Study site and population
Population
Cambodia is inhabited by approximately 90% ethnic Khmer. However, there is a small
ethnic minority population located mostly in the northeast, in Ratanakiri and Mondulkiri
provinces. These populations are part of a larger cultural area, which extends from
Laos in the north through the central highlands of Vietnam in the east and finally
Ratanakiri and Mondulkiri in the south 20]. The largest ethnic groups in Ratanakiri are the Jarai, the Tompuon and the Kreung,
each with a distinct language and cultural system in terms of kinship and political
organization 21]. A characteristic shared by these ethnic groups is that they usually combine slash-and-burn
agriculture with hunting, fishing and gathering various forest products. As distances
between farms in the forest, rice fields and villages can be substantial, many families
maintain residences at each location, and move from one place to another according
to the agricultural cycle and the forest farm or rice field’s requirements 8],21].
Malaria
Malaria transmission is perennial with two peaks, June-July and October-November,
the rainy season lasting from May to October. At the end of the malaria season of
2012, the overall PCR prevalence in Ratanakiri, as recorded by the MalaResT study
(cfr. Infra), was estimated at 4.9% 22]. Species-specific areas with elevated risk of infection have been detected for all
Plasmodium species. The clusters for falciparum, vivax and ovale malaria appear in the north
of the province along the main river, while the cluster for Malariae is situated in
the south of the province 23]. The primary vectors in Ratanakiri are Anopheles minimus and Anopheles dirus, and many secondary vectors are present. These vectors are generally exophagic and
exophilic, and their densities and behaviour vary extensively per village 13]. Early biting proportion (EBP), which is calculated as the biting activity before
22.00 (i.e. assumed human sleeping time), was observed to be around 50% in Ratanakiri
in 2005, the proportion of infectious bites before 22.00 was 29% 13].
Study context
This study took place in the framework of an intervention trial (MalaResT), which
aimed to raise evidence on the effectiveness of the mass use of topical repellents
in addition to LLINs to reduce malaria infections. In this intervention trial, 113
of the most endemic villages in Ratanakiri were randomly assigned to a control arm,
in which every household received one LLIN per one person, or to an intervention arm,
where in addition to LLINs topical repellents were distributed biweekly to every household
one bottle per one person. For the epidemiological trial, sample size calculations
were based on an expected outcome of 40% difference in malaria prevalence between
intervention and control arm. During each of the four malariometric surveys organized
in the trial, the aim was to collect blood samples of 65 randomly selected participants
within each community. Blood samples were analysed using PCR detection in a mobile
laboratory, allowing for a sensitive and rapid malaria diagnostic strategy in the
field 22], alongside a small questionnaire on overnight stays at different locations in the
month prior to the survey.
Research strategy
The research used a sequential mixed methods study design in which qualitative ethnographic
research and quantitative survey research methods were used to complement the qualitative
findings (in standard annotation [QUAL -??quan]) 24]. Qualitative ethnographic data were collected in local communities to acquire an
in-depth understanding of the study setting and population while a cross-sectional
and a structured observation survey aimed at quantifying relevant variables from the
qualitative study. In addition, a malariometric survey was performed, thus enabling
to link previously determined relevant variables to malaria infection, and entomological
surveys were performed in selected villages.
Quantitative study
Data collection
The exploratory and in-depth ethnographic research was done in 2012 in a selection
of villages included in repellent study mentioned above, more specifically in the
intervention villages (with repellents and LLINs) of Kachon Kraom, Lom and Sayos in
respectively Voen Sai, Oyadao and Lumphat district, as well as during shorter visits
in other communities around these villages, including some control villages (with
only LLINs).
Participant observation and in-depth interviewing were carried out within the qualitative
strand of the study. Participant observation consisted of observations and reiterated
informal conversations and was especially used as an exploratory technique to detect
unforeseen variables and to contrast stated opinions with actual behaviour, as it
constitutes a respondent independent data collection tool. In 2012, 153 in-depth interviews
were recorded and transcribed.
Multiple purposive sampling techniques were used, where informants were selected in
relation to emerging preliminary results. In order to increase confidentiality with
respondents and consequent reliability of the data, snowball sampling techniques –
participants introducing us to other participants – were also used.
Quantitative study
Data collection
For the quantitative strand, two surveys were carried out. From August till November
2012, a cross-sectional survey was performed with a close-ended structured questionnaire
based on relevant variables emerging from the ethnographic strand. It explored the
following topics: mobility between farms, fields and villages, repellent use, bed
net ownership and use, evening social activities, use of malaria preventive measures
other than bed nets, perceived mosquito density and nuisance and malaria treatment.
From May until November in 2013, a structured observation survey was carried out.
A first visit, in the evenings between 19.00 and 21.00 depending on the availability
of the household, consisted of the observation of housing structures, people’s resting
behaviour, bed net characteristics and topical repellent use of all household members.
As actual bed net use at night could not be directly observed, bed nets that were
suspended in the evenings before bedtime with at least two corners were considered
ready for use. Holes in bed nets were observed but not systematically measured or
counted. The next morning, a follow-up questionnaire was administered, exploring socio-economic
status, seasonal sleeping spaces, perceived insect and mosquito protection, (alternative)
use of nets, child care system, (alternative) repellent use, perceived mosquito nuisance,
and previous malaria episodes was carried out with the household leader. Results regarding
repellent use are not elicited in this paper.
Sampling
For the cross-sectional survey, 900 individuals were randomly selected from the MalaResT
study population. In total, 393 individuals from 56 intervention villages and 431
from 57 control villages were located and answered the structured questionnaire.
For the structured observation survey in 2013, ten intervention villages and four
control villages were purposively selected based on the criteria of malaria incidence.
In each village, half of all households were randomly selected from the population
census. Although there was no prior information available on which household had a
farm or not, based on exploratory qualitative research, it was assumed that the majority
of households did have a farm and commuted between farm- and village house, favouring
the farmhouse during the rainy season and the village house during the dry season.
As such, each selected household was assigned either to a farm list (meaning this
household had to be observed and interviewed on their farm or rice field, if they
had any) or to a village list (the interviewer had to observe and interview this household
in their village house, if they had one) to explore potential differences between
locations. A total of 653 households were randomly selected for the farm lists, of
which 472 households were eligible (meaning they had a farm house where they stayed
overnight); and 260 of those households were reached. A total of 655 households were
randomly selected for the village lists, of which 555 were eligible (meaning they
had a house in the village where they stayed overnight); and 291 households were reached.
The main reason for not reaching households was because the selected family was not
staying overnight at the respective location within the timeframe of the survey. Finally,
a total of 431 households from the intervention arm and 120 households from the control
were observed and interviewed.
For both surveys, an additional non-response form – recording reasons for not reaching
the household – was used to measure possible systematic self-selection bias. The possible
correlation between malaria infection and overnight stays at different locations was
additionally tested in a malariometric survey in October 2012, on a sample of 6,640
individuals selected randomly out of the population censuses of the 113 villages included
in the MalaResT study. Out of the list of 6,640 selected individuals, 4,996 individuals
were reached.
Entomological collections
To estimate human exposure to malaria vectors, pooled data of different entomological
surveys conducted in Ratanakiri province are presented here. Prior to the MalaResT
project, indoor and outdoor human landing collections were carried out in two villages
(Phi and Lom) during four surveys of five to eight days in July-August 2009, July-November
2010, and July-August 2011. Human landing collections lasted from 18.00 until 06.00.
Three collection points were chosen in each village, with paired indoor and outdoor
collections per collection point. Within the MalaResT study only outdoor human landing
collections were carried out in two intervention (Koy, Chrung) and two control (Kreh,
Klis) villages selected out of 113 villages included in the project based on their
malaria incidence, their accessibility and the availability of mosquito collectors.
In every village, eight entomological surveys of ten successive nights were organized
every two months between April and October of 2012 and 2013. Human landing collections
lasted from 17.00 until 22.00 and from 17.00 until 08.00. Seven collection points
were chosen per village in front of houses across the village, making a collection
effort of 70 man-nights per village per survey. The same collection points were maintained
throughout both studies. A rotation of collectors was ensured. Mosquitoes were identified
and processed according to procedures described in Durnez et al.13]. Mosquito collections were pooled per collection context and per collection hour
for data visualisation. Results regarding changing mosquito behaviour in relation
to repellent use are not elicited in this paper as they are subject of a separate
paper.