In 2000, an estimated 12% of adults aged 15–49 in Mozambique were infected with HIV.1 Most estimates show an increasing prevalence of infection, and about one-half of new infections occur among 15–24-year-olds.2 Nevertheless, according to the 1997 Mozambique Demographic and Health Survey (DHS), young adults had an extremely low rate of condom use.3
Individuals' knowledge of HIV transmission and accurate assessment of their own risk seem to be among the key factors in adoption of safer sexual practices.4 Policymakers must understand these factors to design effective policies in the fight against HIV and AIDS. Adolescent and young adult behavior is of special interest for several reasons. First, the number of life-years saved is greatest when infections are averted in relatively young individuals. Second, preventing HIV infection in women of childbearing age prevents transmission from mother to child. Finally, it may be easier to change sexual attitudes, practices and risky behaviors among the young than among older people.5
In Mozambique, the response to HIV/AIDS has included educational programs, media campaigns, social marketing of condoms, and voluntary counseling and testing services. All of these programs aim to provide individuals with information about HIV transmission, thereby allowing them to make informed choices. The hope has been that acquired knowledge will lead to changes in sexual behavior that will help prevent the spread of the disease. However, policymakers and program officials have not had access to adequate population-based data for adolescents and young adults to develop targeted programs and evaluation strategies.
KNOWLEDGE, RISK PERCEPTION AND BEHAVIOR
According to the AIDS risk reduction model, knowledge of AIDS is a prerequisite to recognizing risky behavior and taking action to change it,6 but findings regarding the relationship between knowledge and behavior have been inconsistent. In a review study, Becker and Joseph found an ecological correlation between AIDS knowledge and behavior among bisexual or homosexual men and young heterosexual adults.7 However, they did not find clear evidence on the mechanism of this correlation and suggested that a third factor may link knowledge to behavior. In Zambia, Magnani et al. found HIV/AIDS knowledge to be associated with a reduced probability of sexual experience and an elevated probability of condom use among boys.8 A cross-sectional study using DHS data from Uganda, Kenya and Zambia showed that knowing somebody with AIDS was predictive of protective sexual behavior, as were knowledge of HIV prevention methods and correct beliefs regarding AIDS patients.9 The authors concluded that knowledge of someone who had AIDS or who had died of AIDS may increase an individual's awareness of the consequences of HIV/AIDS and may lead to safer sexual practices.
Other studies have reported a lack of association between HIV/AIDS knowledge and sexual behavior. In a survey of almost 1,500 Rwandan women, the vast majority had correct knowledge of HIV transmission, but only a small proportion had adopted any protective behavior in the last year.10 Sexually active students in Dar-es-Salaam, Tanzania, who knew that condom use prevents HIV infection had a reduced likelihood of always using condoms.11 In Ethiopia, a cohort study of factory workers with a high prevalence of HIV reported high-risk sexual behavior and low condom use, even though the majority mentioned condom use as the best way to prevent HIV.12 A study using 1998 Kenyan DHS data reported that the odds of having risky sexual behavior were more than tripled among men and women who perceived their risk of HIV/AIDS as high, and found no association between knowledge of HIV transmission and sexual behavior.13
The relationship between perception of risk and sexual behavior is complex and poorly understood. Studies conducted in different cultures have associated HIV risk perception with a wide range of variables: number of sexual partners, knowledge of sexual partners' past sexual behavior, fear of AIDS, shame associated with having AIDS, community perception of AIDS risk, knowing someone with AIDS, discussing AIDS at home, closeness of parent-child relationships and religious affiliation.14 In Sub-Saharan Africa, sociocultural norms and practices are major determinants of sexual risk-taking behavior.15
A study of South African couples found that women who considered themselves at risk of HIV because of their husbands were four times as likely to use condoms as women who did not.16 In a study of university students in Zimbabwe and Nigeria, those who used condoms were more likely than nonusers to have an accurate perception of their HIV risk.17 Personal risk perception was also associated with increased condom use among urban youth in Cameroon.18 In Ghana, self-perceived high risk among youth was associated with sharply increased odds of condom use at last sex.19
The goal of our study is to examine whether correct assessment of HIV risk is associated with condom use at last sex among young adults in Mozambique. We describe youths' attitudes toward HIV and AIDS, and determine whether they have correct information on how the virus is transmitted. We also analyze whether individuals' self-assessment of risk is accurate in light of their sexual behavior.
Our study uses data from the Adolescent and Young Adult Reproductive Health and Behavioral Risk Survey (known by its Portuguese acronym, INJAD), a national, population-based survey conducted from July to November 2001.* The survey included questions about school attendance and sex education, knowledge of contraception, sexual experience and current sexual activity, opinions about condoms, sexuality and gender issues, prenatal care and maternal morbidity, HIV/AIDS, and physical and sexual abuse.
Details about the INJAD methodology can be found in the survey report.20 The survey followed a two-stage probability household sample design, using the master sample from the 1997 census as a sampling frame. Complete interviews were conducted with 5,338 females aged 15–24 (for an 88% response rate) and 5,150 males in the same age-group (81% response rate). Our analyses are based on sexually experienced respondents—3,831 males and 3,986 females. The female and male samples were independent of each other. Differences between sample estimates were calculated using an average design effect of 1.5.
Respondents assessed their risk of HIV infection as none, small, moderate, high or not known. To evaluate the accuracy of perceived risk, we compared these self-assessments with assessments imputed on the basis of current and past sexual behavior. For the imputed assessment, we summed the number of risk factors respondents reported from among the following: having not used a condom at last sex, having had an occasional partner at last sex, having had three or more partners in the last 12 months, having had six or more cumulative partners, having a history of STIs and, for never-married women, having received money or gifts in exchange for the most recent sexual encounter. Respondents reporting fewer than two risk factors were defined as having no risk or a low risk of HIV infection; those reporting two or more were defined as having a moderate or high risk of HIV.
We used chi-square and t tests to compare the proportions of respondents, within and across subgroups, who correctly assessed their HIV risk and the proportions who reported condom use at last sex. To test the association between condom use and correct assessment of HIV risk, controlling for demographic, behavioral and HIV-related variables of interest, we used a probit regression model with condom use at last sex as the outcome variable. To run the model, we removed condom use from the risk definition variable, and removed all of the variables included in the imputed risk definition from the righthand side of the equation. In this way, each variable appears only once, in one side of the equation, and the problem of endogeneity is avoided.
Data analysis was performed with Stata 6. Coefficients shown are equal to the change in the probability of condom use associated with a change of one unit in the independent variable. All results are given separately for females and males; within each gender, results are disaggregated by marital status. Marital status was coded as ever-married or never-married; the ever-married category includes respondents who were married, cohabiting, separated, divorced, widowed, or legally married but not living in the same house as their spouse.
Slightly more than half of both sexually experienced males and their female counterparts were 20–24 years old, and about seven in 10 of each came from rural areas (Table 1, page 193). The vast majority of respondents had no formal schooling. Twenty-eight percent of men and 71% of women had ever married.
To explore respondents' reproductive lives, we used questions that asked them to recall the first time a life event occurred. For women, the median age at menarche was 14.0 years, the median age at first sex was 15.9, and the median age at first pregnancy and at first marriage was 19.0 (not shown). For men, the median age at first sex was 15.6 years, the median age at the birth of the first child was 23 and the median age at first marriage was 24.
Fifty-seven percent of sexually experienced males had had four or more partners; 12% of women reported this cumulative number of partners (Table 1). Similarly, 53% of men and 9% of women had had two or more partners in the last 12 months (not shown). Nine percent and 7%, respectively, had last had sex with an occasional partner. Among never-married women, 19% reported receiving money or goods in exchange for sex at the time of the last sexual encounter. Among the men, one in five did not know their last partner's age, and three in 10 said that she was 1–2 years younger. Almost a third of the women did not know the age of their last partner; another third said that he had been five or more years their senior.
Some 18% of men and 22% of women reported a history of symptoms or signs associated with an STI. Thirty-eight percent of men and 34% of women had known someone who was HIV-positive or who died of AIDS. However, the use of voluntary counseling and testing services was very low; fewer than 5% of either men or women had obtained services. About three-fourths of respondents said that they would care for a family member who was HIV-positive, but 55% of women and 47% of men would keep it a secret if a relative had AIDS (not shown).
When asked how HIV is transmitted, only 8% of men and 3% of women gave answers that were incorrect (e.g., by kissing or holding hands, or through witchcraft or mosquito bites); 98% of men and 91% of women spontaneously mentioned at least one documented cause of transmission (not shown). However, only 79% of men and 74% of women knew that a person could look healthy and be infected with HIV. These findings are consistent with previous survey reports that people may know some specific modes of transmission, but lack a general understanding of the disease.21
HIV Risk Assessment
Overall, 32% of women considered themselves at no or low risk of contracting HIV, 22% thought they were at moderate or high risk, and 46% did not know how to assess their HIV risk. According to our imputed estimates, 27% of women who considered themselves at no or low risk, and 23% who reported not knowing how to assess their risk, were at moderate or high risk of HIV infection (Table 2). The situation for men is even more troubling. Thirty-eight percent said that they had no or low risk of acquiring HIV, 46% that they had a moderate or high risk, and 17% that they did not know (not shown). However, 80% of those who considered themselves at no or low risk, and 92% of those who did not know how to assess their risk, were classified as having a moderate or high risk of HIV infection.
According to our risk reassessment, 53% of men and 46% of women had an accurate perception of their risk (Table 3). The proportions were 45% and 43%, respectively, among ever-married men and women, and 56% and 50% among their never-married counterparts; the differences by marital status were statistically significant (not shown). Furthermore, within each subgroup of marital status, correct assessment of risk differed according to an individual's characteristics. For example, correct assessment of risk increased with level of education among women, particularly among never-married women, whereas it showed no clear pattern in relation to education among men. Correct assessment also increased with number of lifetime partners for women, regardless of marital status, but showed no clear pattern for men.
Among never-married respondents, the likelihood of correct risk assessment rose significantly with age at first intercourse among women but was not associated with this characteristic among men. However, among ever-married respondents, it was inversely associated with men's age at first sex and not associated with women's. Overall, the proportion of respondents who correctly assessed their risk was significantly higher among those who reported condom use at last sex (70% of men and 71% of women) than among those who did not (48% and 42%, respectively). Furthermore, the differences in levels of correct assessment between never-married men and women who had used a condom at last sex (72% and 76%, respectively) and their ever-married counterparts (61% and 58%, respectively) were statistically significant (not shown).
Partner type was related to correct assessment except for ever-married females and never-married males; never-married women who had last had sex with an occasional partner were less likely than those reporting a regular partner to assess their risk correctly (24% vs. 56%). Across all subgroups, those with a history of STI symptoms were more likely than others to assess themselves correctly. For women in both marital status groups and for ever-married men, the likelihood of correct assessment was elevated among those who had known someone with AIDS. Although the use of counseling and testing services is extremely rare in Mozambique, females who had used these services appeared to be more likely than others to make a correct assessment of risk (72% vs. 45%); this relationship holds both for those who had ever married (76% vs. 43%) and for those who had not (70% vs. 50%). Additionally, a greater proportion of females than of males who had used counseling services made a correct assessment of their risk.
Condom Use at Last Sex
According to INJAD, the prevalence of condom use was low overall in 2001: Twenty-two percent of men and 10% of women reported use at last sex (Table 4), and the difference was statistically significant (not shown). Prevalence among never-married women was much higher than average (25%) and was similar to the level among never- married men (26%). The likelihood of condom use was positively related to age only among never-married men; in all subgroups, it was higher in urban areas than in rural areas, and it increased with level of education. Condom use was low among ever-married individuals, but was significantly more common among those whose last partner had been someone other than a spouse. Among ever-married males (but not their female counterparts), condom use was significantly more common with regular or occasional partners than with spouses. Respondents who did not know their last partner's age—particularly females, regardless of their marital status—were less likely than others to have used condoms at last sex. Having a history of STI symptoms was related to an increased likelihood of condom use at last sex for all subgroups except never-married men, and having known someone with AIDS and having used counseling and testing services were positively associated with condom use for every subgroup. Finally, condom use was more than twice as common among respondents who assessed their risk correctly as among those who did not (for males, 30% vs. 14%; for females, 16% vs. 6%).
According to our probit regression model, sexually experienced men overall were 15% more likely to use condoms if they correctly assessed their risk of HIV infection than if they did not (Table 5). The differential was more than twice as high among never-married men as among ever-married men (18% vs. 7%). Males who had used counseling and testing services were 16% more likely to use condoms than were those who had not. This difference was significant only for never-married men. Having known someone with AIDS was associated with a significantly increased likelihood of condom use by males, regardless of their marital status. Findings for the demographic characteristics in the probit regression model were consistent with the bivariate results.
Among women overall, correct assessment of risk was associated with a 5% increase in the likelihood of condom use (Table 6, page 198); the association was significant only for never-married women, for whom the differential was considerable (17%). Similarly, the use of counseling and testing services was associated with a significantly higher likelihood of condom use, but only among never-married women (19%). Having known someone with AIDS was associated with increased condom use among women, regardless of their marital status. Findings for the demographic variables in the model were consistent with the results obtained in the bivariate analysis.
Studies of the influence of AIDS knowledge on condom use have reported mixed results. Some have found positive associations,22 and others no association after various social and demographic characteristics were controlled for.23 However, these studies have involved different population subgroups at different risk levels (e.g., gold miners and commercial sex workers). In addition, they have used an array of questions to measure knowledge of AIDS, from general knowledge about HIV transmission and accuracy of responses given about AIDS, to a general understanding of the disease (e.g., how one can became infected or how one can avoid contracting HIV). Our study, which uses a national population-based household sample, representative of all Mozambican youth aged 15–24, contributes to the literature on condom use predictors by looking at the accuracy of individuals' assessment of their own risk, based on current and past sexual behavior.
We found a tendency for young adults, especially young men, to underestimate their risk of contracting HIV: Some 27% of women and 80% of men who considered themselves at low or no risk were actually at moderate or high risk of HIV infection. Furthermore, even though men and women overall had accurate knowledge of HIV transmission modes, 17% and 46%, respectively, did not know how to assess their risk. Two possible reasons for this discrepancy are that women may not apply their knowledge of disease transmission to assess their risk level every time they engage in sexual activity and that women were more reluctant than men to report their self-assessment of risk.
Existing literature has shown mixed results on the association between risk perception and sexual behavior. Our findings suggest that condom use at last sex in this sample was related to a number of factors operating at the individual and environmental levels. It was associated with most of the demographic and behavioral variables included in the analysis. Previous studies have shown positive associations between education and condom use;24 ours shows a similar trend, except for women with a secondary or higher education.
Our AIDS-related variables—having known someone with AIDS, having used voluntary counseling and testing services, and correctly assessing one's own risk—were also positively associated with condom use. However, results differ between men and women, and by marital status. Condom use at last sex appears to be more strongly related to correct assessment of risk among never-married than among ever-married men and women, and more among males than among females.
Similarly, the use of counseling and testing services was associated with condom use, but only among never- married men and women. Because respondents who had access to and chose to use services are a highly self-selected subsample, these results should be interpreted cautiously. Factors associated with choosing to obtain services may be the cause of the difference.25 A study by Sweat et al.26 found that individuals obtaining voluntary counseling and testing services in Tanzania and Trinidad were 3–4 times as likely as those in comparison household probability samples to report unprotected sex with nonprimary and commercial partners; following voluntary testing and counseling, condom use increased by 66%. Because of low access to and availability of voluntary counseling and testing services in Mozambique, participation in counseling and testing is unlikely to be associated with an increase in condom use of the same magnitude reported by Sweat et al. in Tanzania and Trinidad. Studies in Sub-Saharan Africa have shown that as many as 90% of individuals of reproductive age would use voluntary counseling and testing services if they were available.27 It is plausible that Mozambican youth who chose to obtain services were more likely than others to change their behavior. On the other hand, those who used services may have done so because they were in relatively high-risk situations, and they may have had little control over behavior such as condom use.
The levels of condom use at last sex reported in our study (22% among men and 10% among women) represent significant increases over the levels found in the 1997 DHS (7% and 2%, respectively28). The finding that use remains lower among ever-married than among never-married individuals is consistent with findings from previous work.29 With HIV prevalence high in Mozambique, it is important to encourage condom use in all types of sexual relationships, including consensual and legal unions, since both married and unmarried individuals engage in risky sexual behaviors (e.g., multiple partners and unprotected sex with nonregular partners). Furthermore, condom use may signal mistrust.30 Addressing these issues in information and education campaigns, including encouraging testing for married and regular partners and continued use of condoms until the couple is able to undergo counseling and testing, should be an important component of HIV prevention efforts.
The AIDS epidemic is complex, and successful efforts to limit transmission must rely on a combination of medical, social and behavioral approaches. It is clear that young people should be at the center of strategies to control HIV infection. The population-based data from this survey provide information that will enable program officials and policymakers to design more effective HIV prevention programs for Mozambican youth. Adolescents' and young adults' reproductive health knowledge and their sexual and contraceptive behavior can have important implications for their health and well-being, as well as for the continuation of their education. In many countries, few young people are equipped with the information, skills and resources needed to deal with a healthy transition to adulthood. Adequate programs, including media campaigns, and quality sex education for both in-school and out-of-school youth, would provide important information that would likely enhance young people's ability to correctly assess their risk of HIV infection and increase their use of condoms.