|Home | About | Journals | Submit | Contact Us | Français|
Domestic violence has harmful physical and psychological health correlates, but there is little evidence regarding a relation between domestic violence and malnutrition. To investigate this relation, the authors analyzed data from 69,072 women aged 15–49 years and 14,552 children aged 12–35 months in the 1998–1999 Indian National Family Health Survey. Physical domestic violence victimization was self-reported by the women. Aspects of nutritional status included in this study were anemia and underweight. Anemia was measured with a blood test for hemoglobin. Underweight was calculated from anthropometric measurements and was determined as body mass index for women, and it included stunting and wasting for children. Results indicate associations of multiple incidents of domestic violence in the previous year with anemia (odds ratio = 1.11, 95% confidence interval: 1.04, 1.18) and underweight (odds ratio = 1.21, 95% confidence interval: 1.13, 1.29) in women and a suggested relation among children. Possible mechanisms for this relation include withholding of food as a form of abuse and stress-mediated influences of domestic violence on nutritional outcomes. These findings indicate that reducing domestic violence is important not only from a moral and intrinsic perspective but also because of the instrumental health benefits likely to accrue.
Malnutrition is the leading risk factor for mortality in the world, contributing to nearly 12 percent of all deaths and 16 percent of all disability-adjusted life years lost globally (1) and one third of all disability-adjusted life years lost in low-income countries (2). Anemia is a risk factor for a number of health outcomes, including poor cognitive development, weakened resistance to infection, maternal mortality, pre- and perinatal loss, stunting, and endocrine disruption (3). Nutritional anemia results from inadequate availability of micronutrients required for hemoglobin synthesis, accounting for 841,000 deaths and 35,057,000 disability-adjusted life years lost annually, predominantly in low-income countries (4). Alternatively, hemolytic anemia results from the premature destruction of red blood cells before their normal 120-day lifespan has been reached (5). Underweight due to energy deficiency, another form of chronic malnutrition, is associated with poor birth outcomes, inability to breastfeed, suppressed immune function, and elevated risk of death in women (6). Underweight among children annually accounts for 3,748,000 deaths and 9.5 percent of all disability-adjusted life years lost worldwide (7).
The burden of chronic malnutrition falls disproportionately on developing countries, including India. In 1998–1999, anemia afflicted 74 percent of children and 52 percent of women of childbearing age in India, while 47 percent of children and 36 percent of women of childbearing age were underweight (8). Recent estimates suggest that prevalence of underweight remains virtually unchanged (46 percent for children, 33 percent for women of childbearing age), while anemia increased (79 percent for children, 57 percent for women of childbearing age) (8).
Although biomedical and socioeconomic risk factors for malnutrition in India are well known (3, 9–11), the role that psychosocial factors play in anemia and underweight is less understood. Exposure to domestic violence is a psychosocial factor that may be associated with anemia and underweight for women as well as their children. Domestic violence increases psychological stress among Indian women (12, 13), and witnessing maternal domestic violence increases psychological stress among children (14). Psychological stress increases oxidative stress (15–18) and metabolic levels (19), risk factors for anemia and underweight. Domestic violence could also increase the risk of poor nutritional outcomes through environmental circumstances such as the withholding of food by abusive family members (20).
Exposure to domestic violence is highly prevalent worldwide (21, 22), particularly in south Asia (23–25), and is associated with disadvantaged socioeconomic (23, 24, 26–29), demographic (30–32), and contextual (33–35) characteristics. This widespread exposure to domestic violence indicates that abuse may exacerbate the negative influence that cumulative disadvantage has on nutritional status. We are aware of only one study, based on a community sample of rural women and children (36), that has examined this relation in India. Drawing upon the most recent nationally representative data with objective measures of anemia and underweight, we investigated the extent to which domestic violence predicts the likelihood of being malnourished among women and children in India.
We utilized the 1998–1999 Indian National Family Health Survey, a nationally representative cross-sectional study of 92,447 households (37). Trained data collectors interviewed an adult member in each selected household to obtain demographic information about the household and its family members, achieving a household response rate of 98 percent. From these households, the data collectors conducted face-to-face interviews with 90,303 ever-married women aged 15–49 years and took anthropometric measurements and blood samples. The women's survey had a response rate of 96 percent and provided information on domestic violence. We excluded women who were pregnant, nonmarried, or missing information on anemia, body mass index, domestic violence, and other predictor variables, yielding a final sample of 69,072 women located in 3,190 primary sampling units in 26 Indian states. Primary sampling units, hereafter termed neighborhoods, were villages or village clusters in rural areas and census enumeration districts in urban areas.
Each woman was also questioned about the births of her youngest two children born in the previous 3 years, yielding a sample of 33,026 children. Anthropometric measurements and blood samples were taken from those children who were both alive and present in the household. Children who were less than age 12 months, had died, or for whom information on anemia, anthropometric measurements, maternal domestic violence, and other covariates was missing were excluded from the analysis, yielding a sample of 14,552 children in 3,027 neighborhoods within 26 Indian states.
We analyzed two outcomes for the women—underweight and anemia—and five outcomes for the children—anemia, low height for age (stunting), low weight for height (wasting), low weight for age (underweight), and low body mass index for age. The anthropometric outcome variables measuring children were selected to provide a complete assessment of nutritional status (38). Specifically, stunting is a measure of chronic undernutrition, wasting is a measure of acute caloric deficiency, and weight for age and body mass index for age provide a general picture of overall nutritional status accounting for age (38, 39). Women and children were weighed with a solar-powered scale accurate to within 100 g and were measured with an adjustable measuring board accurate to within 1 mm (37). For the women, binary variables were created to define underweight and severe underweight as a body mass index of less than 18.5 kg/m2 and of less than 16.0 kg/m2, respectively (40). For the children, binary variables were created to define stunting, wasting, low weight for age, and low body mass index for age. Each anthropometric measure of child malnutrition was defined as more than two standard deviations (moderate) or more than three standard deviations (severe) below the mean scores determined by the World Health Organization (41).
Anemia status was determined by using the concentration of hemoglobin in capillary blood. After a finger prick, the first two drops of blood were removed. The third drop of blood was drawn into a cuvette for analysis using the HemoCue system (37). For women, binary variables were used to define anemia and severe anemia as a blood hemoglobin level of less than 12.00 g/dl and less than 7.00 g/dl, respectively (42). For children, the analogous levels for anemia and severe anemia were less than 11.00 g/dl and less than 7.00 g/dl (3).
Women's experience of domestic violence was the main exposure of interest in this study. Although abuse can take many forms, we conceptualized domestic violence as physical abuse. We created a categorical exposure variable to measure whether the woman reported being physically abused by a family member more than once in the previous 12 months, once in the previous 12 months, since age 15 years but not in the previous 12 months, or never since age 15 years. The exposure in the children's sample was based on which domestic violence category the child's mother reported.
We included several socioeconomic and demographic variables: age, religion, caste, education, living standard, employment status, urban-rural status, illness, breastfeeding, number of children, and decision-making autonomy (table 1). Age was specified in 5-year categories. Religion was grouped as Hindu, Muslim, Christian, Sikh, or other. Caste was based on the identification of the household head as belonging to a scheduled caste, scheduled tribe, other backward class, or the general class. Scheduled castes are those whose members have suffered the greatest burden of deprivation within the caste system (43). Scheduled tribes include approximately 700 officially recognized social groups that have historically been geographically and socially isolated from the rest of India (44). Other backward class is a legislatively defined group representing those who have historically suffered significant deprivation, but not as severe as scheduled castes and tribes. General class is a residual category traditionally viewed as better-off than the above groupings.
Education of the woman was defined in terms of the following categories, each representing significant milestones in the formal Indian education system: 0 years, 1–5 years, 6–8 years, 9–10 years, 11–12 years, or 13 or more years. Living standard, defined in terms of material possessions, is a reliable and valid measure of household material well-being or wealth (45). Each person was assigned a living standard score based on a combination of the responses for 19 different household characteristics, such as the quality of the home and ownership of a bicycle, that were weighted according to a factor analysis procedure (46). The analytic models used quintiles of these weighted scores. Employment was classified according to whether the woman was not working or was working in a manual, nonmanual, or agricultural profession. Information from the 1991 Indian National Census was used to create categories defining whether each neighborhood was in an urban area of over 1 million people (large city), an urban area of between 100,000 and 1 million people (small city), an urban area of less than 100,000 people (town), or a rural area (village).
A binary illness variable measured whether a woman suffered from one or more of the following conditions: tuberculosis currently, malaria in the previous 3 months, or jaundice in the previous 12 months. A binary breastfeeding variable measured whether the woman was currently breastfeeding a child. Number of children born included all livebirths and was categorized as none, one, two, three, or four or more. Decision-making autonomy was a categorical variable indicating the number of types of family decisions a woman participated in making, including what to cook, whether to obtain health care, whether to purchase jewelry or other household items, and whether to visit and stay with her relatives (47).
In the analyses of the children's data set, we included a number of covariates analogous to those in the women's data set: rural/urban residence, religion, caste, maternal education, maternal employment, living standard, and maternal decision-making autonomy. The analyses additionally included gender, age, maternal age at birth, birth order, and recent illness. A binary gender variable distinguished boys and girls. Child age was measured in 6-month categories. Maternal age at birth was categorized as 12–19, 20–29, 30–39, and 40–49 years. Birth order determined whether the child was the first, second, third, or greater than third livebirth from the same mother. A binary illness variable indicated whether the mother reported that the child had had a fever, cough, or diarrhea in the previous 2 weeks.
To adjust for clustering of observations at multiple levels, data were analyzed by using logistic multilevel modeling procedures (48), the strengths and relevance of which are well described (49–51). We specified a three-level model for each binary response (y, severely anemic or not) for individual i living in neighborhood j in state k of the form, πijk : yijk ~ Bernoulli (1, πijk). The probability πijk was related to a set of predictors, X, and a random effect for each level, by a logit-link function as Logit (πijk) = log [πijk/(1 − πijk)] = β0 + βX + u0jk + v0k. The linear predictor on the right-hand side of the equation consisted of a fixed part (β0 + βX) estimating the conditional coefficients for the exposure variable (and covariates) and two random intercepts attributable to neighborhoods (u0jk) and states (v0k), with each assumed to have an independent and identical distribution and variance estimated at each level. All models were created by using penalized quasi-likelihood approximation with second-order Taylor linearization as implemented in MLwiN 2.02 software (52, 53).
Study details were approved by the Harvard School of Public Health Institutional Review Board. Data collection, including informed consent, was performed by the International Institute of Population Sciences under the direction and supervision of staff of Measure DHS, a project overseen by the US Agency for International Development, in accordance with principles outlined in the Declaration of Helsinki (54).
Approximately 19 percent of women reported domestic violence, and 19.8 percent of the children had mothers who reported domestic violence (table 1). Women suffered high rates of malnutrition, with nearly half having anemia (severe: 1.4 percent) and nearly a third being underweight (severely: 5.6 percent). Children also experienced high levels of malnourishment: 70 percent had anemia (severe: 4.8 percent), over half were stunted (severely: 33.6 percent), 42.7 percent were wasted (severely: 17.3 percent), 16.2 percent were underweight for age (severely: 5.0 percent), and 11.9 percent had low body mass index for age (severely: 4.0 percent).
In unadjusted models, women who reported more than one instance of domestic violence in the previous year had a substantially increased likelihood of having anemia (odds ratio (OR) = 1.22, 95 percent confidence interval (CI): 1.14, 1.30; p for trend < 0.0001) and severe anemia (OR = 1.54, 95 percent CI: 1.25, 1.91; p for trend < 0.0001). Although adjustment for demographic variables attenuated these associations, reporting more than one instance of domestic violence in the previous year was still found to be strongly associated with anemia (OR = 1.11, 95 percent CI: 1.04, 1.18; p for trend = 0.01) and severe anemia (OR = 1.27, 95 percent CI: 1.02, 1.57; p for trend = 0.04) (table 2). In unadjusted models, women who reported more than one instance of domestic violence in the previous year also had an increased likelihood of being underweight (OR = 1.51, 95 percent CI: 1.41, 1.61; p for trend < 0.0001) and severely underweight (OR = 1.48, 95 percent CI: 1.31, 1.67; p for trend < 0.0001). Adjustment for demographic variables attenuated the strength of the associations, but they remained strong for both underweight (OR = 1.21, 95 percent CI: 1.13, 1.29; p for trend < 0.0001) and severe underweight (OR = 1.20, 95 percent CI: 1.06, 1.35; p for trend < 0.001) (table 2).
The associations between domestic violence and nearly all nutritional outcomes were in the same direction and of the same approximate strength for children as they were for women (table 3). In unadjusted models, maternal report of multiple instances of domestic violence in the previous year was found to be associated with anemia (OR = 1.27, 95 percent CI: 1.08, 1.48; p for trend = 0.001) and severe anemia (OR = 1.50, 95 percent CI: 1.17, 1.93; p for trend = 0.002), but adjustment reduced these associations substantially (OR = 1.06, 95 percent CI: 0.91, 1.24; p for trend = 0.37 and OR = 1.26, 95 percent CI: 0.97, 1.63; p for trend = 0.12, respectively). In unadjusted models, associations of maternal report of multiple instances of domestic violence in the previous year with all anthropometric measures of nutritional status were statistically significant. Although adjusting for demographic characteristics reduced the strength of the associations between maternal report of multiple instances of domestic violence in the previous year and all anthropometric measures of malnutrition among children, the relation remained marked for wasting (OR = 1.18, 95 percent CI: 1.03, 1.35; p for trend = 0.02), stunting (OR = 1.14, 95 percent CI: 0.99, 1.31; p for trend = 0.04), severely underweight for age (OR = 1.34, 95 percent CI: 1.05, 1.70; p for trend = 0.01), low body mass index for age (OR = 1.21, 95 percent CI: 1.01, 1.46; p for trend = 0.08), and severely low body mass index for age (OR = 1.50, 95 percent CI: 1.15, 1.95; p for trend = 0.0006).
Additional analyses were performed by using data on children less than 12 months of age. However, none of these associations approached traditional levels of statistical significance.
The results from this study of two large, nationally representative data sets provide evidence for a relation between domestic violence and malnutrition among women and children in India. These findings persisted, even after adjusting for geographic variation and a number of demographic characteristics, including multiple domains of socioeconomic status, that are strong confounders of domestic violence and chronic malnourishment. These findings also exhibited a dose-response relation, with frequent and recent abuse found to be associated with the highest probability of poor nutritional outcomes, particularly among women. To put our findings in context, a review of studies investigating the effect of administration of intestinal anthelmintic drugs on hemoglobin found that mass treatment of communities for intestinal parasites decreases the prevalence of anemia by 1.1–12.4 percent in adults (55), indicating that preventing domestic violence could be just as effective as this pharmaceutical approach in combating anemia among women.
We offer two explanations for the positive association between exposure to domestic violence and malnutrition among women and children. The first relates to empowerment. Perpetrators of domestic violence often use several types of abuse, such as physical and psychological, to control behavior of their family members (56). The withholding of food is a documented form of abuse in Indian households and is likely correlated with the perpetration of physical violence (20). An inadequate diet resulting from this withholding of food through psychological or emotional abuse could mediate the relation between physical domestic violence and nutrient deficiencies that cause anemia and underweight. Additionally, domestic violence is strongly associated with a woman's inability to make decisions for herself and her family, including the choice of types and quantities of food that a woman prepares as she cares for herself and her children (36). This might also explain why no significant associations were found between maternal report of domestic violence and nutritional outcomes in children less than 12 months of age, since a majority of these children in India are breastfed (57), and thus their feeding is more closely controlled by their mothers. Although a measure of decision-making autonomy was included in our models, this measure of autonomy does not have a temporal component. That is, it may be that many women who reported participation in family decision making exercise this power at times but only at the will of their husbands, and that this participation in decision making can be curtailed as a form of abuse.
The second explanation is that the link between domestic violence and nutritional deficiencies may also involve a mediating effect of psychological stress. Women who experience domestic violence tend to have higher levels of psychological stress (58)—a relation that has been documented among south Asian women (12, 59). Children who have witnessed domestic violence in their homes are also more likely to experience psychological stress (14, 60, 61).
There are physiological reasons why psychological stress may be associated with anemia. Psychological stress is a risk factor for oxidative stress (15–18), a term describing a number of chemical reactions that produce free radicals and other organic molecules capable of damaging living tissue (62). This oxidative stress can destroy red blood cells prematurely, acting as a potential cause of hemolytic anemia (63, 64). Additionally, although acute stressful incidents temporarily increase blood levels of both hemoglobin and erythrocytes, chronic stress has been found to result in long-term reductions, suggesting that stress interferes with protein synthesis required to create new red blood cells (65).
Psychological stress could also be linked to underweight status. Most research in food-rich environments in industrialized nations has documented a link between chronic stress and obesity mediated by metabolic changes (66) or behavioral adjustments (67) that promote the depositing of abdominal adipose. However, chronic stress also increases metabolic rate and energy expenditure (19, 68, 69), a process that could cause weight loss in a person with a fixed and limited caloric intake. That psychological stress may be associated with underweight in a low-income country is consistent with the results of a longitudinal study from the United Kingdom finding that, under stressful conditions, lean subjects tend to lose weight while overweight subjects tend to gain weight (70).
We note several important caveats to the findings presented here. First, although psychological, emotional, and sexual abuse constitute a substantial proportion of all domestic violence incidents (56, 71, 72), this information was not available in the current study. Our results present the relation between malnutrition and physical domestic violence, but it could be that other forms of abuse have different associations. Another caveat involves the measurement of domestic violence with a single global question. Previous research has found that behaviorally specific questions asking, for example, whether a woman has been slapped, punched, or kicked result in higher rates of reported domestic violence than the global question used in this survey (73). A local study from Karnataka state also found that over half of all women would keep quiet about it if abused (74). In addition, our analysis investigated only physical domestic violence and could not address other forms of abuse. These circumstances may account for the fact that the prevalence of domestic violence reported in this sample is lower than that previously found in India (12). However, by adjusting our models for a number of important socioeconomic and demographic variables including wealth, education, occupation, religion, and caste, we worked to limit the influence of bias introduced by social patterning of reliability of domestic violence reports on our results. A third caveat relates to the cross-sectional nature of the data. Although we attempted to address the idea of causal direction as best we could with the temporal reporting of domestic violence, it still was not possible to distinguish whether any reported abuse preceded attainment of the current nutritional status. Longitudinal research is called for to untangle this question.
In conclusion, domestic violence adversely impacts numerous health outcomes in India, including gynecologic morbidity (75, 76), contraceptive use (77), asthma (78), tobacco use (79), attempted suicide (13), and early child mortality (80). To our knowledge, this is the first large, nationally representative study using objective measures of malnutrition and measurement of multiple covariates to show a robust association between domestic violence and the likelihood of malnutrition among women and children. Additional adverse health effects of domestic violence suggest that public policy interventions to reduce domestic violence not only are fundamentally important from a moral and intrinsic perspective (81) but that there is a clear case to focus on reducing domestic violence given the instrumental health benefits that are likely to accrue.
S. V. Subramanian is supported by the National Institutes of Health Career Development Award (NHLBI 1 K25 HL081275).
The authors acknowledge the support of Macro International Inc. (Washington, DC) for providing access to the data. They also acknowledge Amy Cohen for her invaluable assistance with programming and data management.
Leland Ackerson led the analysis, interpretation, and writing of the manuscript. S. V. Subramanian conceived the study and contributed to analysis, interpretation, and writing of the manuscript. Both authors reviewed and approved the final manuscript.
Conflict of interest: none declared.