Field assessment of a novel spatial repellent for malaria control: a feasibility and acceptability study in Mondulkiri, Cambodia

The study was conducted in two small villages in Mondulkiri province, Ou Chra and Pu Cha, with a population of 150 and 312, respectively. Ou Chra (12.235721, 106.848192) and Pu Cha (12.207171, 106.857025) are located in forested areas near the eastern border with Vietnam along a stretch of dirt road, circa 15 and 20 km respectively from Kaev Seima town. These villages were chosen as they represent a typical scenario of endemic malaria transmission in Cambodia—remote communities of ethnic minorities, living in proximity to tropical forests where the malaria vectors breed [19].

The great majority of residents in both villages are ethnic Phnong (also known as Bunong), an aboriginal group mainly found in Mondulkiri that rely on subsistence agriculture [20]. Malaria transmission in the province occurs throughout the year at relatively low rates, with peaks during the rainy season (June–July to October–November). The natural forest cover in a radius of 2 km around Ou Chra reduced from 98.0% in 2010 to 77.4% in 2015 [21]. The only road in and out the villages becomes nearly impassable during the rainy season, isolating the communities from the nearest market town and public health facilities. Both Ou Chra and Pu Cha have village malaria workers (VMWs), community health volunteers who are trained to diagnose suspected malaria cases by using a rapid diagnostic test (RDT), administer artemisinin-based combination therapy, and refer patients to the nearest public health facility [22, 23]. Long lasting insecticide-treated bed nets (LLINs) are distributed free of charge by the national malaria programme, and, as in the rest of Cambodia, their continued availability represents the main vector control method [14, 24].

In 2013, a cross-over entomological and parasitological survey was conducted in both villages to assess the effects of the spatial repellent metofluthrin (2,3,5,6-tetrafluoro-4-(methoxymethyl)benzyl(EZ)-(1RS)-cis–trans-2,2-dimethyl-3-prop-1-enylcyclopropanecarboxylate), supplied by Sumitomo Chemical Co. Ltd. (Hyogo, Japan), on anopheline densities and malaria incidence and prevalence. At the start of the study all houses in Ou Chra (n = 33) and Pu Cha (n = 59) were mapped with hand held GPS units. The emanators were made of polyethylene dual layer (15 × 8 cm wide) mesh held in an open plastic frame impregnated with 10% (w/w) metofluthrin. According to the manufacturers, a single emanator can protect a 30 m³ space for 4 weeks. The total estimated volume of each area requiring protection was determined, the number of emanators required and their optimum location (such as by the eave gap, if there was one) established. They were generally installed above head height so that they did not interfere with other activities and were themselves out of reach of curious children. The mean allocation of repellents per household was 2.5, with a range between 1 and 6 units depending on the household and room size.

In coordination with the entomological and parasitological survey, whose findings are presented in a previously published paper [25], an exploratory social science study was conducted in both villages to assess the feasibility and acceptability of the spatial repellent in the communities. The research design involved two stages of data collection, aiming to cover all households. In June 2013, prior to the first instalment of the emanator, a baseline survey was undertaken in both villages to gain an understanding of risk factors and protective gaps as well as social and contextual variables that may influence product uptake and utilization. The baseline questionnaire focused on risk perceptions and disease awareness and the use of other protective measures against mosquito bites, including details on ITN utilization, procurement, and cost (Additional file 1). Respondents were then asked to indicate the location of all household members at different periods of the day preceding the interview, according to local understanding of time (Additional file 2): bpro leum (early morning, 5:00 to 7:00), broek (morning, 7:00 to 12:00), rosial (afternoon, 12:00 to 17:00), l’ngiak (evening, 17:00 to 21:00), and yub (night, after 21:00). Information on socio-economic indicators was also collected to produce an index of socio-economic status, which was used subsequently to test potential differences in willingness to pay (WTP) for the emanator across population groups. Socio-economic variables included ownership of assets (i.e. radio, television, mobile phone, cupboard, bicycle, cart, tractor, and motorcycle), agricultural land, electricity, source of water, and house construction materials. Notes were taken during field trips to record observations about housing characteristics, household assets, and informal conversations with the villagers about social practices in the communities. One month after the first instalment of the spatial repellent, a follow-up survey was conducted in the same households to assess reception and acceptability of the product (in August in Ou Chra and in November in Pu Cha). In keeping with past evaluations of malaria control measures [26, 27], the second survey focused on perceptions about the effectiveness of the product, WTP and preferences for different procurement mechanisms (Additional file 3). Open-ended questions were included in both questionnaires to elicit further input and explanations on structured questions. Estimates of individual WTP were elicited with an iterative ‘bidding game’ approach [28].

Questionnaires were administered to the female head of household or another household member (where the female head was not available). Female heads of households were prioritized as women in rural Cambodia are usually responsible for day-to-day household management, small household purchases, and the management of bed nets [29, 30]. Most interviews took place in the morning, but some took place in the afternoon or early evening, depending on the participants’ availability and whether they were at home/in the village at the time of interview. Participants were interviewed face-to-face at their homes, and were either alone or in the presence of their spouse or other relative. The two surveys were carried out by a team of four researchers, all graduates in the social sciences and native Cambodian speakers, in close coordination with ML, who participated in the early field trips.

Survey data were double-entered and cleaned using an EpiData template. The datasets were then imported into STATA version 13 (StataCorp, College Station, TX, USA) for descriptive analysis and test statistics. Textual information from open-ended responses was translated into English, entered into an Excel file, linked to the survey dataset by unique identifiers, and coded for frequency analysis. A response was given multiple codes if more than one explanation was provided. Principal component analysis (PCA) was used to generate a relative wealth index, based on household-level information on socio-economic indicators [31, 32]. The first principal component was used to derive weights for wealth index. Households were then divided into tertiles based on their individual score against the wealth index: poorest, poor, and least poor. The Kruskal–Wallis test was used to test for significant differences in mean WTP across socioeconomic groups; an independent t test was used to examine associations between gender and WTP. Field notes were discussed collectively at the end of field trips, and reports were written. Key points from field reports are integrated into the presentation of results below to complement findings from the survey analysis.

Thirty of 33 households (90.9%) in Ou Chra and 57 of 59 (96.6%) in Pu Cha were interviewed. Characteristics of the study population (n = 448) are summarized in Table 1. Occupation, educational level, and socio-economic status did not vary significantly between the two villages. The great majority of residents with an occupation were farmers (92.4%), with an almost equal proportion of men (50.9%) and women (49.1%) involved in agricultural work. The mean length of residence of adults in Ou Chra was 23.9 years while in Pu Cha it was 14.1 years.

As observed in both villages, houses were built in a typical Khmer style, but differed in their quality of construction (Fig. 1). Most houses were relatively well built on stilts, but some were built on the ground or in a bad state of repair with large gaps in the walls. The exterior of the houses was reflected in the interior. People living in well-built houses had many more material possessions such as motor vehicles, televisions or radio, while the inhabitants of houses in disrepair had very little. In both villages, televisions (or DVD players) were common. Some farmers also had a second house or shelter in their rice fields or farms, where they could stay overnight during the harvesting season. These dwellings were partly or completely open, providing easy entry to mosquitoes [33].

During the baseline survey, the temporal and spatial dimensions of routines of all household occupants (n = 448) were mapped to understand potential exposure to mosquito bites and gaps in protection. The settlements tended to be sparsely populated during the day (Fig. 2) as farmers would leave home early in the morning to reach their rice fields or farms, tend cattle or work in the forest, while children went to school (Fig. 3), but after dark most residents were home, and the community population peaked.

In the evening, when malaria vectors may be active, residents spent a considerable time outdoors, resting in hammocks or engaging in social activities. By 21:00, the large majority of villagers were inside their houses (Fig. 2) and 59.8% were in bed (Table 2), where ITNs were reportedly used (Table 3). However, 40.2% stayed up longer (both indoors and outdoors), increasing their risk of exposure to mosquito bites. In addition, of residents who were away in the night preceding the interview, 28.6% (n = 12/42) slept in the deep forest where risk of malaria transmission is greater [34, 35].

Of all respondents at baseline (n = 87), 93.1% (n = 81) reported they had recently been bitten by mosquitoes in the village. The area underneath the elevated house on stilts, where people spent most of the daytime when they were at home, was indicated as the place where outdoor mosquito bites occurred most frequently. Furthermore, 50.6 and 59.5% of respondents noted that bites occurred “very often” inside the house and in the outdoor area surrounding the house, respectively. Indoor biting was reported to occur more frequently in the evening before sleeping time, followed by at night, while outdoor biting occurred more frequently in the early morning (Fig. 4).

Awareness that mosquito bites can spread disease was very high in the study population (98.9%; n = 86/87). When prompted to give further explanations, 85.1% (n = 74) mentioned malaria (krun chanh) and associated it with symptoms such as high fever, chill, headache, diarrhea, or headache, while 8% (n = 7) listed correct symptoms without naming a specific disease. Of those who named malaria, 20.3% (n = 15/74) also mentioned that mosquito bites can cause dengue fever (krun chiam). Of all cited information sources on malaria (n = 106), the health centre/post (35.8%) was the most frequent, followed by the local VMW (28.3%), non-governmental organizations (13.2%), personal experience (10.4%), radio advertisement (5.7%), private providers (3.8%) and other information sources such as relatives (2.8%). Only a minority of respondents demonstrated incorrect understanding of disease causation, with 6.9% (n = 6/87) stating that mosquitoes can cause both malaria and typhoid fever (krun pou vien) and 1.1% (n = 1/87) reporting that malaria can be transmitted by drinking unclean water.

The large majority of respondents reported using protective methods against mosquito bites inside the house (98.9%), outside the house (81.6%) and in the rice field, farm or forest (91.7%). In both villages, bed nets were the most common form of protection inside the house (96.5%), followed by mosquito coils (9.3%), indoor sprays (8.1%), and traditional methods such as burning incense (2.3%) (Table 3). Of all nets across the study households (n = 171), 89.3% were reportedly hung inside the house in the villages (Table 4), while a minority were installed under or near the house (5.7%) or in the rice field or farm (5%). The nets distributed as part of a programme (either through the local VMW, health centre, or non-governmental organization) were allocated free of cost, although a small payment of 1000 riels (US$ 0.25) or less was often given as a voluntary, informal contribution to show appreciation and gratitude. Nets purchased in the private sector (either at the market, local shop or from an itinerant seller) accounted for 13.6% of all nets in the villages, and had a mean cost of 30,360 riels (US$ 7.6), range US$ 3.7–16.2. Finally, a discrepancy between the number of reported and observed nets was found, with only 70.2% of ITNs that were reportedly used in the village being observed.

A variety of different methods were used to prevent mosquito bites outdoors (Table 3). In the area surrounding the house, the prevention method most commonly cited were burning leaves (36.6%), clearing vegetation (33.8%) and wearing long clothes (33.8%). In the farm or forest, wearing long clothes was the most common protective method (67.8%), followed by burning leaves (31.0%), using bed nets (17.2%) and hammock nets (12.6%). Of all participants, 80.5% (n = 70/87) said they would like to use additional protective tools. When asked to specify the type of protective tool, mosquito spray was the most frequent answer (32.8%; n = 19/58), followed by coil (20.7%), light traps (15.5%), additional nets (10.3%), topical repellents (6.9%) and spatial (5.2%) repellents. The remaining answers (10.3%) were vague, for example “anything that kills mosquitoes”.

The small size of the study population is perhaps the most important limitation of this study. In addition, differing practices and beliefs could potentially be observed in other endemic areas and ethnic minorities. Long-term ethnographic work in the villages and the use of qualitative methods such as focus group discussions were also not possible due to time and resource constraints. This would have been a valuable addition to the research design, enabling deeper understanding of protective gaps in the study locations and further insights on the feasibility and acceptability of the product. Finally, the product was not tested in environments outside the villages where high risk of malaria transmission exists, such as the deep forest or the shelters where some farmers stay overnight in the harvesting season.