Intimate partner violence (IPV) is defined by WHO as any acts of physical, sexual and emotional abuse by a current or former partner whether cohabiting or not . Other terms used in the scientific literature are domestic violence, battering or wife abuse .
Design and sample
This study was conducted in 2002 within the framework of the demographic surveillance site in Bavi District, Ha Tay Province, called FilaBavi, situated in northern rural Vietnam .
In Bavi district, agricultural production and livestock breeding are the main economic activities of the local people (81%) . Illiteracy is low (0.4%), but higher among women than among men. About 70% of the adult population has completed primary school, 21% secondary level, 9% high school and 0.6% higher education. The higher the educational level, the lower is the number of females .
FilaBavi consists of a cohort of approx. 50.000 individuals (69 clusters) set together through a stratified cluster sampling procedure from the 240.000 individuals living in the district. For this study, 37 clusters were selected through a random cluster sampling technique. A number of households were selected from each cluster, proportional to the total number of households in each cluster. Eligible for the study were married or partnered women aged 17 to 60 years.
Face-to-face interviewing was used for data collection. The 39 female interviewers and six field supervisors engaged in the regular FilaBavi data collections were trained by the principal investigator in how to manage the specific challenges and difficulties encountered in studies on violence. As IPV might generate feelings of insecurity and frustration also among the interviewers, a pilot study was performed and the interviewers were encouraged to renounce participation if not feeling comfortable, but no one did.
Based on power calculations to detect stable significant risks of IPV, a total of 884 households containing a married or partnered woman were randomly selected for participation. Of these, 867 were currently married, and 16 were in a stable sexual relationship with a man, but not married. Only one woman declined to participate due to psychiatric illness. The participating 883 women all completed the interview and are henceforth referred to as married women.
The data collection instrument used was the Multi-country Study on Women's Health and Life Experiences Questionnaire developed by the World Health Organization (WHO) for studies within public health with focus on interpersonal violence . The questionnaire was developed for use in different cultures and is considered to be cross-culturally appropriate. The abuse questions were developed on the basis of a variety of other abuse assessment scales (Index of Spouse Abuse and the Conflict Tactics Scales) with established reliability and construct validity [20, 21]. This instrument was revised and translated into Vietnamese. The revisions made consisted of selected sections and items being removed as this data was either obtained from the Filabavi database (socio-demographic data)  or considered inappropriate in the Vietnamese context (dowry related items). In a one-day seminar and a pilot interview, the questionnaire was further validated through a review panel process where each item was considered for appropriateness.
Only women took direct part in this study and data related to husbands/partners were obtained from the participating women.
Violence occurrence was assessed by types (physical, psychological and sexual abuse), timing (life-time and past year exposure) and frequency (how often it occurred). Physical abuse was assessed by 11 items: slapping, throwing things, pushing or shoving which were classified as moderate physical abuse behaviours. Further hitting, kicking, dragging, beating, choking, burning and threatening with or using a weapon (knife, scissors or object) were classified as severe physical violence . Sexual abuse was assessed by three items: having sexual intercourse against the respondent's will, using physical force for sexual intercourse, and forcing the respondent to sexually degrading acts. Psychological abuse was assessed by four items: insults or degrading activities, belittlement or humiliation, scaring the respondent on purpose including threats of violence.
Two dependent variables were created, physical and sexual violence combined and pure psychological abuse. Physical and sexual violence was defined as the respondent being subjected to any act of physical or sexual violence or both (henceforth referred to as physical/sexual violence); psychological abuse was defined as being subjected to any item of psychological abuse without overlap of any other kind of violence (referred to as psychological abuse alone) .
Lifetime occurrence of any kind of violence was defined as experience of any act of violence to date of the interview from a current or former husband/partner. Abuse taking place within the past year was defined as any act taking place within the past 12 months. For bi- and multivariate analyses the dependent variables were dichotomised into experience of violence as opposed to no experience of violence. For these analyses, those with only one single experience of violence over the lifetime were considered as non-exposed, to strengthen the criterion for violence exposure.
Socio-demographic and psychosocial variables were tried as independent risk factors. Age was divided into three groups. Educational attainment was grouped into primary (5 years) and secondary schooling (9 years) and higher education (>9 years) respectively, and dichotomised with higher education as the reference category. Annual household income was divided into quintiles and later into three groups (lowest income group, < 288 USD, low and middle income groups, from 288 – 570 USD and high and highest income groups > 570 USD) and further dichotomised for the multivariate analyses whereby a household income in the lowest income group (lowest and low income groups, < 425 USD) was treated as the exposure category. Husband's working specifics was also grouped into three categories and dichotomised into professionals as the reference and semi-skilled and unskilled combined as the exposure group.
Data were double entered into the Statistical Package for the Social Sciences (SPSS) version 10.0 which was used for all statistical purposes . Risk ratios were estimated by odds ratios (OR). Statistical significance was determined at the 95 percent confidence interval level.
Bi- and multivariate analyses were adjusted for age apart from when age differences were investigated. For the multivariate analyses, variables of theoretical and empirical (statistically significant in the bivariate analyses) interest were entered one by one in a stepwise fashion. To avoid a correlation effect, the multiple logistic regression models included only items with correlation coefficients below 0.4 .
The World Health Organisation has issued guidelines for violence research  and these were strictly followed. Interviews were held in strict privacy, mainly in the respondents' homes, with no one able to overhear the conversation. In a few cases when privacy was not possible to establish, the interview was performed at a nearby community health centre. The participants were informed about their possibility to withdraw at any point during the research phase and gave written informed consent to participate. This study was approved by the institutional review board of Gothenburg University, Sweden; Hanoi Medical University and Ministry of Health, Hanoi and Bavi district People's Committee and Bavi District Health Center, Hatay province, Vietnam.