Health and nutritional status of orphans <6 years old cared for by relatives in western Kenya
Kim A. Lindblade, CDC/KEMRI, Unit 64112, APO, AE 09831, USA. Fax: +254-35-22981; E-mail email@example.com
One of the consequences of the HIV/AIDS epidemic in sub-Saharan Africa is the increase in the number of orphans, estimated to have reached 6–11% of children <15 years old in 2000. Orphans who stay in their communities may be at increased risk for poor health due to reduced circumstances and loss of parental care. We have used data from a population-based study in rural western Kenya to compare basic health and nutritional indicators between non-orphaned children <6 years old and children who lost either or both of their parents. In June 2000, all children <6 years old who had been recruited for a cross-sectional survey in 60 villages of Rarieda Division, western Kenya, in June 1999 were invited to return for a follow-up survey. Basic demographic characteristics, including the vital status of the child's parents, and health histories were requested from all 1190 participants of the follow-up survey, along with a finger-prick blood sample for determination of malaria parasite status and haemoglobin (Hb) levels. Height-for-age (H/A) and weight-for-height (W/H) Z-scores were also calculated from anthropometric measurements. Overall, 7.9% of the children had lost one or both their parents (6.4% had lost their father, 0.8% had lost their mother and 0.7% had lost both parents). While there was no difference between orphans and non-orphans regarding most of the key health indicators (prevalence of fever and malaria parasitaemia, history of illness, Hb levels, H/A Z scores), W/H Z-scores in orphans were almost 0.3 standard deviations lower than those of non-orphans. This association was more pronounced among paternal orphans and those who had lost a parent more than 1 year ago. These results suggest that the health status of surviving orphans living in their community is similar to that of the non-orphan population, but longitudinal cohort studies should be conducted to determine better the overall impact of orphanhood on child health.
Without doubt, HIV/AIDS is one of the most important public health problem ever faced by sub-Saharan Africa. One of the consequences of the epidemic is the increase in the number of orphans; by the year 2000, it was estimated that 6–11% of children <15 years old in East Africa would have lost either their father or mother (Preble 1990). There have always been many orphans in Africa, but deaths due to HIV/AIDS may be responsible for a 40% increase (Kamali et al. 1996). In Kenya, where HIV prevalence can exceed 25% (Ayisi et al. 2000), it has been estimated that the number of motherless orphans will grow to more than 1.5 million by 2005 (AIDS Control Unit 2001). In Africa, children whose parents die are doubly burdened, losing not only the attention, care and advice that a parent gives, but also access to household resources such as housing and land. As a result, orphaned children are especially vulnerable and potentially at increased risk of poor health (Ayieko 1997).
While the general practice in sub-Saharan Africa is that orphaned children are cared for by the extended family rather than in the orphanages, anecdotal evidence indicates that the increasing number of HIV/AIDS orphans is pushing community resources to the limit (Foster 2002). The few published studies on the health status of orphans, however, do not clearly indicate whether orphans cared for within their community constitute a high-risk population that should be targeted for public health intervention (Ryder et al. 1994; Kamali et al. 1996; Panpanich et al. 1999).
We used data from a population-based study in western Kenya to compare basic health and nutritional indicators between non-orphaned children <6 years old and children who lost either or both parents (orphans) to determine whether orphaned children living with the surviving parent or other relatives are at increased risk of poor health and malnutrition.
Informed consent was received from the caregivers of all participating children. The study protocol was approved by the institutional review boards of both Centers for Disease Control and Prevention (CDC) and Kenya Medical Research Institute (KEMRI).
Study site and population
The study took place within a larger investigation of the impact of insecticide-treated bednets on children in Rarieda Division (Asembo), Bondo District, located in Nyanza Province, north-east of Lake Victoria. The population of rural Asembo (178 km2) was 56 883 in 1999, with children <6 years comprising 19.5% of the population (Central Bureau of Statistics 1999). Inhabitants are predominantly (95%) from the Luo tribe, whose main occupations are subsistence farming and fishing. Extended families generally reside in compounds, i.e. in groups of two or more houses. Polygyny is common and cowives generally inhabit different houses within the same compound. The towns of Kisumu and Siaya, where government hospitals are located, are within 60 km of the centre of the study area.
Children <6 years old are at the greatest risk of illness in this population and account for approximately 94% of both paediatric outpatient visits at peripheral health facilities and inpatient admissions at the two regional district hospitals. Malaria is the predominant cause of illness in children <6 years of age and is responsible for 70–72% of diagnosed cases at both peripheral health facilities and hospital inpatient admissions (CDC, unpublished data). Between 78% and 90% of children <5 years may be parasitaemic at any time (Bloland et al. 1999). While malaria is common throughout the year, peak prevalence occurs at the end of the rains (i.e. December and June). Anaemia is also common; population-based cross-sectional surveys between 1997 and 1999 found 66% of children with haemoglobin (Hb) <11 g/dL (Desai et al. 2002). Between 1992 and 1996, the estimated infant mortality rate was 176 per 1000 live births with under-five mortality at 257 per 1000 (McElroy et al. 2001). Reported HIV prevalence rates in pregnant women living in rural and peri-urban areas of neighbouring Kisumu District are 21–37% (AIDS Control Unit 2001).
Sampling and survey methodology
In June 1999, as part of a cross-sectional survey to compare malaria indicators between children sleeping under insecticide-treated bednets and those without, 40% of all houses located in 60 of the 75 villages of Asembo were randomly selected and all children <5 years old and their caregivers were invited to participate in a health survey. One year later, these children were traced and requested to attend a follow-up survey. Additionally, all new siblings of the study children and other children currently residing in the same house as a study child were invited to participate in the second survey. Information on the vital status of the children's parents was only collected at the follow-up survey in May/June 2000.
Children and their caregivers were requested to assemble in a central location in each village on the day of the survey. After receiving written, informed consent, information was collected from the caregivers on basic demographic characteristics and the child's recent illness and treatment history. Parents' vital status was determined by questioning the caregiver as to whether the child's mother or father were alive. An orphan was defined as a child who had lost either or both parents and were further categorized as maternal, paternal or double (both parents deceased) orphans.
The children's caregivers were questioned regarding the children's history of illness over the previous 2 weeks using illness terms translated into the local language (dhoLuo). In addition to questions about general illness, caregivers were asked to report on two specific health indicators associated with the quality of parenting: diarrhoea and upper respiratory tract infection (URTI) (Ware 1984). The variable indicating history of URTI was constructed from reported runny nose with ear pain or ear infection. Caregivers were also asked whether treatment for any illness had been sought in the previous 2 weeks. Use of bednets by the children for more than 5 days/week was recorded.
We took the axillary temperature of all children using a digital thermometer and used the mean of two independent measurements. Fever was defined as axillary temperature ≥37.5 °C. A finger-prick blood sample was taken for determination of Hb concentration and malaria parasitaemia. Hb concentrations were measured at the time of the survey using a HaemoCue machine (HaemoCue, AB, Angelholm, Sweden) and were later transported to the Kenya Medical Research Institute (KEMRI)/Centers for Disease Control and Prevention (CDC) field station in Kisumu for repeat testing on the same day using a Coulter counter (Coulter Corporation, Miami, FL, USA) (Hb measurements from the Coulter counter were used in the analysis). Blood slides were stained at the KEMRI/CDC lab with Giemsa and examined for asexual malaria parasites using standard protocols.
Two independent readings of both weight (using a Salter scale) and height/length were made and the mean of the two values was used. Weight-for-height (W/H) and height-for-age (H/A) Z scores were calculated using growth reference curves developed by the US National Center for Health Statistics and WHO (Hamill et al. 1979) using Epi Info (2000, CDC Atlanta, GA, USA).
All children were examined at the time of the survey by a trained clinical officer and treated for illness according to Kenya national guidelines. All children with anaemia (Hb < 11 g/dl) were given iron supplementation. Children with either fever (axillary temperature ≥37.5 °C) or anaemia (Hb < 11 g/dl) were presumptively treated for malaria with sulphadoxine-pyrimethamine (SP) at the time of the survey according to Kenya national guidelines (Malaria Control Unit 2000). Any asymptomatic children found to have parasite densities >5000 parasites/μl were treated with SP the next day. Samples from these children were not tested for HIV.
Only data from the follow-up survey in June 2000 were used in the analysis comparing orphans and non-orphans. At this survey, the sample of children was <6 years of age. All analyses were conducted using SAS (version 8, SAS Institute, Cary, NC, USA). Children were first categorized into orphans and non-orphans, and then classified according to the vital status of the parents (paternal, maternal and double orphans) to separate out the singular effects of loss of a mother and loss of a father. Children were further classified according to the time since loss of a parent, i.e. non-orphans, orphans who had lost a parent within the last year and orphans who had lost a parent more than 1 year ago. We used the GENMOD procedure to calculate prevalence ratios (PR) and 95% confidence intervals (CI) for the association between dichotomous variables and orphan status adjusted for clustering by household. The MIXED procedure was used to calculate the difference in continuous variables by orphan status with 95% CI adjusted for clustering by household. Age was included in all models due to the age difference between orphans and non-orphans.
A total of 1347 children <5 years participated in the baseline survey in June 1999; 1055 (78.3%) of these children were traced and participated in the follow-up survey in June 2000. Forty-one (3.0%) of the original sample died before the follow-up survey, 222 (16.5%) moved out of the study area or were away from home at the time of the follow-up survey, and 29 (2.2%) declined to participate. All new siblings or other children currently living in the same house as the study participant were invited to participate in the follow-up survey; an additional 224 children were recruited for a final sample size of 1279 children <6 years old in the follow-up survey. The status of the father was unknown in 87 (6.8%) cases of the 1279 children in the sample; they were excluded from further analysis because it is not clear whether the impact of absence of a father is similar to loss of a father. Due to missing clinical data for two children, the final analytic sample size was 1190.
The children enrolled at baseline who died before the follow-up survey were significantly younger at baseline (mean age: 17.0 months) than those who went on to participate in the follow-up survey (mean age: 28.2 months). The deceased children also had a lower baseline Hb and lower baseline H/A and W/H Z scores than the participants, although they were less likely to be parasitaemic. All 41 children who died had been brought to the baseline survey either by their mother (39 children) or their father (two children). Children who were absent from the follow-up survey because they had moved or were away from home were also significantly younger (23.8 months old) than the participants and had lower Hb but were similar with regard to other variables. Those who refused to participate in the follow-up survey had higher W/H Z scores than the participants but were otherwise similar.
Vital status of parents
Overall, 94 (7.9%) of the 1190 children with complete information on both parents had lost one or both of their parents. Most (6.4%) had lost their father, 0.8% had lost their mother and 0.7% had lost both parents. This distribution is similar to that from a census of the entire population of Asembo conducted in late 2001 (J. Kwach, personal communication) and a study of orphans in Uganda (Kamali et al. 1996). The mothers of seven of the 18 (39%) maternal orphans had died within the last year and a similar proportion (36 of 84; 42.9%) of paternal orphans had lost their father during that period. Overall, 41 (3.4%) children in the survey had lost a parent within the last year.
Most of the non-orphans (92.5%) and paternal orphans (92.1%) were brought to the follow-up survey by their mothers. In contrast, children whose mother had died were brought by a cowife to the mother (33.3%), a grandparent (16.7%) or another relative (41.7%). These proportions were similar for double orphans.
Characteristics and health status of orphans
Orphans (mean age: 41.2 months; range 5.2–70.2) were significantly older than non-orphans (mean age 34.3; range 0.2–71.0; difference 6.9, 95% CI: 2.8, 11.0) in this sample. All subsequent analyses were therefore adjusted for age. Girls were less likely to be in this sample of orphans than boys, but this difference was not statistically significant (Table 1). There was no disparity in the use of bednets between orphans and non-orphans.
Table 1. Associations between demographic and health characteristics and orphan status, May/June 2000, western Kenya
|Female||45 (47.9)||575 (52.5)||1.22 (0.79, 1.87)|
|Slept under a bednet >5 days/week||86 (92.5)||1018 (93.1)||0.99 (0.43, 2.28)|
|Axillary temperature ≥37.5 °C||6 (6.4)||76 (6.9)||1.02 (0.43, 2.38)|
|Malaria parasitaemia||56 (59.6)||563 (51.4)||1.16 (0.75, 1.80)|
|History of illness in the past 2 weeks|
| Any illness||84 (89.4)||966 (88.4)||1.14 (0.62, 2.31)|
| Upper respiratory tract infection||10 (10.6)||169 (15.5)||0.70 (0.35, 1.40)|
| Any diarrhoea||45 (47.9)||485 (44.4)||1.35 (0.85, 2.14)|
| Sought treatment for any illness||74 (79.6)||845 (78.0)||1.18 (0.67, 2.07)|
|Mean (range)||Mean (range)||Difference (95% CI)†|
|Haemoglobin (g/dl)|| 9.2 (6.0–13.9)|| 9.5 (4.1–18.5)||−0.2 (−0.6, 0.1)|
|Height-for-age Z score||−0.97 (−5.80, −1.94)||−1.03 (−5.71, 2.84)|| 0.06 (−0.24, 0.35)|
|Weight-for-height Z score||−0.59 (−4.32, 1.37)||−0.31 (−5.01, 2.93)||−0.28 (−0.50, –0.06)|
There were no significant differences in key health indicators between orphans and non-orphans (Table 1). The proportions of children with fever, parasitaemia, a reported history of any illness, diarrhoea, upper respiratory disease and treatment for any illness in the past 2 weeks did not differ by orphan status, nor did mean Hb values or mean H/A Z scores. Only in W/H Z scores did orphans appear at a disadvantage compared with non-orphans, with W/H Z scores almost 0.3 standard deviations (SD) lower in orphans.
All analyses were repeated with dummy variables indicating whether a child had lost a mother, father or both parents to compare with non-orphans (data not shown). There were no differences in demographic or health variables by parents' vital status except in the case of W/H Z-scores, where paternal orphans had Z scores 0.29 SD less than non-orphans (adjusted 95% CI: −0.54, −0.05). Differences between maternal or double orphans and non-orphans were not significant, although the sample size in both of these orphan categories was very low.
The time since loss of a parent was examined for association with demographic and health characteristics. Children were classified as non-orphans, those who had lost a parent within the last year and those who had been orphaned more than a year ago. Children who had been orphaned for more than 1 year were more likely to have reported diarrhoea in the last 2 weeks (adjusted PR 1.92, 95% CI: 1.04, 3.54) than non-orphans, although there was no difference between orphans who had lost a parent <1 year ago (adjusted PR 0.85, 95% CI: 0.43, 1.66) and non-orphans. The W/H Z scores of children who had lost a parent more than a year ago were on average 0.49 SD (adjusted 95% CI: −0.77, −0.22) lower than non-orphans, although the Z scores of children who had been orphaned more recently were not significantly different from non-orphans (adjusted difference – 0.01 SD, 95% CI: −0.33, 0.31). There were no differences among any of the other demographic or health characteristics by time since loss of a parent.
The proportion of orphans in this sample of children <6 years old living in a rural area of western Kenya was 7.9%, similar to the findings of other community-based studies in sub-Saharan Africa (Foster et al. 1995; Kamali et al. 1996) and within the range for East Africa estimated for 2000 (Preble 1990). A population-based census of the study area conducted in December 2001 found a similar proportion (9.5%) of children in this age group who had lost mother or father or both parents (J. Kwach, personal communication). Extrapolating the proportion of orphans found in this study to the population of Nyanza Province, there are an estimated 15 700 orphaned children <6 years old in the province, most of whom are dependent on the extended family network.
The impact of orphanhood on the health and nutrition of Kenyan children is likely to be greatest on young children. Children <6 years of age are dependent on adults to provide them with food, shelter and care. They are also especially vulnerable to a variety of infectious diseases, including malaria, diarrhoea and respiratory disease, and also anaemia, as evidenced by the high rate of outpatient visits and inpatient admissions in young children. As orphans are at increased risk of being HIV-positive (Ryder et al. 1994; Kamali et al. 1996), in the absence of appropriate antiretroviral treatment orphans <6 years of age are more likely than non-orphans to develop an HIV/AIDS-related illness and die before they reach school age (AIDS Control Unit 2001). Therefore, any impact of orphanhood on health should be detectable in children <6 years of age.
However, we failed to find compelling evidence that orphaned children in this age group constitute a high-risk population as measured by general health indicators. Compared with non-orphaned children, orphans were not at greater risk for fever, malaria parasitaemia, or history of illness, while Hb levels and H/A Z scores were similar between orphans and non-orphans. But W/H Z scores were lower among orphans, and particularly among orphans who had lost a parent more than 1 year ago, possibly indicating that loss of a parent results in decreased household resources with immediate impact on child nutritional status.
Although it is generally held that maternal orphans are at greater risk for health problems because of the loss of their primary caregiver, children who had lost a father were more likely to be malnourished than non-orphans, indicating that loss of a father may be at least as significant as loss of a mother. Among the Luo of Kenya, women keep house, cultivate the shamba (garden), gather firewood, prepare meals and look after children. Fathers provide cash income through fishing, shopkeeping, petty trade or casual labour. A mother may find it difficult to cope without this income after her husband's death. An additional consequence of the death of a male household head may be the loss of household capital through sale of resources to pay for health care, or because the husband's family may reclaim household property after his death. However, the number of maternal orphans in this sample was very low and a larger sample of orphans will be needed to determine whether impact of orphanhood on malnutrition differs according to the sex of the parent lost.
Other studies comparing the health status of orphans to non-orphans found no important differences, even after long follow-up periods (Ryder et al. 1994; Panpanich et al. 1999). However, one of these studies defined orphans only according to the vital status of the child's mother and consequently did not include paternal orphans. The other study did not differentiate maternal from paternal orphans. If paternal orphans are more at risk for malnutrition, these studies would not have been able to detect such an effect.
The results of this study may have been biased because of loss to follow-up. As orphans are more likely to be HIV-infected than the general population of children (Ryder et al. 1994; Kamali et al. 1996), they are also likely to suffer a higher mortality rate. In addition loss of one or both parents may be associated with leaving the study area. Therefore this sample of children taken from a follow-up survey is likely to have lost orphans from the original sample. Unfortunately, we did not gather information on the vital status of parents or the HIV status of children at the baseline survey and are therefore unable to directly evaluate the effect of this potential bias. The children who died between the baseline and follow-up surveys had lower Hb, W/H and H/A Z scores. It is therefore possible that we have underestimated the impact of orphanhood on these indicators.
Survival bias is also likely to have been a factor in this study. If loss of a parent is associated with increased risk of death in children, our sample is likely to have been taken from among those children who survived the death of their parent. Additionally, death of the household head may be associated with migration of the remaining household (Urassa et al. 2001). Without following children from before the death of a parent, it will not be possible to fully examine the impact of orphanhood on child health.
In summary, we have found that surviving orphans <6 years old who remain in their communities are in general at no greater risk of poor health than the overall population of children in that age group although orphans may be more likely to be malnourished. Further research is needed to determine whether paternal orphans are at greater risk of acute malnutrition than maternal orphans, and paternal orphans should be included, where possible, in studies of the impact of orphanhood. Additionally, to avoid potential survival bias, longitudinal cohort studies are needed to more fully explore the effect of orphanhood on the health status of children.
We thank the CDC/KEMRI field staff who assisted in collecting these data, in particular Michael Onyango. We are indebted to Feiko O. ter Kuile, D.J. Terlouw and Arthur Kwena for organizing and supervising the baseline survey. Pauli Amornkul and Laurence Slutsker made important editorial comments that improved the quality of this paper. We are very grateful to the children and their caregivers for participating in this research. This paper was published with the permission of the Director of the Kenya Medical Research Institute (KEMRI). Studies on the efficacy of insecticide-treated bednets were supported by the US Agency for International Development (USAID).