Introduction

In all societies, the death of a parent during childhood and adolescence, especially the death of a mother, is widely acknowledged as having potentially profound consequences on child health and wellbeing [1–8]. Studies in South Africa conducted since the start of the HIV epidemic have shown orphaned children to be at increased risk of poorer health and educational outcomes compared to other children (See for example, [9–11]). The severity and duration of negative consequences are influenced by the type, cause and timing of orphaning. For example, there is consistent evidence that double or maternal orphans are at greater risk than paternal orphans [2,12,13]. Compared to other causes of parental death, children experiencing the death of a parent due to AIDS or injuries have higher risks of poorer child mental health [14] and, significant changes in their living arrangements due to household dissolution and migration [14–16].

In rural KwaZulu-Natal, South Africa, men and women become parents in a context of very high HIV population prevalence [17,18]. In the early 2000s, forecasts based on population data suggested that even if HIV incidence remained stable, in the absence of public HIV treatment, between 24% and 33% of children would go on to become maternal and paternal orphans by their 18^th birthday [19,20]. When antiretroviral treatment (ART) was rolled out from 2004 onwards, overall adult mortality rates declined reflecting reduced adult mortality due to HIV/AIDS [21–23]. Consequently, the previously forecasted level of orphan incidence are no longer expected even in communities that experienced severe HIV epidemics and where the prevalence of HIV is still increasing due to the improved survival of people living with HIV. However, there are very few recent and detailed published population orphanhood estimates for South Africa, especially of orphan incidence rates. Given that orphans and vulnerable children (OVC) are a key population group targeted by statutory and private support services [24–27]; uncertainty about recent changes in the numbers and types of orphans in HIV affected communities is a challenge for planning relevant and effective health and welfare-related policy responses.

This paper contributes to filling this gap with robust estimates of the levels and trends in age-specific orphan prevalence and incidence in rural KwaZulu-Natal over the period 2000–2014. We are able to situate these new estimates of orphanhood in the context of changes in the level and pattern of adult mortality documented by previously published studies from the same area [9,15,19,21,28–31]. We describe changes over the 15 year period (2000–2014) in the ages at which children are orphaned, different types of orphans, children’s exposure to different causes of parental mortality. We discuss the implications of these recent trends in orphaning for programme and research activities focusing on orphans and other children affected by HIV in South Africa.

Methods and data

Results

The levels and trends in age-specific maternal, paternal and double orphan prevalence are presented in Fig 1 for the years 2000, 2005, 2010 and 2014 (S1 Table presents additional detailed data related to Fig 1 including absolute numbers, estimates and 95% confidence intervals). The proportion of children and adolescents (<20 years) for whom one or both parents had died rose from 26% in 2000 to peak at 36% in 2010, followed by a decline to 32% in 2014. In all age groups and years, the level of paternal orphan prevalence was three-five times higher than maternal orphan prevalence. The pattern of maternal orphan prevalence rose from 2000 to plateau between 2005 and 2010, declining slightly by 2014. In contrast, paternal orphan prevalence continued to rise through 2010—declining later and more slowly than maternal orphan prevalence. The largest difference between paternal and maternal orphan prevalence was in the youngest age group (0–4 years). In this age group, the seven-fold difference in 2000, 15% and 2% respectively, remained very similar across the period. Orphaning is a cumulative experience with age, with the oldest age group of children (15–19 years) experiencing the highest prevalence of every orphan type. By 2010, 53% of children aged 15–19 years had experienced the death of one or both parents; with 10% in this age group being maternal orphans and 27% paternal orphans. Increases in double orphaning among older age groups of children were very pronounced through 2010; more than doubling from 7.5% (2000) to 16% (2010) in 15–19 year olds. By the end of the study period in 2014, the prevalence of all orphan types had declined close to, but had not yet reached, the lower levels observed at the start of data collection in 2000.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Orphanhood prevalence by orphan status and age group in children and adolescents <20 years in 2000, 2005, 2010, 2014.

https://doi.org/10.1371/journal.pone.0238563.g001

Fig 2 presents age-specific incidence of maternal and paternal orphaning (S2 Table presents additional detailed data related to Fig 2 including numbers and 95% confidence intervals). Estimates of maternal and paternal orphaning incidence (age <20 years) peaked in 2005, 17 and 27 per 1,000 person years respectively. As with orphan prevalence, the difference between maternal orphaning incidence and the higher rates of paternal orphaning incidence is greatest in the youngest children (0–4 years). Across the period, the two oldest age groups (10–14 and 15–19 years) experience the highest incidence of maternal and paternal orphaning. The highest level of paternal orphaning incidence estimated in these age groups was in 2005, 30 and 33 per 1,000 person years respectively. Compared with orphan prevalence, orphaning incidence declined earlier and more rapidly; by 2014, the incidence of all types of orphan was less than half those in 2000.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Maternal and paternal orphaning incidence (per 1,000 PYO) by orphan status and age group in children and adolescents <20 years, 2000, 2005, 2010, 2014.

https://doi.org/10.1371/journal.pone.0238563.g002

Causes of parental death

Fig 3A–3D show the trends in annual cause-specific mortality in mothers and fathers by age. HIV/AIDS and TB was the dominant cause of parental mortality throughout the period. Substantial reductions in the rates of mothers and fathers dying due to HIV/AIDS and TB occurred after 2007 and 2009 respectively. The cause-specific mortality of older mothers (35+ years at death) closely resembles younger mothers (<35 years at death) with HIV/AIDS and TB remaining the largest category of mortality across the period 2000–2014. In contrast, at the beginning of the period and most years from 2008, a larger proportion of deaths in older fathers (40+ years at death) were attributable to non-communicable diseases and injuries than HIV/AIDS and TB. For younger fathers (<40 years at death), the rate of deaths due to HIV/AIDS and TB only fell below those due to non-communicable diseases and injuries several years later, from 2012.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Mortality rates per 10,000 PYO by year (2000–2014), overall and by cause of death, among deceased mothers (Fig 3A aged <35 years and Fig 3B aged 35+ years) and deceased fathers (Fig 3C aged <40 years and Fig 3D aged 40+ years) of children and adolescents <20 years.

a. Person years observed (PYO) 2000–2014 = 306,729; Number of maternal deaths <35 years 2000–2014 = 3,896. b. Person years observed (PYO) 2000–2014 = 132,477; Number of maternal deaths 35+ years 2000–2014 = 1,710. c. Person years observed (PYO) 2000–2014 = 89,928; Number of paternal deaths <40 years 2000–2014 = 1,547. d. Person years observed (PYO) 2000–2014 = 139,041; Number of paternal deaths 40+ years 2000–2014 = 3,659.

https://doi.org/10.1371/journal.pone.0238563.g003

Table 1 shows the estimated incidence of maternal and paternal orphaning overall and by separate age groups for each cause of death in 2000, 2005, 2010 and 2014. The incidence of maternal and paternal orphaning due to HIV/AIDS and TB reduced by more than half over the period in all age groups of children. The exception was the slightly smaller decline in the incidence rate of paternal orphaning due to HIV/AIDS and TB in the oldest age group (15–19 years). In 2000, at the beginning of the period the age-specific rate of orphaning was much higher in this age group, 14.5 per 1,000 person years, than in younger age groups. Rates of paternal orphan incidence due to injury-related deaths were more than four-fold higher than maternal orphan incidence. The overall incidence of each cause of parental death declined across the period. In contrast, the incidence of orphaning due to communicable diseases and undetermined causes increased slightly; more so with respect to deaths of mothers than fathers.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Cause-specific maternal and paternal orphaning incidence rates per 1000 PYO in children and adolescents <20 years by age group in 2000, 2005, 2010, 2014.

https://doi.org/10.1371/journal.pone.0238563.t001

Discussion

Our analyses of long-term trends over the pre- and post-ART periods (2000–2014) show recent declines in maternal and paternal orphan prevalence and incidence. Having already reached high levels by the start of the period (2000), the proportions of children orphaned continued to rise further, only starting to decline after 2010. Despite huge progress in scaling up the HIV treatment programme since the mid-2000s, these longitudinal population-based data provide evidence that the experience of a parents during childhood or adolescents remained very common in this rural South African community. In 2014, a third (32%) of all children and adolescents (0–19 years) and half (51%) of the older adolescents (15–19 years) had experienced the death of one or both parents respectively. As expected given the broad similarity of social and economic characteristics and public services in the study area to rural communities elsewhere in rural KwaZulu-Natal; these estimates of orphanhood mirror provincial those from national Census and survey data [9,28,48]. However, compared with other South African provinces, KwaZulu-Natal has higher levels of orphanhood. In 2011, the province had more than double the number of single and double orphans (<18 years) then the next highest province—Eastern Cape based on the 2011 Census data [49].

The peak levels of paternal and double orphan prevalence (<20 years) in the study population 2010 (S1 Table) are much higher than the national estimates in the 2011 Census: 21% vs 15% and 7% vs 4% respectively [49]. Whereas estimates of maternal orphan prevalence (<20 years) was the same in the study and nationally (7%). In KwaZulu-Natal, exceptionally high levels of excess male mortality in young adults have been documented at different stages of the HIV epidemic. The causes of excess male mortality particularly in young men, is not only due to HIV/AIDS and TB deaths but also higher rates of road traffic accidents, intentional injuries and non-communicable diseases [20,50,51].

A strength of the directly observed data available from longitudinal population surveillance is that it enables us to estimate and monitor trends in orphan incidence, a much rarer indicator. The levels of both orphan incidence and prevalence that we report for the early period before public ART rollout are very similar to the estimates and forecasts of orphanhood reported on the basis of studies conducted in the early 2000s in the same population [19] and, in KwaZulu-Natal [52,53]. However, after ART rollout, our study shows that there was an immediate decline in orphan incidence but not orphan prevalence. The trend in orphan incidence closely mirrors the timing of reversals in adult mortality reported in the study population [21,23]. In contrast, orphan prevalence is a cumulative indicator and therefore, is less responsive than orphan incidence to immediate impacts of HIV treatment or other factors that improve parental survival. As discussed by Reniers et al (2017), recent declines in adult mortality due to HIV/AIDS is not only a consequence of ART roll-out but of long-term declines in HIV incidence and improvements in coverage and effectiveness of HIV treatment and care programmes [23]. In the study population, estimated HIV incidence in women and men declined year on year from 2014 and 2012 respectively [38].

Understanding the trends in survival of HIV-infected parents in the period post-ART rollout is complicated by changes in treatment access and, testing and treatment guidelines for ART eligibility over the same period. After national roll-out in 2004, local ART services were initially available from the hospital clinic [40]. Later ART services rolled out to all sixteen clinics in the health district offering ART initiation and follow-up by late 2007. Until March 2010, treatment guidelines were for ART initiation of adults with poor levels of immunity and health (CD4 cell count < 200 cells/μl or WHO stage 4) [54]. Eligibility criteria were first increased to CD4 <350 cells/μl for pregnant women and people with active tuberculosis (TB) disease, then later for all HIV-infected adults. After the end of study follow-up for this paper, eligibility criteria for ART initiation rose to CD4<500 cells/μl in January 2015 and irrespective of CD4 count from September 2016 [42]. Despite increased availability of ART, the estimated population ART coverage in HIV-positive adults in this area only increased from 31.7% in 2012 to 45% in 2017 [38,42].

Several factors may contribute to the consistently higher proportion and incidence of death among fathers compared with mothers. Mortality rates increase with age and irrespective of the effects of the HIV epidemic, fathers are on average older than mothers. Among married and regular couples belonging to the same household in the study area, men are more than 6 years older than their female partner on average [55]. There are also reasons to expect an earlier and larger impact of ART treatment on the survival of HIV-infected mothers than fathers in this population and others in South Africa. The Prevention of Mother-to-Child Transmission (PMTCT) national programme started in 2002 i.e. before the national rollout of ARV started [56]. Similar to other areas of South Africa, pregnant and recently delivered mothers have very high rates of contact with antenatal and delivery and child vaccination services [57,58]. During the study period therefore, not only were HIV-infected mothers more likely than HIV-infected fathers to have been tested for HIV, there would have been earlier improvements in the average time from HIV infection to ART eligibility and survival for women on ART treatment. A prospective study of a large cohort of HIV-infected pregnant women in the study area (2010–2014) showed that the majority of women (96%) were either on lifelong ART or ART prophylaxis within 6 months of the first antenatal visit [57]. Men’s poorer levels of engagement with healthcare generally, and specifically with HIV care, has been well documented in the study area and KwaZulu-Natal [22,42,59]. Compared with women, HIV-infected men are more likely to have a late HIV diagnosis, delay in seeking healthcare, and not remain engaged [60].

In the period before ART rollout, increasing trends in adult mortality in HIV-infected and HIV-uninfected adults and, as a result in maternal, paternal and double orphan prevalence and incidence, were reported in rural study populations in Malawi and Tanzania (1998–2004) [19] and Zimbabwe (1998–2003) [61] where comparable demographic and HIV surveillance was undertaken. The levels of all orphanhood indicators reported in the South African study population were however, much higher than those in the other rural study populations. In the years after the implementation of national HIV treatment programmes across the region, we expect more variation to have emerged in the pattern and timing of reversals in orphanhood due to differences in the underlying HIV prevalence between populations, the timing of ART rollout and variations in the extent of ART coverage [62,63]. Few comparable data in other African communities are available on changes in orphan incidence and prevalence trends before and after ART rollout. A similar pattern of decline in orphanhood in the Rakai study population in Uganda following HIV treatment was reported [64]. The prevalence and incidence of children (0–14 years) with one or both parents deceased declined significantly pre-HIV care period (2001–2003) levels of 17.2% and 2.10 ⁄ 100 person years respectively to 12.6% and 1.07 ⁄ 100 person years after HIV care was expanded (2006–2009).

The demographic surveillance data analysed in this study are of high quality and provide very detailed information on all children and adults belonging to households in this rural community over a 15-year period. There are two particularly useful features of ACDIS data for orphanhood studies: i) the deliberate distinction of biological mothers and fathers from non-biological parents and from non-parental primary carers; and, ii) the follow-up of biological parents who are not resident with their child in the study area, for example, as labour migrants [65,66]. Thus, our estimates of orphanhood are reliable and robust, as well as generalizable to other communities in KwaZulu-Natal with similar demographic and health characteristics. Some limitations remain. Although efforts were made to maximize the completeness and ensure the quality of repeated rounds of parental survival data [65,66], ‘missing’ or erroneous information about parental survival may result in under- or overestimations of orphanhood in this population. An ‘adoption effect’ in which interviewers or respondents report information about non-biological rather than biological parents has been described elsewhere [65]. In South Africa, Udjo (2011) has also suggested that African households are more likely to over-report paternal deaths than other population groups [67]. In this study however, the effects of response bias on orphanhood estimates are likely to be small given the numerous visits to households to update ACDIS information and, specific attention in questionnaire design and interviewer training. Orphanhood estimates could also be biased by parents, particularly fathers, whose survival status was not known for the whole period of follow-up [68]. However, although we expect rates of mortality to be higher in parents of unknown survival than in parents with complete data, parental survival status was unknown for only 0–2% of children at each mid-year time point. We therefore, grouped children with unknown parental survival status together with children known to be orphaned. This approach is similar to that used by UNICEF and other agencies in defining children with one deceased parent as an orphan irrespective of the survival status of the other parent [69]. This study has implications for future research on children and parents in communities with high HIV prevalence. Our results, together with the findings of other recent longitudinal studies of HIV and adult mortality [22,23] suggest that we can anticipate further gains in life-expectancy of HIV-infected adults and a continued decline in HIV/AIDS as a cause of premature adult death. Given the scale of the HIV epidemic in rural KwaZulu-Natal, there is also likely to be an increase in the number of children living with HIV-infected parents and carers for some years to come. However, the multiple determinants of overall and age-specific fertility and mortality means less certainty in the future levels of different orphanhood types [7].

The value of continuing to collect and measure orphanhood in longitudinal African population cohorts is also given emphasis by recent studies highlighting important discrepancies between the international modelled orphanhood prevalence estimates reported by the United National Programme on HIV/AIDS (UNAIDS) and similar estimates derived from survey and census data [70]. There is no current consensus about how changes in the HIV epidemic such as age and sex patterns of HIV incidence and changes in HIV treatment coverage are affecting differences in orphanhood estimates or which statistical methods are appropriate for adjusting the modelled estimates [71].

The findings of this population-based study have implications for South Africa’s formal statutory policies and social assistance programmes targeted towards orphans such as the Foster Care Grant (FCG). FCGs are a non-means tested cash payment to foster carers with legal custody of a child <18 years. Widely perceived as a de facto orphan grant [72], in reality only a small proportion of orphans are subject to foster care court orders [73]. In 2014, 500,000 FCGs were paid to foster carers (80% of whom were caring for orphaned children). The FCG was provided to 7.1% and 1.2% of carers for maternal and paternal orphans compared with one in three carers of double orphans (33.8%) [74]. The most commonly accessed form of assistance is the child support grant (CSG), a means-tested, unconditional grant aimed at reducing poverty and promoting investment in poor children [48]. However, parental and non-parental carers of children who are single orphans do not currently have preferential access to the CSG [75]. Government efforts to improve the capacity and delivery of South Africa’s social welfare system including the foster care arrangements and financial assistance for orphans, are ongoing [76]. Analyses of detailed, longitudinal population-based data can make a valuable contribution in ensuring statutory definitions and eligibility criteria align more closely with the diversity of residential, parenting and care arrangements in which children live [65,75,77].

In the last decade, policy and research focusing specifically on orphans and childhood adversity in South Africa have increasingly concentrated on children affected by HIV and AIDS (See for example, [78]). Related intervention programmes typically include targeted interventions to prevent sexual violence and HIV acquisition in orphans affected by HIV and AIDS, especially during adolescence [3,78]. Although the incidence of HIV/AIDS and TB maternal and paternal deaths was still higher than any other causes among older children and adolescents by 2014, it is problematic to overlook children experiencing parental deaths from other causes, particularly with respect to the impacts on early development in younger children. For example, the challenging situations faced by children and households who experience a sudden and/or violent adult death of a family member, may require specialist support. Studies conducted before HIV treatment was available in this study population described significant differences in the impact of parental death on health and wellbeing outcomes of orphans depending on the cause of parental death and the child and family circumstances before and after the death [9,15]. Consequently there is a need to continuously re-evaluate the value and effectiveness of support, particularly in relation to education, training and entry into employment, for children, adolescents and possibly young adults, who are affected by parental illness and death [3,12,79].

In conclusion, these empirical findings demonstrate that widespread availability of HIV treatment has made a profound contribution to checking and reversing mortality in mothers and fathers of young children and adolescents. Therefore, the bleakest forecasts of orphanhood levels based on orphan incidence in the 1990s and early 2000s have not been realised. However, given the dynamic and complex epidemiological and demographic changes in South Africa and other countries with HIV epidemics, we should continue to monitor the nature and level of impacts of all causes of parental ill-health and mortality in order to ensure that programme and policy responses best meet the needs of all children and adolescents.

Supporting information

Acknowledgments

We thank the communities that contributed their data to the demographic surveillance system and to other linked studies. We acknowledge the Africa Health Research Institute (formerly Africa Centre for Health and Population Studies) in making data available and to the many operational and scientific colleagues involved in the collection, validation and management of data. We also thank PloS One anonymous reviewers for their helpful comments and suggestions.