In what UNICEF has called a ‘crisis of gargantuan proportions’, an estimated 15 million children under age 18 have lost one or both parents to AIDS. It is expected that the absence of adult protection and provision leave orphaned children susceptible to abuse and desperation, thereby increasing their vulnerability to HIV infection and perpetuating a cycle of orphanhood and risk .
Despite this concern, only two published studies from Africa have explored links between orphanhood and sexually transmitted risk among adolescents. In rural Zimbabwe, an analysis of 1523 teenagers found that young women were at highest risk of HIV if they had lost their mother or both parents; no associations were found among orphaned males, or with the loss of a father among females . A study of 1694 teenagers in Kwa-Zulu Natal, South Africa, found that both male and female orphans were more likely than non-orphans to have had sex, and to have initiated sex earlier . The authors did not present differences by type of orphanhood.
Related research has focused on the influence of parental presence versus absence (for any reason) on the sexual health of adolescents. Among adolescents living in slums of Nairobi, never-married 12 to 19-year-old girls who lived with their father or both parents were less likely to have ever had sex and to have experienced unwanted pregnancies than those whose fathers did not live in the same household . In a cross-sectional survey in Lusaka, Zambia, 10 to 24 year olds who lived with both parents were less likely than those not living with both parents to have had sex . A nationally representative survey of unmarried 12 to 24 year olds in Ghana found that girls but not boys who lived with both parents were less likely to have had sex than those who lived with neither parent .
The Zimbabwean context
There is evidence that both the incidence and prevalence of orphans are rising in Zimbabwe, with the majority of children orphaned as a result of AIDS [7,8]. In Uganda, orphan incidence continued to rise for 14 years after the peak in HIV incidence was reached . In Zimbabwe, the highest increases are predicted among those who lose their mother or both parents [7,8,10].
A heightened HIV risk among the growing population of orphans in Zimbabwe could slow or reverse recent declines in HIV prevalence, which have, so far, been concentrated at young ages . This study was designed to assess this risk among adolescent girls in Harare.
The study was conducted in Highfield, a high-density residential area of Zimbabwe's capital, Harare in 2004, in collaboration with the Child Protection Society (CPS). In order to meet the estimated sample size of 800 girls (required to detect a doubling of HIV risk among each type of orphan, with 80% power and a 5% level of significance, assuming 10.6% HIV among non-orphans and up to 3.6 non-orphans to each orphan – based on the Zimbabwe Young Adult Survey 2001–2002), a random sample of 80 of the 246 census enumeration areas in Highfield was selected, stratified by electoral ward. Within each area all households were visited by community volunteers of CPS, and all 15 to 19-year-old girls were invited to participate. A household survey was conducted in the home of each invited girl, including questions about family composition, physical aspects of the dwelling, material assets and needs, and conditions of the street. As exploratory research revealed that few girls of the target age lived outside of homes (e.g., in institutions or the street), a household sample was considered adequate to capture a representative sample.
Participants of the community survey who gave informed consent were interviewed at a central, convenient and private location, by a young female social science graduate. Interviews included modules on: living situation; education; parents/caregivers and family; pregnancy; sexual experiences; and HIV/AIDS prevention and services.
After further consent, a nurse collected a biological specimen (either a venous blood sample, a dried-blood spot sample, or an oral fluid sample, depending on the participant's preference). Biological samples were transported daily to the Biomedical Research and Training Institute (BRTI) laboratory, Harare, and tested for HIV (Vironostika, Lab Assist; confirmation of positives with Determine; Abbott Diagnostics, Abbott Park, Illinois, USA) and herpes simplex virus type-2 (HSV-2; Kalon Biological, Guildford, UK). The HSV-2 tests were only possible on the venous blood samples.
Confidentiality of the participants was protected through linked anonymized testing. To know their HIV status, all participants were given a voucher for free voluntary counselling and HIV testing at a New Start Centre. New Start offered high-quality, long-term services (i.e., voluntary counselling and testing, ongoing counselling, peer support groups, nutritional programmes, Moving On clubs) and afforded appropriate privacy to the adolescents, avoiding pressure on project staff and participants to disclose test results to parents and guardians. Participants were offered drinks and snacks during the interview; a thank-you packet containing personal toiletries; reimbursement for transport; and referrals to nearby health services for adolescents. Participants who requested help due to recent sexual abuse were accompanied to the Family Support Clinic at Harare Central Hospital for free medical examination, counselling, HIV/sexually transmitted infections (STI) testing and legal services.
Ethics permission for the study was received from MRC-Zimbabwe (MRCZ/A/1126), BRTI (AP62/03), and the London School of Hygiene & Tropical Medicine ethics committee (1023).
Data were double-entered in EpiData2 (EpiData Association, Odense, Denmark) and analysed using STATA 9 (Stata Corp., College Station, Texas, USA). Participants were classified according to whether they had lost their father only (paternal orphans), mother only (maternal orphans), both parents (double orphans) or whether both parents were still alive (non-orphans). The main sexual health outcomes were HIV and HSV-2 infection. In never-married participants, these were combined with pregnancy as a marker of unprotected sex, and referred to as ‘sexual risk’. This combination of outcomes was decided a priori, as pregnancy outside of marriage is generally considered socially undesirable in this cultural context.
The prevalence of adverse sexual health outcomes was compared among paternal, maternal, double and non-orphans, distinguishing those who lost a father, mother or both parents before or after age 12. Twelve was chosen to represent the transition from childhood to adolescence, and to be comparable with age groupings used by UNAIDS .
Life tables based on person-time were constructed to explore the timing of sexual debut and marriage in relation to orphanhood, using age when the parent died as a time-varying covariate . Person-time was censored at the date of the interview. Poisson regression analyses were used to compare rates of initiating sex by orphanhood status.
Participation in the survey
Community volunteers visited 2742 households in the selected areas, and conducted surveys in 1057 households in which at least one 15 to 19-year-old girl was living. They invited 1283 girls to participate: 863 (67%) came for an interview, of whom 839 (97%) agreed to provide a specimen (749 venous blood, 76 fingerprick sample and 14 oral fluid). The proportion agreeing to venepuncture was similar among orphans (87%) and non-orphans (88%).
A comparison of data from the household surveys suggested that girls who did not participate were from better-off households than the study participants (e.g., families of non-participants owned more assets and ate more meals per day); however, these differences were small and not statistically significant.
Although 15 to 19 year olds were recruited for the study, according to the girls' reported ages in the household surveys, when ages were calculated from dates of birth provided in the detailed interviews, 40 were aged 14 and two were 20 years old. Date of birth has been used as the indicator of age in the analysis.
Results were identical whether or not analyses accounted for clustering within households, and there was no evidence that outcomes among girls in the same enumeration area were correlated. Analyses that allowed for such clustering gave almost identical results to those that assumed independence, and the latter are reported.
Prevalence and experience of orphanhood
Half of the adolescent girls had lost one or both parents. One-quarter, 26% (n = 222), had lost their father only; 9% (n = 78) their mother only; and 15% (n = 127) both parents. In addition, 2% did not know the vital status of one parent. Among the 127 double orphans, most (63%) lost the second parent within 5 years of the first parents' death. In most cases (58%), the father died before the mother, and 10 double orphans lost both parents in the same year.
Participants who lost their parent while aged 12 or older (and thus more likely to remember) reported that 40% of their fathers' and 67% of their mothers' deaths were preceded by a chronic, debilitating illness. Few orphans said their father or mother had died of AIDS (3 and 5% respectively; n = 11 in both cases); the most common reasons cited were tuberculosis and headache (for either parent) and accident (for fathers only). When asked about recent parental illness, 3% (n = 18) of those with a living father and 10% of those whose mother was alive (n = 70) said their parent was severely ill during the past year.
Sexual health outcomes
Overall, 8% (63/826) of the blood and saliva samples tested positive for HIV, and 12% (87/746) were positive for HSV-2 (Table 1). The prevalence of both infections increased significantly with age; from 5% among 14 to 15 year olds to 12% among 19 year olds for HIV, and from 4 to 22% for HSV-2 (P < 0.001 in each case). Girls who were or had been married (n = 120) had higher HIV (18%) and HSV-2 (42%) prevalence than girls who had never been married [6% HIV; 6% HSV-2; age-adjusted odds ratio (aOR) = 2.6; 95% confidence interval (CI), 1.4–4.9 for HIV, and aOR = 7.7; 95% CI, 4.5–13.3 for HSV-2]. Age-specific prevalence, by marital status, is presented in Fig. 1.
Overall, compared to non-orphans (12%), the prevalence of HIV and/or HSV-2 was higher among girls who had lost one or both parents (17%; aOR = 1.5; 95% CI, 1.0–2.3). The association was significant among maternal orphans (24%; aOR = 2.1; 95% CI, 1.1–4.1), and maternal combined with double orphans (21%; aOR = 1.8; 95% CI, 1.1–2.9) but not for paternal orphans (14%; aOR = 1.3; 95% CI, 0.7–2.2).
There was evidence that the associations between orphan status and both HSV-2 and HIV were modified by marital status (tests for interaction P = 0.006 and P = 0.06, respectively). We therefore stratified the analysis by whether or not participants had ever been married (Table 1). There was no evidence to show that married orphans were at higher risk of HIV or HSV-2 than married non-orphans. Strong associations between orphan status and adverse sexual outcomes were, however, evident among never-married girls. Table 1 shows higher proportions of each sexual health outcome among orphans than non-orphans. For the never-married, further analyses used the combined sexual risk indicator of either HSV-2-positive, HIV-positive or ever pregnant. In comparison with non-orphans, higher levels of sexual risk were seen among maternal orphans, double orphans and, to a lesser extent, among girls who lost their father before age 12 but not after age 12.
Associations were also explored between sexual risk and a parents' recent illness. There was no consistent pattern of risk among girls whose mother was recently ill (n = 70), in comparison with girls whose mother was healthy, and the number of girls with a sick father was small (n = 18). With little evidence of increased sexual risk, girls with a sick father or mother were not distinguished in subsequent analyses.
Initiation of sex and marriage
At the time of the interview, 27% of all participants (n = 234) and 16% of never-married participants (n = 115) reported that they had had sex.
Figure 2 shows person-years in three states: time each girl spent as a virgin; sexually active and not married (single); and married (one of whom claimed she had never been sexually active). These show a sharp rise in sexual activity with age after a mother's death and to a lesser extent, after a father's death, compared to the time while both parents are alive.
To quantify differences illustrated in the plots, relative rates of sexual debut and marriage were compared while a parent was still alive and after a parent's death (Table 2). Girls were significantly more likely to initiate sex after their mother died than while both parents were alive. Similar results were found for girls whose mothers had died when they were children or adolescents. Girls were also more likely to initiate sex after their father died, but only if he died while she was a child and not while an adolescent (P = 0.08 for interaction).
Most sexually-active participants were or had been married, with the exception of girls who lost their mother as an adolescent, most of whom were sexually active but not married at age 19. No significant associations were found between orphanhood or any form of orphanhood and rate of marriage.
Sexual behaviours among orphans and non-orphans
Lifetime sexual experiences of unmarried girls
Specific sexual experiences are presented by orphan status in Table 3. Maternal orphans were the most likely to have had sex (25%), and twice as likely as non-orphans (14%; aOR = 2.0; 95% CI, 1.0–3.8). Most girls had had only one sexual partner in their lifetime (80% of the sexually active unmarried girls), but maternal and double orphans were most likely to have had two or more partners. In other aspects, maternal and double orphans differed: while double orphans were most likely to have a regular sex partner, maternal orphans were the least likely; none of the unmarried maternal orphans were currently with a regular sexual partner.
Half of the unmarried girls who had had sex said they had been forced, at least once, to have sex when they did not want to. In most cases (82%), this included physical force, and 26 girls were physically forced more than once. Experiences of forced sex were high in all groups, ranging from 44% of paternal orphans who had had sex to 65% of maternal orphans, and there was no evidence of a difference with non-orphans.
First sexual experiences of all sexually active participants (married and unmarried)
Circumstances around first sexual intercourse were assessed according to girls' orphan status at the time of first sex (Table 3). Thirty-seven percent of all sexually active girls said their first sex was forced, with double orphans the least likely to have experienced forced first sex. Condoms were used in few first sex acts (23%; n = 48). Paternal and double orphans (31 and 40%, respectively) were more likely than non-orphans (16%) and maternal orphans (14%) to have used a condom during first sex. Double orphans were also the most likely to have used a condom during their most recent sex (47 versus 19% of non-orphans; aOR = 4.1; 95% CI, 1.7–9.8), whereas there were no differences between paternal or maternal and non-orphans.
Experiences within marriage
Of the 103 (14%) girls who identified themselves as ‘currently married’, most described their marriage as traditional (70), three were church marriages, one was a civil marriage, and the remainder were ‘living-in’ or cohabiting (25). Sixteen girls were divorced, and none had been widowed. Seventeen girls got married before they reached the legal age of 16, and 25 married at age 16. The median age difference between married girls and their husbands was 6.5 years; only four girls were the same age or older than their husband.
Maternal orphans were the most likely to marry a man at least 7 years older (67 versus 39% of non-orphans; aOR = 4.6; 95% CI, 0.8–24.7). This was followed by double orphans (53%; aOR = 1.7; 95% CI, 0.6–5.1). Most of the married girls had been pregnant at least once (81%; n = 120); maternal orphans were most likely to say they got married because they were pregnant and least likely for love. The numbers in each of these sub-analyses were small, however, and the differences were not statistically significant.
Associations were found between orphan status and adverse sexual health outcomes among never-married girls in Highfield, Harare: levels of HIV, HSV-2 and pregnancy were higher among maternal and double orphans and girls who lost their father at a young age, in comparison with their non-orphaned peers. Assessment of sexual behaviours helped to explain the heightened risk: rates of initiating sex were higher after the death of a mother (at any age) and a father if the girl was under age 12 at the time of his death. Other differences in sexual experiences emerged among different types of orphans: for example, maternal and double orphans were most likely to have had multiple sexual partners in their lifetime. Whereas maternal orphans were least likely to use a condom at first sex, double orphans were most likely to do so, however; and double orphans were most likely to have a regular sex partner whereas maternal orphans were the least likely.
A study conducted in rural Zimbabwe also found the highest HIV risk among maternal orphans followed by double orphans , but found no association with paternal loss. The study set in KwaZulu Natal, South Africa, did not distinguish risk by type or timing of orphanhood . By looking at HIV/HSV-2 by age at orphanhood, as well as the timing of sexual debut in relation to orphanhood – techniques that have not been applied to this question previously – the Highfield analysis identified a risk among paternal orphans.
The high levels of HSV-2 and HIV infections among married girls in Highfield indicate that girls are at risk within marriage regardless of their orphan status. Whereas maternal orphans in rural Zimbabwe were more likely than non-orphans to be married, this was not the case in Highfield. By stratifying the analysis by girls' marital status, we saw the increased sexual risk for orphans is present among the unmarried, and not the result of early marriage in this context.
The present study, which was designed specifically to explore the links between orphanhood and sexual risk among adolescents, allowed us to draw a representative sample of adolescents, with sufficient numbers to compare outcomes among orphans, by type, with non-orphans. We found a particularly high proportion of double orphans (15 versus 7% in other studies) and maternal orphans (9 compared to 5–7%) [2,8]. This may be due to more recent data collection and maturity of the AIDS epidemic. It may also reflect migration of maternal and double orphans to urban settings such as Highfield, a trend which was evident in the Highfield data (50% of maternal orphans and 38% of double orphans had lived less than 1 year in Highfield, versus 20% of non-orphans and 22% of paternal orphans). The triangulation of reported behaviour and biological markers (both HIV and HSV-2) provided a strong case for sexual risk among young women in this setting. Collaboration between academic and community partners helped to maximize the study's relevance to the community.
A relatively low response rate (67%) may have introduced bias into our sample if we did not capture the girls at highest risk (for fear of HIV/STI testing) or if those in most financial need were more likely to participate (to receive the token of appreciation and bus fare, or in the hope of receiving tangible benefits). There was, however, no statistically significant difference in measures of socio-economic status among girls who did and did not participate, suggesting that sampling bias was not a major problem. There is some evidence that self-reported sexual behaviours underestimated the true prevalence, most likely due to social desirability bias; almost 5% of girls who said they never had sex (n = 29/629) tested positive for HIV or HSV-2. The cross-sectional nature of the study led us to rely on recall for the timing of events, which may have resulted in some misclassification of when events occurred; however, we expect the importance of a parents' death and relative recency of first sex and marriage in the lives of adolescents helped to minimize recall bias. We were not able to differentiate timing of HIV/HSV-2 infection in relation to marriage, but this limitation did not affect our analysis of unmarried girls, where higher associations with orphanhood were found. In some cases, results were based on small sample sizes and meaningful comparisons between groups (e.g., types of orphans) could not be made.
In conclusion, the present study has offered the most detail to date regarding the nature, magnitude and timing of adolescent sexual risk in relation to the loss of a mother and/or father, in a context of high HIV prevalence. The prevalence of adolescent orphans in the Highfield sample at 50% was higher than levels found in earlier Zimbabwean studies [2,8], and may reflect the rising incidence of orphans – particularly maternal and double – detected in recent analyses. Thus, with the highest sexual risk documented among maternal and double orphans, and the numbers of such orphans expected to rise in Zimbabwe and elsewhere, HIV prevention programmes must address the unique needs of orphaned adolescents and appreciate the high sexual health risks to married teens. Following a workshop to discuss study findings with the participants, this project began a pilot programme in Highfield – community-based HIV education combined with bereavement counselling, peer support groups, clothing and food donations – and is assessing its impact locally and relevance more widely. A national workshop organized with UNICEF and the Government of Zimbabwe urged greater emphasis of HIV prevention among orphans in the National Plan of Action on Orphans and Vulnerable Children (OVC), and better synergy of HIV/AIDS and OVC programming.
We thank the Child Protection Society and the community volunteers and young women of Highfield. We are also grateful to BRTI, under the leadership of Peter Mason, and the cooperation of Junior Mutsvangwa and Shungu Munyati. We offer special thanks to Liz Corbett for helping to implement the study, Jimmy Whitworth for procuring laboratory equipment, Yasmin Madin at Populations Services International for the link to New Start Centres, and Mike St Louis and his team at CDC-Zimbabwe.
Sponsorship: This study was funded by UNICEF NYHQ, the DFID Knowledge Programme on HIV/STI at LSHTM, and the Andrew Mellon Foundation. I.B. was funded by a European Union Marie Curie Fellowship and Chadwick Trust Travelling Fellowship, and is currently funded by an ESRC-MRC Fellowship. J.G. is funded by the UK Department of Health (Public Health Career Scientist Award).
Conflict of interest statement: All authors declare that they have no conflict of interest.
© 2008 Lippincott Williams & Wilkins, Inc.