The study design was based on a methodology previously used in other conflict settings . A two-stage cluster design was executed in two districts in the Indian part of Kashmir (Kupwara and Badgam). These districts were chosen because MSF intended to start working there, an operational decision based on anecdotal evidence of mental health problems among populations living in these areas. The districts have a combined population of 145,000 residents living in 101 villages (3750 square kilometres). The predominantly Muslim, rural and indigenous population of these districts do not differ from other districts in Kashmir except for the capital, Srinagar. Both districts are close to the Line of Control and have experienced high numbers of violent incidents, although to what degree the level of violence differs from other districts is unknown due to lack of reliable information.
For the calculation of sample size we assumed a prevalence of trauma-related psychological problems of 20% , and using a precision of 5% (confidence interval 95%) and a design effect of 2, the minimum sample size was estimated at 492. A two-stage cluster sampling design was used to cover 30 villages, resulting in 17 randomly selected households per village. Research teams started at the centre of the village, spun a bottle, and began the interviews according to the direction in which the bottle pointed. The first encountered household was selected, after which the next household in the same direction was approached. Within the household the participant was also selected randomly.
Ethics and interview procedures
The survey was conducted over a period of eleven weeks, from 4 June 2005 to 16 August 2005 in Badgam and from 4 July 2005 to 18 August 2005 in Kupwara. The informed consent procedure consisted of two steps. In the first step the head or most senior adult present in each selected household was asked permission to interview a person over the age of 18 years. The purpose of study, guarantees of anonymity and confidentiality, the use of data (including public dissemination and scientific publication), and the possibility to withdraw from interview at any time was explained. It was made clear that no (financial) compensation was given. Written consent was then sought. The head of household assisted the interviewer in making a list of all household members and from this list one person (the respondent) above 18 years of age was selected randomly. If the selected person was not at home, another person in the household (>18 years) was selected. Step two of the interview process consisted of repeating the above introduction to the potential participant. Once written consent was given, the interview was conducted.
The survey team consisted of four senior national and expatriate staff that supervised 20 trained local interviewers. Interviews were done in pairs, each pair conducting two to three interviews each day. Each team consisted of both male and female interviewers and respondents could choose who did the interview. The average time for interviewing was 50–60 minutes. The interviewers were recruited from Srinagar University Department of Psychology and Sociology and received a salary for their work. Teams stopped their activities at any moment if they were worried about their own safety or that of the population or if they judged their activities to be counterproductive to the program (for instance, when security incidents such as strikes or 'Hartals' occurred, forcing the survey team to postpone the survey).
Interviewing people on traumatic experiences carries a risk of contributing to psychological distress of both interviewee and interviewer. To respond to this, one experienced counsellor supervised each survey team to give immediate (technical or emotional) support if required. Also, referral to MSF operated counselling centres in another location was offered to all interviewees and interviewers (although none were referred).
To manage potential overwhelming emotions among the interviewer, staff training was given in communication and handling of difficult or upsetting situations. Staff were debriefed daily for both technical and emotional issues. For those interviewers who were overwhelmed or needed follow-up support counselling services were available.
The study received ethics approval from MSF's independent Ethical Review Board.
The survey questionnaire was based on previous formats used in similar studies elsewhere  and focussed on the following four subjects: baseline demographics, confrontation with and consequences of violence, mental health, and sources of support. This paper focuses on the first two issues. Tools to assess mental health, and sources of support are described in a second paper .
We assessed confrontation with violence both since the beginning of the conflict and in the three months preceding the survey. Proximity to violence was defined as either exposure ('Being in the vicinity of a violent event but not witnessing or self-experiencing'), witnessing ('Witnessing an event so close it could have happened to you or you were forced to see it'), or self-experience ('The event happened to you'). Violence categories were based on a review of violent incidents as reported in newspaper articles (such as Kashmir Affairs, Greater Kashmir, and Jammu Kashmir) of the past two years and consultation with national staff. We used rape in the witnessing section and a broader concept of 'violation of modesty' in the self-experience section because national staff felt that interviewees would feel more comfortable with this term. Violation of modesty is the local equivalent for sexual violence and includes inappropriate touching, in accordance with the WHO's definition of sexual violence .
The survey was translated from English to Urdu and phonetic Kashmiri, then back-translated from Urdu and phonetic Kashmiri to English using a different translator. After revisions, the questionnaire was piloted in a community close to Srinagar. For the definition of the start of the conflict (1989), the definition of torture ('Unbearable physical pain deliberately inflicted by others who have complete control'), maltreatment ('cruel and inhumane treatment'), and round-up raids the local population and national staff were consulted. Examples of physical and mental maltreatment such as 'Being kicked at checkpoints', and 'For body searching males being forced to undress in front of their family' were discussed among interviewers, as were forced labour and violation of modesty.
Data entry was standardised and checked by supervisors. As an additional control, 5% of the forms were randomly checked. Data were entered in an EXCEL program spreadsheet and exported into EPIINFO-2002 for analysis. Previous studies have consistently shown gender to be a risk factor for developing psychological problems (most notably post-traumatic stress disorder) after exposure to traumatic events [9, 10]. Analysis of our data also revealed gender as a confounder for many variables. Therefore we stratified results by gender (see Tables).