Study site and population
The study was carried out in 10 communities (~20,000 people) situated in the hilly forested areas of the Bac Ai and Ninh Son districts of Ninh Thuan Province in south-central Vietnam in the framework of a community-based, cluster randomized trial assessing the effectiveness of long-lasting insecticidal hammocks (LLIH) in controlling forest malaria . Malaria transmission is perennial with two peaks (May-June and October-November), at the start and at the end of the rainy season. In a survey carried out in November 2003, the overall parasite rate was 13% (up to 40% in some villages) and the prevalence of anti-malarial antibodies 37% (up to 75% in some villages) . Malaria control is based on early diagnosis and treatment with artemisinin-based combination therapy and the provision of insecticide-treated nets (ITNs), distributed free-of-charge and insecticide-treated twice a year by the national malaria control programme. Reported ITN use in the study area was high (86.3%, in April 2004) with a median coverage of 2.5 people per bed net . Similar to other forested areas of central Vietnam, despite intensive efforts, malaria remains difficult to control due to the complex interactions between humans, vectors and environmental factors .
The Ra-glai, or 'people of the forest' in Ra-glai and Cham languages [13, 14], are of Malayo-Polynesian descent and represented 86% of the study population . Linguistic evidence suggests that they are possibly descendents of coastal Cham that were historically pushed inland to the forested and mountainous regions [15, 16]. Today, the Ra-glai are a largely impoverished ethnic minority in Vietnam, almost exclusively dedicated to small-scale subsistence slash and burn agriculture in fields located in the surrounding forests. According to a survey carried out in 2003, 80% of the active population can be categorized as 'forest worker', 99.5% of which participate in slash and burn agriculture, occasionally combined with hunting, gathering of forest products and logging . The Ra-glai's heavy dependence on the forest for subsistence activities places them at greater risk for malaria infection, a risk that is further increased by staying overnight in the forest [12, 17]. Indeed the main malaria vector Anopheles dirus sensu stricto, is a sylvatic and highly anthropophilic species, resting mainly outdoors and biting early in the evening .
The implemented research strategy was based on methodological triangulation (the combination of various methods and data sources to refine and test particular interpretations and hypotheses), combining both qualitative and quantitative methods in order to limit bias and enhance the strengths of the respective research methods [19, 20]. The combination of both qualitative and quantitative methods during fieldwork allows for the confirmation of specific patterns while also facilitating the detection of new and unexpected variables. Methodological triangulation has the additional advantage of being able to contextualize trial data through qualitative sampling techniques and data analysis, and of facilitating the quantification of qualitative variables from focused ethnography through random sampling and statistical analysis.
Field work was carried out during three field stays of approximately one month each, between July 2005 and September 2006. The aim of the stays was to provide contextualized data on the study area and population, to refine existing research questions and to detect new potential risk factors for malaria infection. During fieldwork, 12 villages were selected for their theoretical relevance, representing seven communes distributed in the two study districts. A total of 101 Ra-glai households were included: 58 households (57.4%) were informants selected during participant observation at public spaces (forest fields, roads to and from the forest, local shops and community health centres), while an additional 43 households (42.6%) were selected due to the recent malaria episode of one of their household members. The latter were identified by passive case detection (PCD) between August 1 and October 15, 2005, and represent all malaria cases identified during that period in their respective villages, and 68.3% (total = 63) of those identified in the five communities belonging to Bac Ai District. Additional interviews were held with health staff, including Ra-glai Hamlet Health Workers (HHWs) and general health staff from the seven CHCs and different institutions, including the Ninh Thuan Provincial Malaria Station (PMS), the National Institute of Malariology, Parasitology & Entomology (NIMPE) Hanoi and the Institute for Tropical Medicine Antwerp (ITM).
Participant observation consisted of participating in everyday activities, observing events in their usual context and carrying out reiterated informal conversations and interviews in order to build up confidence with informants and understand more sensitive subjects, such as forest activities, some of which are illegal. This technique proved useful in identifying bias in overall response during interviews and surveys.
One to five open or semi-structured formal and/or informal interviews were held with all included households (total = 101) during fieldwork. Households and key informants were contacted and interviewed reiteratively during the same or different field stays, building up confidence between researchers and respondents, which played a key role in distinguishing between biased and genuine response.
Focus group discussions
Several attempts were made to organize focus group discussions with local health staff but this technique was abandoned as the discussions tended to reflect the groups' internal hierarchical organization and adherence to social order rather than the aimed diversity of opinions and knowledge.
During the follow-up survey carried out in December 2005 (hereafter "2005-survey"), a questionnaire was administered to 3,685 randomly selected Ra-glai individuals surveyed bi-annually within the LLIH study . A total of 20 questions were asked as part of a standardised pre-coded questionnaire specifically related to the Ra-glai house-settlement system, forest activities, sleeping habits and knowledge and practices related to malaria exposure. The estimation of bed net use in homes and plot huts at forest fields was done on the sub-sample of 635 respondents that stated having slept at their fields in the month prior to the survey (17.2% of the total survey respondents).
Mosquito collections were carried out in three villages in Ma Noi commune and five villages belonging to Phuoc Binh commune. In total, five surveys were included (Nov 2004, Sept/Oct 2005, Nov/Dec 2005 & 2006). Human outdoor landing collections were done inside the villages (subsequently called 'village') and near forest shelters ('forest') for eight nights per survey in each location. Mosquitoes were collected from 18.00 until 06.00, stored by hour and morphologically identified in the field using a standardised key for medically important anophelines in Southeast Asia (for more details please refer to ).
In accordance with the research strategy, data analysis was a continuous, flexible and iterative process: preliminary data from different techniques were collected and analysed; further research was then conducted confirming or refuting temporary results until saturation was reached and the data theoretically supported. Qualitative data were systemised and analysed with N/Vivo Qualitative Analysis software (QSR International Pty Ltd. Cardigan UK). Percentages presented as a result from qualitative data analysis are illustrative of field research, hence not based on random sampling and should, therefore, not be equated to survey data.
Survey data were double entered and cleaned in Epi Info 6.04 (CDC, Atlanta; WHO, Geneva 1996), and analysed in STATA 9.0 software (Stata Corp., College Station, TX). Descriptive statistics were computed using the "svy" command in STATA, in order to take into account the survey characteristics. During the last phase of the fieldwork, preliminary results of the 2005-survey were analysed and further qualitative research was conducted to acquire supplementary data for the interpretation of those results.
In accordance with existing social science theory , risk perception for suffering from malaria was operationalized as encompassing the following components: (i) the perceived exposure to the disease, (ii) the perceived susceptibility to malaria infection and (iii) the perceived severity of malaria.
The study was approved by the ethical committees of the Institute of Tropical Medicine, Antwerp, Belgium and the NIMPE, Hanoi, Vietnam. During field work, all interviewers followed the Code of Ethics of the American Anthropological Association (AAA) . As proposed by the AAA, all interviewees were informed before the start of the interview about project goals, the topic and type of questions, their right to refuse being interviewed, to interrupt the conversation at any time, and to withdraw any given information during or after the interview, and the intended use of the results for scientific publications and reports to health authorities. Oral consent was preferred since the interviewees were not put at any risk of being harmed physically or psychologically and because the act of signing one's name when providing information during informal conversations could be a potential reason for mistrust .