The growing number of orphan and vulnerable children (OVC) in Africa—approaching 12 million children—has been described as the most tragic and long-term legacy of the HIV/AIDS pandemic in Africa (UNAIDS, UNICEF, & WHO, 2008). The primary caregivers of OVC face the dilemma of maintaining their own health while addressing the physical and emotional needs of additional family members (Muliira, 2011). These women are experiencing more severe economic and social problems (e.g., discrimination, stigma due to HIV/AIDS, and lower income) than other women from their own towns (Tarimo et al., 2009; Singh et al., 2011). Taking responsibility for the care and protection of orphans has negative implications for caregiver health over time and OVC caregivers experience higher levels of distress and chronic illness and less support from other adults compared to women who are not OVC caregivers (Littrell et al., 2009; Govender et al., 2012). Regardless of the large number of women who are primary caregivers of OVC, current studies involving OVC have focused on children’s needs and overlooked challenges faced by their primary caregivers including nutritional effects of orphan care. Studying OVC primary caregivers’ nutritional status is important because child caregiving could impact children, families, and the health of these caregivers.

The ongoing African nutritional transition is characterized by a shift toward a more energy dense diet, with increased fat, sugar, processed foods, and reduced physical activity (FAO, 2006). The BMI of 15–49 year old women is much higher in southern Africa than in other African regions (Garenne, 2011). According to 2006–2007 Demographic Health Surveys (DHS) country reports, the overweight prevalence among 15–49 year old women is greater in Namibia and Swaziland (28.0%) than in Zambia (19.2%), (Kanamori et al., 2013). Among OVC primary caregivers, overweight prevalence (Namibia: 33.2%, Swaziland: 61.3%, Zambia: 26.9%) is higher than underweight prevalence (Namibia: 13.1%, Swaziland: 1.7%, Zambia: 7.2%), (Kanamori, 2013).

Previous studies in Africa have shown that women with lower levels of education and/or household decision-making autonomy are more likely to be underweight (Woldemicael, 2007; Elfhag & Rossner, 2005). Therefore, promoting women’s autonomy has been identified as a critical and cost-effective nutritional intervention for helping households break the cycle of intergenerational deprivation (United Nations, 2012).

This study’s objective was to investigate the association between married women’s autonomy and their nutritional status and to explore whether this relationship was modified by OVC primary caregiving. Our study focused on two dimensions of women’s autonomy: decision-making about a variety of household issues and knowledge autonomy measured by educational attainment (Woldemicael, 2007). Three OVC primary caregiving categories were included: a) OVC primary caregivers, b) non-OVC primary caregivers, and c) non-primary caregivers. Research questions were: 1) What is the association between married women’s autonomy and their underweight/overweight status? 2) Does OVC primary caregiving status modify the effect of the association between married women’s autonomy and their underweight/overweight status? and, 3) Does OVC primary caregivers’ decision-making autonomy mediate the association between AWI and their BMI?

A cross-sectional study was conducted using secondary analyses of DHS data from Namibia (2006–2007), Swaziland (2006–2007), and Zambia (2007). This study included 20–49 year old married women who gave oral consent for participation and reported de jure household residency. Pregnant and three months or less postpartum women were excluded to avoid the impact of the fetus and lactation on BMI (Bitew, 2010). Women with BMI less than 12.0 or greater than 60.0 were also excluded as these might be cases of extreme anthropometric measures or resulting from data errors (Garenne, 2011). The Opinion Research Corporation Company Macro Institutional Review Board approved every DHS surveys (ICF, 2012).

The dependent variable, BMI (kg/m²) was categorized in three groups: underweight (BMI < 18.5), normal weight (18.5 ≥ BMI < 25.0), and overweight (BMI ≥ 25.0).

The independent variable, household decision-making autonomy included five situations to determine whether a woman had the final say on: (1) her own health care, (2) large household purchases, (3) purchases of everyday household items, (4) visits to family or relatives, and (5) what to do with the money the husband earns. The response options were: (a) respondent alone, (b) respondent and husband/partner, (c) respondent and other person, (d) husband/partner alone, (e) someone else, f) other. For each question, a value of 1 was assigned if the response was (a), (b), or (c), and 0 for (d), (e), or (f). Decision-making autonomy was included as a dichotomous variable for each of the five situations as well as a continuous measurement (the values were added resulting in a score from 0 to 5), (Rahman, 2012). Educational attainment included: no education, primary, secondary and higher education.

The potential effect modifier, OVC primary caregiving status, included three categories. Primary caregivers were women who lived with a biological child under the age of 18 and/or who were primary caregivers of a non-biological child under the age of 18. This study uses DHS’s definition for OVC: children with one or both parents deceased (orphans); and vulnerable children who a) have a chronically ill parent (sick for more than 3 consecutive months during the past 12 months) or b) live with an adult who was chronically ill or died during the past 12 months. The OVC status of the primary caregiving was based on the question: “How many orphans and vulnerable children live in your household?” OVC primary caregivers were primary caregivers who mentioned that one or more children living in their house. Non-OVC primary caregivers were primary caregivers who mentioned that no OVC lives in their house. Non-primary caregivers were women who did not live with a biological child under the age of 18 and/or who were primary caregivers of a non-biological child under the age of 18.

The Absolute Wealth Index (AWI) developed by Hohmann and Garenne (2011) and adapted from the Filmer and Pritchett’s approach captures a wide range of household wealth distribution (Filmer & Pritchett, 1999; Garenne, 2011; Hohmann & Garenne, 2011).

Potential confounders included: women’s age, education, work status, region of residence, parity and relationship with the household head; presence of a child 5 years old or younger living in the house; sex of the household head; and, number of household members, 18–49 year old women and men in the household.

Statistical analyses included data screening to check for outliers and errors, as well as descriptive statistics for continuous and discrete variables. Estimation of marginal means used analytic weights to correct for over-sampling and variations in survey response rates by region. Logistic regression analyses controlled for potential confounders and interactions. Additional simple and adjusted logistic regression models were used for selecting confounders. A variable was included as a confounder if the adjusted odds ratio was 10% different than the unadjusted odd ratio.

Collinearity between and within independent variables and potential confounders was tested by using Phi analyses when both variables were dichotomous, and Cramer’s V for variables with three or more levels. A pair of independent variables with a Phi or and Cramer’s V value of at least .70 was considered to have substantial collinearity and one variable was removed from the analysis. Multicollinearity among the independent variable and potential confounders was tested using two statistics: the tolerance value and the variance inflation factor (VIF). Multicollinearity was also considered to exist if a tolerance value less than 0.1 or a VIF greater than 10 were found (Menard et al., 2006; Myers et al., 2006). Mediation analyses were performed using Baron & Kenny’s approach (1986). Significant associations were determined using 95% confidence intervals and p-values (α <0.05) using SPSS® 19.

Our findings demonstrate the important role that educational attainment has in the nutritional status among African married women. Having education decreased the odds for being an underweight woman in Namibia, and increased the odds for being an overweight woman in Swaziland and Zambia. In Namibia, women who had at least a secondary education were less likely to be underweight as compared to women with no education. Education has been identified as crucial for poverty reduction; improvement in gender equality, health and nutrition; reductions in infant and child mortality; and a reduction in the prevalence of HIV/AIDS transmission in sub-Saharan Africa (Fotso et al., 2009). It has been suggested that education protects women’s nutritional status because educated women are more likely to have better skills that could benefit their children and/or their own health and nutrition (e.g., use health care facilities; keeping their places clean; access to health and nutrition information; from knowledge gleaned from reading newspapers, listening to the radio and watching television; effective interaction with doctors and nurses; better compliance with treatment recommendations), (Joshi, 1994).

In developed countries, education has also been associated with eating more frequently out of the home and lower adherence to the WHO recommended daily intake of vegetables (Krige et al., 2012). Further work should explore whether the association between education and these health behaviors is a possible explanation why our study also found that being a woman with at least primary education (Zambia) or at least secondary education (Swaziland) increased the odds for being overweight. Future studies should also explore the reasons why education is not associated with overweight status among non-child caregivers in Zambia.

This study found that Namibian women who had the final say on what to do with their husband’s salary were less likely to be underweight. This finding supports UNESCO guidelines for addressing undernutrition problems in Africa including the fight against traditional stereotypes in the minds of both men and women that are reinforcing gender inequality (United Nations, 2012). The acquisition of basic literacy and numeracy is essential to promote women’s capacity for taking charge of the household income and budgeting. Our analyses including OVC primary caregivers showed that decision-making autonomy mediated the association between AWI and women’s BMI in Zambia. Previous studies did not find women’s autonomy to mediate the association between socio-economic factors and pre-natal and child delivery care services (Woldemicael, 2007).

As documented in DHS’s country reports (2006–2007), the reasons why we found significant interaction and mediation effects only in Zambia could be related to the comparatively lower prevalence of households in Zambia than in the other two countries that have a female headship, a higher percentage of men with two or more wives, and lower women’s mean age (years) at first marriage in urban (Namibia: 29.1, Swaziland: 27.9, Zambia: 19.6) as well as rural areas (Namibia: 27.4, Swaziland: 22.8, Zambia: 17.8). These factors could reflect different reasons and motivations for getting married, their levels of decision-making autonomy and their likelihood of continuing with higher education. Another important rationale could be differences in women’s status in society by country as measured by levels of tolerance for violence against women. DHS reports (2006–2007) show that a much higher percentage of women and men in Zambia than the other two countries consider that a husband is justified in hitting or beating his wife in several situations (e.g. his wife burns the food, argues with him, goes out without telling him, neglects the children or refuses sexual intercourse with him).