recent trends in African demographic surveys. Sex Transm Dis
. 2004;80 (Suppl II):28-35.
18. Konings E, Bantebya G, Carael M, et al. Validating population surveys for the measurement of HIV/AIDS prevention indicators. AIDS
19. Catania JA, Turner H, Pierce RC, et al. Response bias in surveys on AIDS-related sexual behavior. In: Ostrow DG, Kessler RC (eds). Methodological Issues in AIDS Related Behavioral Research
. New York: Plenum Press; 1993:133-158.
20. Carballo M, Cleland J, Deheneffe JC, et al. Research agenda: a cross national study of patterns of sexual behavior. J Sex Res
21. Bloom S, Banda C, Songolo G, et al. Looking for change in response to the AIDS epidemic: trends in AIDS knowledge and sexual behavior in Zambia, 1990 through 1998. J Acquir Immune Defic Syndr
22. Fylkesnes K, Musonda RM, Sichone M, et al. Declining HIV prevalence and risk behaviors in Zambia: evidence from surveillance and population based surveys. AIDS
23. Singh S, Darroch J, Bankole A. A, B and C in Uganda: the Roles of Abstinence, Monogamy and Condom Use in HIV Decline
. Occasional Report No. 9. New York: The Alan Guttmacher Institute; 2003.
24. Wawer M, Gray R, Serwadda D. Declines in HIV prevalence in Uganda. Not as simple as ABC [abstract 27LB]. Presented at: 12th Conference on Retroviruses and Opportunistic Infections, Boston, MA; 2005.
25. Cohen S. Beyond slogans: lessons from Uganda's experience with ABC and HIV. Reprod Health Matters
26. Wilson D. Partner reduction and the prevention of HIV/AIDS. BMJ