Introduction
Sub-Saharan Africa has been severely affected by the HIV/AIDS epidemic with an estimated 3.1 million new infections, and 2.3 million deaths in 2004 [1]. AIDS is rapidly emerging as the leading cause of adult death [2], and mortality attributable to HIV/AIDS has led to an increase in the number and proportion of children under 15 years who have lost one or both parents. Models suggest that in sub-Saharan Africa, the proportions of children orphaned due to HIV/AIDS will increase to 47.8% in 2010 [3,4]. In Rakai, Uganda, the prevalence of orphanhood in 1990 was 15% [5], and similar levels of orphanhood have been found elsewhere [6-10]. However, these studies did not differentiate between orphanhood associated with parental HIV and unrelated to HIV. One study in Zaire [11] found an incidence of orphanhood of 8.2 per 100 person-years among children of HIV-infected women, and a recent study from Zimbabwe reported an incidence of orphanhood of approximately 6.3 per 100 person-years [12]. However, the rates at which children become orphaned are likely to vary by the stage of the HIV epidemic and levels of adult mortality due to HIV. Model estimates and projections of the number of orphans are based on assumptions about the epidemic dynamics and HIV mortality [3,4], and errors in these projections may lead to inappropriate policies for alleviating the orphan problem [13]. Therefore, we used empirical data to assess the prevalence and incidence of orphanhood and to assess the population attributable fraction of incident orphanhood due to parental HIV infection in Rakai district, Uganda.
Methods
Study population
Data are derived from a community cohort in Rakai District, Uganda [14]. Household censuses were conducted in 56 communities to obtain data on marital status, sex, and age of each resident, and their relationship to the household head. The absence or presence of biological parents of resident children, and the survival status of their parents were also ascertained. Data on household possession of modern objects, housing structure and household size were used as measures of economic status. All consenting resident adults aged 15-59 years who provided informed consent were then interviewed, and asked to provide blood for HIV testing. The study was approved by institutional review bodies in Uganda and the United States.
For the purpose of this analysis, data from 10 657 households included in a census and surveyed in 1996/1997 provided baseline information. A total of 22 712 children aged 0-14 years were enumerated during the 1996/1997 census and provided cross-sectional data for estimating the prevalence of orphanhood. Children were classified as orphans if they had lost at least one parent or as non-orphans if both parents were still alive, irrespective of whether they were currently co-resident with the child. Orphaned children were further categorized into maternal orphans (children who had only lost a mother); paternal orphans (those who had only lost a father) and double orphans if they had lost both parents. For the cross-sectional analysis, parental HIV status could not be determined for all deceased parents, so analysis of the prevalence of orphanhood by parental HIV status was done for a subgroup of children with at least one parent of known HIV status, irrespective of whether this parent was alive or dead.
Parental HIV infection was determined from blood collected after the interviews. Serum was tested for HIV-1 using two enzyme immunoassays (EIA; Vironostika HIV-1; Organon Teknika, Charlottesville, North Carolina, USA and Cambridge Biotech, Wooster, Massachusetts, USA) with western blot (HIV-1, Western Blot; BioMerieux Vitek, St Louis, Missouri, USA) confirmation of discordant EIA results and HIV seroconverters. Parents who were interviewed but not bled at the 1996/1997 survey were classified as HIV seronegative if they tested negative at the subsequent study visits and seropositive if they had tested positive at earlier study visits. Parental HIV infection was stratified into three categories: (1) at least one HIV-infected parent, (2) unknown parental HIV status, and (3) both parents known to be HIV negative (used as the referent category). The unknown parental HIV status category included children for whom both parents' HIV infection status was unknown, or one parent had unknown HIV status and the other parent was HIV negative. From other data we estimate that if one parent is HIV negative, there is 88% likelihood that the other parent is also uninfected [15,16].
The households were followed up 10 months later in 1997/1998 and data were collected on parental survivorship (for both resident and non-resident parents), and on migration/mobility for all household members. A total of 4953 children (21.8%) were already orphaned at the 1996/1997 census, and among the 17 759 non-orphans identified in 1996/1997, 2478 children (14.0%) had out migrated, 272 (1.5%) had died at the end of the 10 months inter-survey period, and 38 children (0.2%) had missing data on outcome variables. These children were excluded from the incidence analysis since their orphanhood status could not be established. The incidence of orphanhood during the 10 months inter-survey period was determined in a total of 14 971 surviving children, who had both parents living at the initial 1996/1997 census. An incident orphan was defined as any child less than 15 years of age at the 1996/1997 census who lost one or both parents during the inter-survey period. Using unique identifiers, we were able to link household information on the socio-demographic characteristics of the mother, and HIV status of the mother and father to children with co-resident parents who participated in the survey. The mother's characteristics including age, education, and main occupation were ascertained by interview, and household characteristics including possessions of modern objects (radio and bicycle), housing structure (roofing, floor, and walls), and household size (number of rooms) were ascertained from the census information.
Statistical analysis
The prevalence of orphanhood was estimated as the proportion of children with one or both parents deceased at the 1996/1997 census. The incidence of orphanhood was estimated as the number of children who lost at least one parent during the 10 month follow-up, divided by the total number of children with both parents living at the initial 1996/1997 census.
We used a generalized linear model (GLM) with a Poisson distribution to estimate crude and adjusted incidence rate ratios (IRR) and 95% confidence intervals (95% CI) of becoming orphaned over the 10 months follow up. Covariates included the child's age and gender, parental HIV infection, mother's socio-demographic characteristics, household characteristics and rural/urban residence. In univariate analyses, we estimated the crude incidence rate ratios of orphanhood by these covariates, and in multivariate analyses, we adjusted for potential confounders and checked for interactions. All covariates that were significant at P < 0.15 in the univariate analysis, or those having a univariate IRR greater than 2.0 or less than 0.5, and suspected confounders were included in the multivariate analyses. The statistical significance of individual regression variables was determined by the Wald test or the log likelihood ratio test (LRT) [17,18]. Since some children came from the same household, there might be correlation between individual children at the household level. Therefore, we used the Huber-White sandwich robust estimator to account for non-independence, and the 95% CI are based on robust standard errors [19-21].
We estimated the population attributable fraction (PAF) of incident orphanhood associated with parental HIV infection. The PAF was estimated from newly orphaned children with at least one HIV-infected parent as a proportion of the total newly orphaned children (Pi), and the adjusted incidence rate ratio (IRR) of becoming orphaned among children with at least one HIV-infected parent, relative to children whose parents were both uninfected, [PAF% = ΣPi × ((IRRi - 1)/[(IRRi - 1) + 1]) × 100], where Pi is the proportion of incident cases in the ith exposure level, and IRRi is the adjusted incidence rate ratio comparing the ith exposure level relative to the unexposed group. The 95% confidence intervals (95%CI) of the PAF were estimated using the aflogit Stata ado program [22-24]. The duration of orphanhood (D) was estimated from the ratio of prevalence (P) to annualized incidence (I), based on the formula P = I × D (i.e., D = P/I). The under-five (U5) mortality rates were estimated as the number of deaths in children aged 0-4 years, during the inter-survey period, divided by their total number of person years of observation. Person years of observation were obtained as time between the two censuses, and ascribing half the time for those who died between baseline and follow up censuses, on the assumption that deaths were evenly distributed throughout the follow up interval. Statistical analyses used STATA software package version 7.0 (College Station, Texas, USA).
Results
Prevalence of orphanhood
The adult (15-49 years) HIV prevalence was 14.8% and was significantly higher among females (17.3%) than males (11.7%, P < 0.0001). Table 1 shows the number of children and the prevalence of orphanhood among the 22 712 children aged 0-14 years identified at the 1996/1997 census. The overall prevalence of orphanhood was 21.8%, and significantly increased with the child's age from 7.3% among children aged 0-4 years to 40% among the 10-14 year-olds (χ2 for linear trend P <0.0001). Paternal loss was the most common cause of orphanhood (11.2%), followed by double orphanhood (6.4%), and maternal orphanhood (4.3%). Among the 0-4 year-olds, the prevalence of maternal and double orphanhood were low (1.4 and 1.1%, respectively), but paternal orphanhood was significantly higher (4.8%, P < 0.001). Among the 2540 paternal orphans, 927 (36.5%) had a non-resident surviving mother, and among the 967 maternal orphans, 549 (56.8%) had a non-resident surviving father. Thus, even children who had lost a single-parent were likely not to be co-resident with the surviving parent, particularly if they had lost a mother.
Among the 22 712 children, 2997 (13.2%) had at least one HIV-infected parent. The proportion of children with a known HIV-infected parent was highest among the 0-4 year olds (15.5%, 1523 of 9840), and declined with the child's increasing age to 9.4% in the 10-14 year olds (539 of 5761) (χ2 for linear trend, P < 0.001). If one or both parents were HIV-infected, the prevalence of orphanhood was 22.7% (681 of 2997), compared with a prevalence of 7.9% (931 of 11 815) if the parents were uninfected (P < 0.001).
Incidence of orphanhood
There was a total of 14 971 children followed up between 1996-1997 and 1997-1998 who provided data for the assessment of the incidence of orphanhood. Of these children, 5590 had parents who were both HIV negative (37.3%), 1877 had at least one HIV-infected parent (12.5%), and 7504 had parents of unknown HIV status (50.1%). Table 2 shows the incidence rate, and the crude and adjusted incidence rate ratios of orphanhood. Three hundred and twenty-three (323) children lost at least one parent during follow up and the incidence of orphanhood was 2.2% over 10 months (i.e., 2.6% per year). The incidence of orphanhood was 6.8% over the 10-month follow up if at least one parent was HIV-infected (annual incidence 8.2%), compared with 0.4%/10 months (annual incidence 0.5%) if both parents were HIV uninfected (P < 0.001). Increasing child and maternal age, parental HIV infection, non-residence of both parents or the presence of already orphaned or foster children in the household were significantly associated with an increased risk of incident orphanhood.
In the multivariate analysis, the adjusted IRR of orphanhood associated with parental HIV infection was 18.93; (95% CI, 8.34-42.99), and the adj. IRR was 4.77; (95% CI, 2.04-11.15) among children with unknown parental HIV status, compared to the referent group of children with HIV-negative parents. The adjusted risk of incident orphanhood was significantly increased among children with older mothers (adj. IRR = 3.11; 95% CI, 1.34-7.22 for mothers aged 45 or more), and children in households already caring for orphans or foster children (adj. IRR = 1.57; 95% CI, 1.06-2.33). Furthermore, older age of the child was significantly associated with increased risk of incident orphanhood (adj. IRR = 1.32; 95% CI, 1.03-1.69 for 5-9 year olds; adj. IRR = 1.64; 95% CI, 1.20-2.25 for the 10-14 year olds). Household size, affluence, and rural/urban residence were not significantly associated with the incidence of orphanhood.
The duration of orphanhood was estimated from the ratio of prevalence to annualized incidence rates (prevalence/annual incidence). The overall estimated mean duration of orphanhood was approximately 8.3 years (21.8/2.6). Among children with HIV-positive parents, the mean duration of orphanhood was approximately 3.4 years (22.7/6.8).
Population-attributable fraction of incident orphanhood due to parental HIV infection
Table 3 shows the population attributable fraction (PAF) of incident orphanhood associated with parental HIV. The overall PAF of incident orphanhood due to parental HIV infection was 37.3% (95% CI, 27.7-45.7), and was highest among younger children (adj. PAF = 50.6% for 0-4 year olds; adj. PAF = 37.3% for 5-9 year olds and adj. PAF = 22.1% for children aged 10-14 years). The PAF of incident orphanhood associated with parental HIV infection was particularly high among children with younger mothers (adj. PAF = 75.7% for mothers aged < 25 years; adj. PAF = 59.3% for mothers aged 25-34 years; and adj. PAF = 32.9% for mothers aged 35-44 years).
Incidence of double orphanhood
The incidence of double orphanhood was assessed among 2864 single-parent orphans identified in 1996-1997. There were 59 children who lost a second parent during follow up and the incidence of double orphanhood was 2.1% over 10 months (2.5% per year). The 10-month incidence of double orphanhood among single-orphans with an HIV-infected parent was 4.1% over 10 months (22 of 536, annual incidence 4.9%), compared with an incidence of 1.6% (37 of 2328, annual incidence 1.9%) among children with HIV-negative parents (adj. IRR = 7.19; 95% CI, 2.37-21.82). The PAF of double orphanhood associated with parental HIV-infection was 41.9%.
Child mortality and orphanhood
Mortality among children aged 0-4 years was estimated per 100 person years). The mortality rate among 6744 non-orphaned children with co-resident parents was 3.1/100 person-years (197/6370 person-years), compared with 4.8/100 person-years (11/227 person-years) in children who had lost a mother (adj. IRR = 5.58; 95% CI, 2.57-11.67), and 2.1/100 person-years (9/427) in children who had lost a father alone (adj. IRR = 1.09; 95%CI, 0.56-2.13).
Discussion
This prospective study provides empirical data on the incidence of orphanhood, and the PAF of incident orphanhood due to parental HIV infection. The effects of HIV infection on orphanhood were profound. The incidence of orphanhood was 8.2% per year if one or both parents were HIV-positive, compared with 0.5% if the parents were HIV-negative. The annual incidence of orphanhood among children with HIV-infected parents in Rakai, is comparable to estimates from a smaller study of children of 466 HIV-infected women in Zaire [11], but is somewhat higher than the incidence of 6.3 per 100 person-years in a population-based cohort in Zimbabwe [12]. The data are also compatible with the annual mortality of HIV-infected adults in the Rakai cohort population [25,26] as well as model projections [4].
The PAF of incident orphanhood attributable to parental HIV was 37.3% overall, and was especially high among children aged 0-4 years (adj. PAF = 50.5%) and children of mothers aged under 25 years (PAF 75.7%). These findings suggest that if parental mortality due to HIV/AIDS could be eliminated, there would be a 37.3% reduction in the incidence of orphanhood in this population, assuming that the association between parental HIV and orphanhood was causal [22].
The prevalence of orphanhood was 22.7% if one or both parents were known to be HIV-infected in comparison with 7.9% if the parents were HIV-uninfected. The overall prevalence of orphanhood in 1996/1997 was higher than the previous 15% prevalence estimated for Rakai data in 1990 [5], suggesting an increase in mortality among HIV-infected adults due to disease progression in the absence of antiretroviral therapy [25,26]. The prevalence of orphanhood in Rakai is also higher than national Demographic and Health Survey (DHS) estimates for Uganda in 1995 and 2000 (12.9 and 12.4%, respectively) [27], and HIV prevalence in Rakai (14.8%) was higher than the national average (7%).
Loss of both parents (6.4%) was more frequent than maternal orphanhood (4.3%), which is contrary to some [5,8,10], but not all previous studies [28]. Moreover, even among children who had lost a single parent, a high proportion had no co-resident surviving parent, reflecting the vulnerability of these orphaned children. In addition, the presence of resident orphans or foster children in the household was significantly associated with the incidence of new orphanhood (Table 2), suggesting that families which absorb new orphans or foster children may already be carrying a substantial burden of childcare [3]. Orphaned children clearly constitute a vulnerable population. In our study, children who had lost their mothers were at increased risk of death, and this is comparable to findings from Zimbabwe [12]. Other surveys indicate that orphans are less likely to attend school [29].
Data from Rakai [26] and Zimbabwe [12], suggest a progressive increase in the prevalence of orphanhood over time, due to increasing HIV-associated adult mortality. It is, therefore, likely that the prevalence and incidence of orphanhood due to HIV will increase in the future unless sustainable antiretroviral therapy can reduce adult AIDS mortality. Furthermore, programs for the prevention of mother-to-child HIV transmission may reduce the rate of HIV infection at birth and thus increase child survival, which in turn will increase the population of surviving children at risk of orphanhood.
This study has limitations. Parental HIV status was unknown for 7504 (50.1%) children included in the incidence analysis, because the parents either declined to provide a blood sample for HIV testing or they were absent at the time of the 1996/1997 Survey (Table 2). However, the proportion of children with missing parental HIV status was lower in Rakai, than in a study from Zimbabwe (61%) [12]. A proportion of these parents with unknown HIV status were likely to be HIV infected, and this would lead to misclassification, which may result in underestimation of incident orphanhood among children with known infected parents. The children of parents with unknown HIV status were included in the analysis to provide a comprehensive estimate of the PAF of orphanhood associated with parental HIV-infection in this population. Children with HIV-infected mothers have high mortality because of mother-to-child HIV transmission [30], and the risk of orphanhood among HIV-infected children is truncated by the competing risk of child death, which might lead to underestimation of parental loss. However, the study has several advantages. It was population-based, and should capture the majority of orphaned children even if they are dispersed among different households or if the initial household was dissolved as a consequence of parental death. Furthermore, Rakai is representative of a mature, generalized HIV epidemic in rural Africa, and thus allows generalization to other comparable populations.
In conclusion, the prevalence and incidence of orphanhood among children of HIV-infected parents is high in this population. There is an urgent need to provide support for these vulnerable children, and to reduce adult HIV infection and mortality to prevent orphanhood.
Contributors
F.E.M. was the lead author responsible for the data management, statistical analyses, interpretation and manuscript preparation. R.H.G. contributed to the analyses, interpretation and manuscript preparation. R.H.G., N.K.S., M.J.W., D.S. and F.W.-M. were responsible for the design and implementation of the original trial, and manuscript preparation. M.K. and F.N. contributed to the review of the manuscript.
Acknowledgments
We thank the former UVRI Director, Dr S. Sempala and the people of Rakai who provided these data.
Sponsorship: This study is part of the main author's PhD dissertation supported by Fogarty International Center of the National Institutes of Health grant (numbers 1D43-TWO 150801 and 3D43TW-00010), and the Gates Institute for Population and Reproductive Health at Johns Hopkins University, Bloomberg School of Public Health. The research was supported by grants RO1 A134826 and RO1 A134826S, National Institutes of Allergy and Infectious Diseases; the Rockefeller Foundation; the World Bank Uganda Sexually Transmitted Infection (STI) Project, and the Johns Snow Inc. grants 5024-30.
References
2. Gregson S, Garnett GP, Anderson RM. Is HIV-1 likely to become a leading cause of adult mortality in sub-Saharan Africa? J Acquir Immune Defic Syndr 1994; 7:839-852.
3. UNAIDS/UNICEF.
2003 Fact Sheet: Children Orphaned by AIDS in sub-Saharan Africa. Fact Sheet - 21/10/2003 - en UNAIDS, UNICEF Fact sheet, produced jointly by UNAIDS and UNICEF on the topic of children orphaned due to AIDS in sub-Saharan Africa.
http://www.unaids.org/en/.
5. Nalugoda F, Wawer MJ, Konde-Lule KJ, Menon R, Gray RH, Serwadda D, et al. HIV infection in rural households, Rakai District, Uganda. Health Trans Rev 1997; 7(Suppl 2):127-140.
6. Preble EA. Impact of HIV/AIDS on African children. Soc Sci Med 1990; 31:681-690.
7. Hunter SS. Orphans as a window on the AIDS epidemic in sub-Saharan Africa: initial results and implications of a study in Uganda. Soc Sci Med 1990; 31:681-690.
8. Foster G, Shakespeare R, Chinemana F, Jackson H, Gregson S, Marange C, et al. Orphan prevalence and extended family care in a peri-urban community in Zimbabwe. AIDS Care 1995; 7:3-17.
9. Foster G, Makufa C, Drew R, Kambeu S, Saurombe K. Supporting children in need through a community-based orphan visiting Program. AIDS Care 1996; 8:389-403.
10. Kamali A, Seeley JA, Nunn AJ, Kengeya-Kayondo JF, Ruberantwari A, Mulder DW. The orphan problem: experience of a sub-Saharan Africa rural population in the AIDS epidemic. AIDS Care 1996; 8:509-515.
11. Ryder RW, Kamenga M, Nkusu M, Batter V, Heyward WL. AIDS orphans in Kinshasa, Zaire: incidence and socioeconomic consequences. AIDS 1994; 8:673-679.
12. Watts H, Lopman B, Nyamukapa C, Gregson S. Rising incidence and prevalence of orphanhood in Manicaland, Zimbabwe, 1998 to 2003. AIDS 2005; 19:717-725.
14. Wawer MJ, Gray RH, Sewankambo NK, Serwadda D, Paxton L, Berkley S, et al. A randomized, community trial of intensive sexually transmitted disease control for AIDS prevention, Rakai, Uganda. AIDS 1998; 12:1211-1225.
15. Porter LE. HIV Status and Union Dissolution in Rakai, Uganda [Dissertation]. Baltimore, Maryland: The Johns Hopkins University; 2002.
16. Serwadda D, Gray RH, Wawer MJ, Stallings RY, Sewankambo NK, Konde-Lule JK, et al. The social dynamics of HIV transmission as reflected through discordant couples in rural Uganda. AIDS 1995; 9:745-750.
17. Collett D. Modelling Survival data in Medical Research. Texts in Statistical Science, 1st edn. London, New York: Chapman & Hall/CRC;1994.
18. Breslow NE, Day NE. Statistical Methods in Cancer Research, Vol. II The Design and Analysis of Cohort Studies. International Agency for Research on Cancer (IARC) Scientific Publications No. 82. Oxford: Oxford University Press.
19. Rogers, W. H. 1993. Regression Standard Errors in Clustered Samples. STATA Technical Bulletin 13:19-23. Reprinted in STATA Technical Bulletin Reprints, vol. 3, 88-94. College Station, Texas: STATA Corporation.
20. Williams RL. A note on robust variance estimation for cluster-correlated data. Biometrics 2000; 56:645-646.
21. Wooldridge JM. Econometric Analysis of Cross Section and Panel Data. Cambridge, Massachusetts: The MIT Press; 2002.
22. Rockhill B, Newman B, Weinberg C. Use and misuse of population attributable fractions. Am J Public Health 1998; 88:15-19.
23. Bruzzi P, Green SB, Byar DP, Brinton LA, Schairer C. Estimating the population attributable risk for multiple risk factors using case-control data. Am J Epidemiol 1985; 122:904-914.
24. Hamilton LC. STATA 7. Statistics with Stata, Updated for Version 7. College Station, Texas: STATA Corporation.
25. Sewankambo NK, Wawer MJ, Gray RH, Serwadda D, Li C, Stallings RY, et al. Demographic impact of HIV infection in rural Rakai district, Uganda: results of a population-based cohort study. AIDS 1994; 8:1707-1713.
26. Sewankambo NK, Gray RH, Ahmad S, Serwadda D, Wabwire-Mangen F, Nalugoda F, et al. Mortality associated with HIV infection in rural Rakai District, Uganda. AIDS 2000; 14:2391-2400.
27. Uganda Bureau of Statistics and ORC Macro. Uganda Demographic and Health Survey, 1995 and 2000-2001. Calverton, Maryland: Uganda Bureau of Statistics and ORC Macro.
28. Bicego G, Rutstein S, Johnson K. Dimensions of the emerging orphan crisis in sub-Saharan Africa. Soc Sci Med 2003; 56:1235-1247.
29. Case A, Paxson C, Abeidinger J. Orphans in Africa: parental death, poverty and school enrolment. Demography 2004; 41:483-508.
30. Beckerman KP. Mothers, orphans, and prevention of paediatric AIDS. Lancet 2002; 359(9313):1168-1169.
© 2005 Lippincott Williams & Wilkins, Inc.