Prevalence of domestic violence against women in informal settlements in Mumbai, India: a cross-sectional survey

Objectives Domestic violence against women harms individuals, families, communities and society. Perpetrated by intimate partners or other family members, its overlapping forms include physical, sexual and emotional violence, control and neglect. We aimed to describe the prevalence of these forms of violence and their perpetrators in informal settlements in Mumbai. Design Cross-sectional survey. Setting Two large urban informal settlement areas. Participants 5122 women aged 18–49 years. Primary and secondary outcome measures Prevalence and perpetrators in the last year of physical, sexual and emotional domestic violence, coercive control and neglect. For each of these forms of violence, responses to questions about individual acts and composite estimates. Results In the last year, 644 (13%) women had experienced physical domestic violence, 188 (4%) sexual violence and 963 (19%) emotional violence. Of ever-married women, 13% had experienced physical or sexual intimate partner violence in the last year. Most physical (87%) and sexual violence (99%) was done by partners, but emotional violence equally involved marital family members. All three forms of violence were more common if women were younger, in the lowest socioeconomic asset quintile or reported disability. 1816 women (35%) had experienced at least one instance of coercive control and 33% said that they were afraid of people in their home. 10% reported domestic neglect of their food, sleep, health or children’s health. Conclusions Domestic violence against women remains common in urban informal settlements. Physical and sexual violence were perpetrated mainly by intimate partners, but emotional violence was attributed equally to partners and marital family. More than one-third of women described controlling behaviours perpetrated by both intimate partners and marital family members. We emphasise the need to include the spectrum of perpetrators and forms of domestic violence—particularly emotional violence and coercive control—in data gathering. Trial registration number ISRCTN84502355; Pre-results.

An-Najah National University Nablus Palestine REVIEW RETURNED 01-Aug-2020

GENERAL COMMENTS
Overall; this is an important article in the field of GBV, shedding the light on special, and disadvantage population. Kindly find enclosed my comments; 1. Abstract; The authors indicated interviewing 100 women per 50 clusters, this would leave us with 5000 women being interviewed. It is unclear for me how the number become 5122. 2. Methods; Interviews were carried on household. As authors described, it is overcrowded residency, so how the safety of the women as well as researchers was granted 3. Results; We have added more information to the section on data collection: "Interviewers were all women and provided both time and sufficient information for women to consider whether to participate. They were supported by three field supervisors with direct linkage to counselling services, available by phone at any time. The interview team visited the local police station and social services to discuss their activities before starting in each area. Interviewers worked in groups of 7-8 in one cluster at a time, accompanied by a supervisor. A pair worked together in each lane and administered interviews in adjacent households.
To ensure privacy, interviews were arranged by advance appointment and avoided times when partners or children were likely to return from work or school. Women were interviewed at home or in a local community office if they preferred it. The interview began with general questions about demography, household residents, education, socioeconomic position, maternity, and health. If a family member, neighbour, or friend entered, the interviewer went back to asking questions about general health. If the person showed signs of staying, the interview was terminated and completed over up to three repeat visits. As a result of the gatekeeper consent process, community members were aware that interviewers would be visiting people in their area and this limited curiosity and intrusion." Discussion: Discussion section need revision. In second paragraph authors discussing about limitation. I suggest having separate section on study limitation before conclusion section.

Response
The STROBE guideline suggests placing limitations in the current position. We have added subheadings to structure the Discussion section according the STROBE guideline, with some additions: key results, limitations, (comparison with other studies), implications, generalisability. Fourth paragraph of discussion section saying page 11 line 46-48 stating "it is also possible that women answer questions differently when they are responding to a survey on violence than when the questions on violence are included in a broad survey like NFHS" I suggest authors to support this statement with evidence.

Response
The statement was based on the article by Saltzman, which reviews a number of sources of estimates of prevalence of intimate partner violence. We have made this clear by adding the following: "In a consideration of methodological issues in research on intimate partner violence, Saltzman suggests that questionnaire context and cues given to the participant can influence response (Saltzman, 2004)." Study should state some policy or programme implications of in the slums.

Response
We have rewritten the conclusion section: "We have two recommendations for research and action. First, social workers and healthcare providers should be aware of the importance of emotional violence and coercive control. Both are common and cause substantial suffering, particularly to women's mental health. These forms of violence need to be considered in interactions with women because of their subtlety and their intersection with poverty in urban informal settlements. Researchers should make sure that they are included in studies of domestic violence. Second, violence is often perpetrated by family members other than intimate partners. Again, this is particularly true of emotional violence and coercive control. In interacting with clients, social workers and healthcare providers need to be aware that family members accompanying them might be involved in abuse, and studies should assume the possibility of other perpetrators." Reviewer: 2 Abstract; The authors indicated interviewing 100 women per 50 clusters, this would leave us with 5000 women being interviewed. It is unclear for me how the number become 5122. Response Interviewers aimed to achieve 100 interviews per cluster, giving a total of 5000. As is usual with field data collection, which was divided among 16 people, there was some variation in the numbers achieved. A median 101 interviews were achieved in each cluster (interquartile range 100-103, range 94-118). We have added this to the Results section. Methods; Interviews were carried on household. As authors described, it is overcrowded residency, so how the safety of the women as well as researchers was granted Response We have added text on this to the section on data collection: "Interviewers were all women and provided both time and sufficient information for women to consider whether to participate. They were supported by three field supervisors with direct linkage to counselling services, available by phone at any time. The interview team visited the local police station and social services to discuss their activities before starting in each area. Interviewers worked in groups of 7-8 in one cluster at a time, accompanied by a supervisor. A pair worked together in each lane and administered interviews in adjacent households.
To ensure privacy, interviews were arranged by advance appointment and avoided times when partners or children were likely to return from work or school. Women were interviewed at home or in a local community office if they preferred it. The interview began with general questions about demography, household residents, education, socioeconomic position, maternity, and health. If a family member, neighbour, or friend entered, the interviewer went back to asking questions about general health. If the person showed signs of staying, the interview was terminated and completed over up to three repeat visits. As a result of the gatekeeper consent process, community members were aware that interviewers would be visiting people in their area and this limited curiosity and intrusion." And in the section on data management: "Mobile connectivity allowed immediate contact between interviewers and supervisors." And in the section on ethical considerations: "We made provision for storage of participant information sheets on women's behalf if they were concerned that the paperwork might be seen by others." "Participants were able to speak with counsellors immediately by phone. When a survivor was not ready to disclose violence, the interviewer provided her with information on available services and legal rights and gave her a small card that was easy to hide and listed essential contact numbers and addresses for 24-hour crisis support, medical emergencies, and the police. She took consent for any action from the participant herself." Results; We have restructured the discussion section and added subheadings to make the flow clear. Cluster effect was not discussed. It is unclear for me whether the authors had account for it during the analysis.

Response
The analysis used survey commands which accounted for cluster as the primary sampling unit and for strata describing broader settlements.
Effect of inter-observer variation was not discussed.

Response
We assessed inter-observer variation in supervisory meetings with interviewers. In line with good data monitoring practice, we examined variation in prevalence rates by interviewer and selected interviewers for field observation who showed early signs of deviation from the group average. We have added to the data collection section: "We examined variation in prevalence rates by interviewer and discussed performance in supervisory meetings. We selected for field observation interviewers who showed signs of deviation from the group average and provided feedback where necessary." The conclusion does not reflect on policy implication of the results Response We have added to the conclusion: "We have two recommendations for research and action. First, social workers and healthcare providers should be aware of the importance of emotional violence and coercive control. Both are common and cause substantial suffering, particularly to women's mental health. These forms of violence need to be considered in interactions with women because of their subtlety and their intersection with poverty in urban informal settlements. Researchers should make sure that they are included in studies of domestic violence. Second, violence is often perpetrated by family members other than intimate partners. Again, this is particularly true of emotional violence and coercive control. In interacting with clients, social workers and healthcare providers need to be aware that family members accompanying them might be involved in abuse, and studies should assume the possibility of other perpetrators."

GENERAL COMMENTS
I appreciate authors for making substantial changes to bring out clarity. The manuscript may be accepted for publication.

VERSION 2 -AUTHOR RESPONSE
Thank you for accepting the paper for publication. As far as we can see, the reviewer did not suggest any revisions.