Intimate partner violence against women during and after pregnancy: a cross-sectional study in Mumbai slums

IPV in India

Although population-based surveys underline the ubiquity of IPV, its occurrence and impacts are frequently hidden and the figures are usually underestimates [5]. India’s third National Family Health Survey (NFHS-3 2005–2006) suggested that 33% of ever-married women aged 15–49 years had faced IPV at some point over the age of 15 [6]. A second large cross-sectional survey involving over 14 000 women in 18 states yielded an estimate of 39% [7]. Other community-based surveys suggest a prevalence of physical IPV of 35-60% [8–10].

IPV in urban India

A secondary analysis of NFHS-3 data suggested that 28% of ever-married urban women aged 15–49 years had experienced physical violence in the last year [11]. There is less information on IPV in the slums that characterize India’s cities. The NFHS-3 estimated the lifetime prevalence against ever-married women at 23% in Mumbai’s slum population [6]. The figure was 54% in a Kolkata slum [12]. Over shorter recall periods, 37% of women in a Mumbai slum survey reported violence in the preceding year (verbal 32%, physical 23%, and sexual 9%) [13], compared with 62% in Pune [14], 17% in Kolkata [15], and 27% in Bangalore in the preceding six months [16].

IPV during maternity

Pregnancy does not protect women from IPV [17–19]. A global systematic review described a prevalence during pregnancy of 1-20% [20], a review of studies in Asian countries 4-48% [21], and a review of Indian studies 21-28% [22].

Consequences of IPV during maternity

IPV leads to both acute injuries and profound longer-term challenges to health and wellbeing [23, 24]. Studies from India suggest that women who experience physical IPV during pregnancy are more likely to suffer depression and mental ill-health [25]. They are more likely to miscarry, have premature labor [26], and deliver low birth weight infants [27, 28]. The survival of their infants is reduced in the perinatal [29, 30], neonatal [29, 31], infant [32], and childhood periods [33]. Women who report IPV are also less likely to have prenatal care [34, 35], may be more likely to terminate pregnancy [36], and are at greater risk of sexually transmitted infection [37, 38].

Risk factors for IPV

Figure 1 categorizes factors available in our dataset that have been described as potential determinants of IPV in India and included in recent reviews [35, 39]. Each level interacts with others. For example, improvements in women’s education may lead to changes in societal attitudes to IPV, and have a bearing on employment and socioeconomic status. At community level, IPV appears to be less common in urban than in rural environments [11, 40]. Although it is difficult to quantify, several studies have suggested that an absence of social support might make women more vulnerable to IPV [10, 41, 42].

At the household level, there is a consensus that poverty increases the risk of IPV. This is seen in Indian studies of IPV in general [10, 43–48], in pregnancy [7, 11], and in slum communities [12, 15]. IPV has been described as more likely to be experienced by women from lower castes [43], larger families, and certain faith groups [7, 11, 18, 49]. There is some evidence that women who have witnessed abuse in their families are more likely to experience it themselves [10, 45].

Having more educated partners appears to protect women against IPV [42, 50]. The same is true of employment, in that women whose partners are in work are less likely to suffer IPV. Spousal alcohol or drug use is a risk factor in many studies [10, 41–43, 47, 51, 52], including those that consider IPV in pregnancy [7, 11, 13], and in slum communities [12, 15]. Risk-taking behavior such as spousal gambling [7] and extramarital sex increase the risk of IPV [15, 53].

Women who marry younger (under 18 years in most studies) are at greater risk of IPV [11, 54–56], although this finding is not always replicated [57] and older women are also at increased risk [13, 46, 47]. Longer marriages may be associated with increased [15] or decreased risk [7]. As with their partners, women with less education are at greater risk of IPV [7, 11, 13, 18, 27, 45–47, 50, 54]. It has been suggested that women who are more educated than their husbands are at greater risk [50], but this has not been substantiated and a recent study found the opposite [58]. The effects of other risk factors are unclear, particular examples in the Indian context being son preference and women’s employment. We discuss these later.

SNEHA (Society for Nutrition, Education and Health Action) is a non-government organization working to improve the health of women and children in Mumbai’s disadvantaged settlements, in four domains: prevention of violence against women and children, reproductive health, maternal and newborn health, and childhood nutrition. We offer a range of counseling, legal, and liaison services for women and children experiencing violence. Despite our work in the area for over a decade, we had little information on the prevalence of IPV within slum communities. We had also become concerned about the impact of IPV on women’s experience of maternity. During a community-based program that focused on maternal and newborn health [59], we interviewed women after childbirth. We included a series of questions about violence within a larger questionnaire. Our objectives were to describe the prevalence of IPV during pregnancy and postpartum in an urban slum setting, to review potential risk factors in the urban context, and to explore its effects on maternal and newborn health.

Setting

The capital of Maharashtra state, Mumbai has 12.4 million residents. Although 63% of households fall into India’s highest wealth quintile [6], 53% of residents live in slums [60]. Indicators of women’s status such as literacy, workforce participation, age at marriage, and healthcare seeking tend to be better than the national averages. We used data collected during a cluster randomized controlled trial of community mobilization to improve maternal and newborn health [59, 61]. The trial involved 48 slum clusters, each of ~1200 households, covering a population of ~283 000. 26% of homes were of insubstantial fabric (corrugated iron, planking, tarpaulin) and sizeable proportions did not have metered electricity (28%), access to individual or communal piped water (21%), or individual toilet facilities (94%). More than one-third of clusters were adjacent to hazards such as railway lines, garbage dumps and polluted bodies of water.

Data collection

A registration system monitored live births, stillbirths, maternal and neonatal deaths in all 48 slum clusters. Births were identified by 99 locally resident women, covering ~600 households each. After confirming a birth, one of 12 researchers visited mothers 6 weeks later for a postnatal interview. The interview included questions on demographic characteristics, socioeconomic indicators, maternal and newborn care, and was framed as a women’s health study with a particular focus on maternal and newborn health. After an explanation of the data collection activities, participants were asked for verbal consent to interview and were assured of data confidentiality. Completed interviews were checked by supervisors and project officers, both systematically and through random visits, and were entered in a database in Microsoft Access (Microsoft Corporation). Information provided by the participants remained confidential and no analyses or outputs included their names. Access to data was restricted to the analysts and was password protected.

Because information on the epidemiology of IPV is limited, particularly at disaggregated urban levels, we added a module to the routine questionnaire. All women who gave birth from March to September 2009 and consented to interview received the module, which included questions about justifications for IPV, instances of physical, emotional, or sexual IPV in the preceding year, injuries, help seeking, and spousal drug and alcohol use. The questions were adapted from existing versions in the Demographic and Health Surveys [6].

We followed World Health Organization ethical and safety recommendations [62]. Participants were told that the study was about maternal and newborn health, and that women in the study area who had given birth in a certain period would be approached for interview. During the interview, participants were informed of the nature and potential sensitivity of the IPV module and consent was taken before administering it. One woman per household received the module. Most of the interviews were conducted in participants’ homes. The researchers made efforts to interview women alone, but the density of slum homes and a desire to make respondents comfortable meant that there were limits to their ability to achieve privacy. At the first visit, if family members were present and the researcher was concerned about privacy, she asked the respondent for a suitable time to make a second visit. If family members were present again, information about the maternal and newborn health study was shared with them. The researcher told them that some questions might be embarrassing for the participant to answer in the presence of others. The interview began with the routine questionnaire and, when family members were reluctant to leave, the researcher encouraged them to stay for a while to reassure themselves about the nature of the questions. Most family members were generally amenable to leaving before the IPV module. The researchers carried a resource list of support services near the study area. They visited providers and informed them about the study, and that their help might be needed. All participants received information on resources, irrespective of whether or not they reported violence. They were told that the information could help them to support other women if they suffered violence, and referral was arranged if requested.

Given available estimates of the prevalence of IPV, a sample size of 1800 would give an estimated proportion with a precision of 5% at a confidence level of 95%. Recruitment was planned to be sequential until 300 questionnaires had been collected in each of the six municipal wards across which the 48 clusters were distributed. We assumed a high attrition rate because of the sensitive nature of the questions, and interviewed more women than the required sample size.

Variables

We defined physical IPV according to the WHO Multi-Country Study [63], as an instance in which her partner beat, punched, kicked, dragged, or slapped a woman, twisted her arm, pulled her hair, hit her with an object, choked, burned, or physically restrained her. Emotional IPV was defined as an instance in which a spouse showed jealousy, humiliated his partner in front of others, accused her of infidelity, threatened to evict her or actually did so, threatened her or her children with violence, or forcibly took something from her. Sexual IPV was defined as an instance in which a woman was forced to have sex with her partner when she did not want to. Respondents were asked the questions about IPV in the preceding year. The timing of the interview meant that the preceding year encompassed less than two months before the pregnancy, pregnancy itself, and the postpartum period. For each type of IPV, a subsidiary question on frequency followed (once, sometimes, often). We also asked respondents who reported physical IPV whether it had happened more before, during or after pregnancy. We asked about injuries or complications that might have resulted from it (cuts, bruises, pain, burns, wounds, fractures and dental trauma, fetal death, premature labour, vaginal bleeding), and whether it led to problems (self-care, infant care) such as prevention of breastfeeding. We asked whether respondents had sought help from family members, friends or neighbours, local or religious bodies, women’s or other non-government organizations, doctors, the police, lawyers or a helpline.

Socioeconomic status was described by asset scores assigned to respondents on the basis of standardized weights for the first component of a principal components analysis, ordered and divided into quintiles [64, 65]. Assets included a range of consumer durables and house ownership and construction. The remaining variables included miscarriage (cessation of pregnancy before 22 completed weeks) in a previous pregnancy, any reported illness during the index pregnancy, receipt of any prenatal care, home delivery, preterm delivery before 37 completed weeks of gestation, low birth weight (< 2500 g), infant sex, and use of a modern family planning method in the three years preceding the index pregnancy (oral contraceptive pill, condoms, intrauterine device, injectable or implantable contraception).

Statistical analysis

We summarized responses to questions about IPV with frequencies and percentages, categorized according to type of IPV. Analysis of determinants followed the conceptual framework in Figure 1. We hypothesized that the odds of IPV during maternity would decrease with rising socioeconomic status, woman’s age, age at marriage, parity, and education, and with her husband’s education and employment; would increase if her husband used alcohol or drugs; and would be affected in an unspecified direction by family structure and religion. We entered IPV as a binary dependent variable in multivariable logistic regression models with a random effect for cluster. Hypothetical determinants were entered as independent variables in both univariable and multivariable models [66]. At household level, socioeconomic quintile, family unit, and religion were entered as categorical indicator variables. At woman level, age-group, schooling, age-group at marriage, and parity were entered as categorical indicator variables, and employment as a binary variable. At partner level, husband’s schooling was entered as a categorical indicator variable, and employment and alcohol use as binary variables. We did not expect the intervention under test in the trial in which the study was nested to affect IPV. Including a covariate for allocation in the analysis made no substantial difference to the findings, and it has not been included in the analyses presented.

In an additional analysis, we hypothesized that having a female baby might increase the odds of IPV, during pregnancy if the infant sex was known, or postpartum. We therefore added an independent variable for infant sex to a model including the covariates already mentioned. We also hypothesized that IPV might be more likely to occur in an environment in which a woman had had a previous miscarriage. Again, we added an independent miscarriage variable to the existing model.

Ethical approval

Data for the study originated from a larger process of data collection approved by the Municipal Corporation of Greater Mumbai and the Independent Ethics Committee for Research on Human Subjects (Mumbai committee, reference IEC/06/31).