Study Context and Data Source
This study uses data from a World Bank-funded study of the welfare of orphans and vulnerable children (OVCs) of primary school-going age (6–14 years) in urban poor areas29. The OVC study was carried out in two informal settlements in Nairobi, Kenya, where the African Population and Health Research Centre (APHRC) has run a health and demographic surveillance system, the Nairobi urban health and demographic surveillance system (NUHDSS), since 2001. The NUHDSS, in which the study was nested, involves a systematic quarterly recording of vital demographic events including births, deaths and migrations occurring among household residents. The NUHDSS also regularly collects data on other health and socioeconomic issues such as household assets and amenities, morbidity, and cause of death, using verbal autopsies and education. The two slum areas that comprise the study site (Korogocho and Viwandani) are densely populated (63,318 and 52,583 inhabitants per square kilometer, respectively) and are also characterized by poor housing, high unemployment rates, lack of water supply and sanitation services, high levels of violence and general insecurity and poor health indicators.8,30 Viwandani, which is located near the industrial area, has relatively higher levels of education, employment and population mobility, while the population in Korogocho is more stable and with higher levels of co-residence of spouses.
The OVC study, which was carried out between January and June 2007, investigated various domains of child welfare. This paper uses data on nutritional status and food security among orphans and non-orphans. In common with other studies, the term orphan in this study refers to children who have lost either one (paternal/maternal) or both parents (double).
The target minimum sample size calculated for the study was 2,122. We then sought to include all orphans in the NUHDSS database (n = 1,202), with an equal number of non-orphans, randomly selected from the NUHDSS database; matched upon age, gender and location of residence at the population level. Hence, the target sample was 2,404 children. Anthropometric measurements (height and weight) were taken from the child; and interviews regarding food security were done with both the child and his/her caregiver.
Ethical approval for the study was obtained from the Kenya Medical Research Institute’s National Ethical Review Committee. Written informed consent was obtained from the child’s caregiver both for interviewing the caregiver and the child. In addition, verbal assent was obtained from the child.
Study Variables
Dependent Variables
Child nutritional status was derived from anthropometric measurements taken from all the children. All measurements were carried out using standard procedures.31 Height was measured using an inelastic tape measure with the child standing on level ground against a flat perpendicular surface and was recorded in centimeters to one decimal point. Weight was measured using an electronic scale (Seca 881 U, obtained from United Nations Children’s Fund) and was recorded in kilograms to one decimal place. Through use of the World Health Organization (WHO) 2005 Anthro program,32 height-for-age and weight-for-age z scores were generated using the 1977 National Center for Health Statistics/World Health Organization (NCHS/WHO) reference. Nutritional outcomes included height-for-age score, weight-for-age score, stunting and underweight.
Food security was measured through complementary interviews with both the caregiver and the index child separately. Questions asked sought to assess perceived hunger, regularity of meals, food access and food shortage. Answers were recoded to be unidirectional, with 0/1 being the poorest/lowest and 4/5 the best/highest; a composite measure was then derived by summing up standardized scores of all responses. Cases with missing information on any of the food security variables were not included in the generation of the composite score (a total of 63 cases). Such missing information was mainly due to questions with reference to a specific date, e.g. if the child was away from the household on the reference day (see Table 7 for questions contributing to the food security score).
Independent Variables
The orphan status of children—our key predictor—was defined using two specifications: (1) non-orphan vs. orphan and (2) father/paternal orphan vs. mother/maternal orphan vs. double orphan. Other explanatory variables, mainly extracted from the NUHDSS database, included location of residence (Korogocho, Viwandani); child’s age, sex, ethnicity and relationship to the household head; household head’s age, sex and highest level of education; number of children <15 years in the household; household socioeconomic status (constructed using principal component analysis of the following amenities and assets: electricity supply, bicycle, television, radio, house phone, sofa, table, flush light, kerosene lamp, kerosene stove and wall clock). Household wealth tertiles were generated from the wealth index using the Stata’s xtile command and labeled as poorest (lowest 1/3), middle and least poor (highest 1/3).
Statistical Analysis
Analysis was carried out to test the following hypotheses:
-
Orphans are more vulnerable than their non-orphan counterparts in relation to food security and have poorer nutritional status;
-
Paternal orphans are worse off than maternal orphans in relation to food security and nutritional outcomes;
-
Double orphans are worse off than their paternal and maternal orphan counterparts in relation to food security and nutritional outcomes.
Only children for whom information from two sources (the index child and the caregiver) was captured were considered in the analysis (n = 1,235: 467 orphans and 768 non-orphans). Five hundred nineteen caregivers were unavailable for interview because of migration (314 permanent, 10 temporary), refusals (48), deaths (17) and untraceable (130). A further 797 children were excluded due to migration (312 permanent and 193 temporary), refusals (44), death (3) and untraceable (245). Overall, there were 1,550 children with a corresponding caregiver interview: 950 non-orphans and 600 orphans. Three hundred fifteen of these children were aged 15 years or older (due to some time lapse between sampling from the NUHDSS database and the actual study time), and hence excluded from analysis. Thus, 1,235 children were included in the analysis.
The analysis, done using Stata version 10.0 (StataCorp LP, USA), involved both descriptive and multivariate regression methods. Chi-square test was used to test for differences in proportions by orphan status. Initially, mean group differences with regards to nutritional outcomes and food security score were analyzed through t test for orphan status (orphan/non-orphan) and a one-way ANOVA for categories of orphanhood (father, mother and double). Subsequently, random intercepts regression models were used in the multivariate analysis using the Stata’s xtmixed command (for linear regression) and xtlogit command (for logistic regression) to allow for clustering at the household level, given the structure of the sample. The 1,235 children included in the study were nested within 1,034 households: 1 household had 4 children, 29 households had 3 children each and 140 households had two children each, while the rest of the 864 households hosted one child each. The mixed effect model was used to account for both fixed effects and random effects at the child level and at the household level, respectively.