Three cross-sectional surveys were undertaken, one in each of three geographic areas supported by MSF. Both IDP camps and local villages were included.
Setting
The three sites--Kabizo, Masisi, and Kitchanga--all lie to the northwest of Goma, the capital of North Kivu. The sites are geographically close (less than 100 km between sites), but all have distinct histories. All sites were affected by the renewed fighting as of September 2008. Kabizo town was directly attacked, with the population fleeing to a nearby town and pre-existing IDP camp. In Masisi, the town itself was not the scene of direct fighting, but clashes in the surrounding villages lead to continual displacement, returns, and the creation of new IDP camps. In Kitchanga, the population of a small town has been outnumbered by IDPs in two camps. Additionally, as the ongoing conflict continues in the region, more IDPs are settling further to the north of Kitchanga. In all sites, the MSF health programme, run in collaboration with the Ministry of Health, had been in place for at least 1 year prior to the most recent clashes.
Sample size calculation
Sample size was calculated to estimate a crude mortality rate of 0.7 deaths/10 000 people/day over a period of 8 months, with a precision of 0.3. The point estimate was based on the results of a rapid assessment carried out in another area of North Kivu in February 2009. Assuming an average household size of 5, a minimum of 249 households were required for each survey using systematic sampling. Assuming the same household size and a design effect of 3, 747 households were required for each cluster sampling survey.
Survey procedures
Surveyors were given 2 days of theoretical and 1 day of practical training in the field prior to starting the survey. The three epidemiologists responsible for the surveys all participated in the first training to ensure consistency between sites.
In Kabizo town and camp, where the households were arranged in a semi-regular pattern, systematic sampling was used, with the sampling interval calculated based on the most recent population estimates available. Town population data were obtained from local authorities (2008 census). NGOs provided 2009 population data for IDP camps. The sampling interval (x) was calculated by dividing the total estimated number of households by the required number of households based on the sample size calculation. The first household in each camp or town section was selected by choosing a random number between 1 and x. The sampling interval was then added to this random number to select the next household with the procedure repeated until the end of the section. Local outreach workers were used as guides to ensure that no houses were missed and that section boundaries were respected. One village was excluded from the survey for security reasons.
In Masisi and Kitchanga, two-stage cluster sampling was used following the standard World Health Organisation (WHO) Expanded Programme on Immunization cluster sampling proximity method[7]. Probability of selection was proportional to population size based on the smallest possible geographic area (village or camp section) and the most recent census data (Masisi January 2007 census with 3.0% growth adjustment, Kitchanga 2009 NGO estimates). In Masisi, any area with active fighting or where distance would have required the surveyors to remain in the village overnight was excluded.
In Masisi, the initial household in each cluster was selected using the EPI 2 method[8]. Because of the long distances to travel to reach each cluster site in Masisi, we estimated that 20 households could be interviewed per team per day and the cluster size was defined accordingly. In Kitchanga, the initial household in each cluster was the household closest to a randomly selected global positioning system (GPS) point from within the mapped cluster area. Travel to distant sites was facilitated by the set-up of a sub-base where teams could spend the night, thus reducing daily travel times. We estimated that 25 households could be interviewed per team per day, and the cluster size was defined accordingly.
A household was defined as a group of people under the responsibility of one person (head of household) and who share meals.
If a house was found empty, neighbours were asked to help establish whether the house was abandoned or inhabited, and if inhabited, to find absentees. Survey team members returned at least once to each house to try to find absentees. In the event of absence, refusal, or if no adult (older than 15 years) was present, the house was skipped and replaced with the next nearest house.
The same basic survey questionnaire was used in all three sites, with minor changes made to the recall period to adapt to local events. The questionnaire was written in French, translated into Swahili, and back-translated to French for verification. All surveyors spoke French, Swahili, and/or other local languages. The purpose of the survey was explained to heads of households and written consent (thumbprint or signature) obtained before beginning the interview. Data were collected on household members who had been present at the beginning of the recall period, including those who had since left, resident household members, and household members who had died during the recall period (including babies who were born and died during the period). Age and sex of all individuals, either currently alive or deceased during the recall period, were collected as well as dates of births, deaths, arrivals in, or departures from the household. Information on the reported cause of death was also collected. No detailed verbal autopsy or other verification was done for reported cause of death.
Data on violence against individuals, both the number and type of episode, were collected. If a household member had experienced an episode of violence, they were asked to give the nature of the violence (beating, rape, injury by firearm or other weapon, detention, abduction, forced labour, or extortion). One response could be given for each episode. Individuals were also asked when the episode occurred and to describe the perpetrator from a list of categories (combatant, "incivique" [a local term used for a deserter], civilian, unknown, no response). Each episode during the recall period was described separately.
In addition to individual data, data were collected at the household level on current status (displaced, returnee, resident), number of displacements during the recall period, reason for displacement, distance of displacement (measured in number of days walked), theft of essential household items (jerry can, cooking set, blanket, livestock, or food-stocks), and access to agricultural fields.
To estimate access to health care, information was collected on the last household member to be ill, if the illness occurred during the 2 weeks before the survey. Information on care seeking and payment for services was collected for these individuals.
Questionnaires were checked for completeness and accuracy at the end of each day. Data were entered in EpiData 3.0 (EpiData Association, Odense, Denmark) with 10% general data checks and all major events verified. STATA 10 (StataCorp, College Station, TX), EpiInfo 6.04 d (Centers for Disease Control and Prevention, Atlanta, USA and WHO, Geneva, Switzerland) and Emergency Nutrition Assessment (ENA) for SMART http://www.nutrisurvey.de/ena/ena.html were used for data analysis. For the Masisi and Kitchanga analyses, software or commands that estimate design effect and incorporate them in confidence interval calculations were used.
Ethics statement
Ethical approval was obtained from the Ethical Committee of the School of Public Health, Kinshasa, Democratic Republic of Congo.