Madagascar is a huge island and extends over 500 km from east to west and 1500 km from north to south. The relief divides the island into three bands: a narrow coastal strip in the east, the Highlands in the center, and an area of low plains and plateaus in the West. The country is characterized by a large diversity of climatic variations: the east of the country may suffer from floods at the same time that the South encounters long periods of drought.

Madagascar's diversity is also reflected in the origin of the Malagasy population, which remains an ethnological enigma. The contribution of South East Asia and Central and Eastern Africa is visible in the social organization, language, and religion. All the peoples of Madagascar share a culture that represents a synthesis of these two inputs, one of which may predominate by region [31, 32]. This diversity of origins, climates, and landscapes, induces a mosaic of populations and practices that are difficult to represent in a single study. So, to capture practices as diverse as possible, we chose four different geographical areas among the 31 zones of the MEDALDI project: an area in the East (Mananjary), one in the Fringe (Moramanga), one in the Highlands (Fianarantsoa) and one isolated rural area, located in the north of the island (Antsohihy). The selection of these qualitative investigation areas was based on epidemiological criteria, social and cultural features of ethnic groups and the vector control interventions implemented in these sites: LLIN (Moramanga, Mananjary, Antsohihy) and IRS (Moramanga and Fianarantsoa) (S1 Fig).

In terms of malaria transmission, the selected regions have different epidemiological profiles. The region of Mananjary in the East of the island is characterized by high and permanent transmission; populations are exposed to vectors on a daily frequency throughout the year. The region of Moramanga is located on the fringe between the coast and highlands, and is characterized by a low and seasonal transmission. The region of Fianarantsoa is located on the central highlands where the transmission is almost nul. This area is characterized by an economic immigration that facilitates importation of malaria. The region of Antsohihy in the North is characterized by moderate and seasonal transmission, and variable presence of mosquitoes. In February 2012 an outbreak, which was attributed to malaria after epidemiological and parasitological investigations, caused many deaths in this area. In all areas the main activity is rice cultivation. The difference between the regions lies in their level of wealth as indicated by differences in housing materials and household’s belongings. The northern area (Antsohihy) is characterized by cash crops for export (cocoa, vanilla) which tends to make it richer than other areas. Lack of social and health infrastructures, roads, and markets represents a significant obstacle to development in the all the areas investigated.

Data was collected between May and December 2012. The study was designed in two steps: in August 2012 an exploratory mission was conducted in order to identify the fokontany of investigation- the smallest Malagasy administrative entity—and October and December 2012 in-depth missions between were carried out.

An anthropological approach was taken to data collection, the following qualitative tools were used: (i) semi-structured interviews of various categories of individuals (community members, health workers, traditional healers and traditional birth attendants), (ii) direct observations (living conditions, installation and use of bed nets, environment, water supply point, hygiene, queue for consultation at the health center, availability of health workers) and (iii) informal interviews in the context of a annual mass distribution of LLINs. Observations allowed a comparison between practices and discourses about usage of bed nets.

In order to standardize interviews, we developed interview templates. A total of 71 in-depth interviews were carried out with two categories of participants: community members (men and women) and healthcare providers. More specifically, 54 interviews with community members (26 men and 28 women) and 17 with health workers (biomedical health workers–nurses, doctors, community health agents–and traditional healers and birth attendants) were carried out. For the identification of participants amongst community, purposive sampling was used to ensure gender and age ratio (between 20 and 80 years of age). One out of two houses was invited to participate, until we reached our objectives in terms of gender and age ratios and data saturation. A minimum quota of 14 interviews per area was determined.

All persons solicited for the study agreed to participate. The interview template for villagers focused on fever and malaria (vernacular words used to refer to malaria, perceptions, causes, symptoms and steps of the care seeking), malaria prevention practices, information channels about malaria and knowledge of prevention messages. The health worker’s template focused on treatment of malaria, time spent for providing care, difficulties met in providing care, information channels about malaria and involvement in sensitization and prevention actions. We interviewed villagers at home, and professionals at their work place. The interviews were carried out in the Malagasy language with an interpreter. The National Ethics Committee under the Ministry of Health of Madagascar gave us ethical clearance for this study. We obtained verbal consent from participants before starting or recording interviewees, written consent was not required by the Committee for qualitative studies. Additionally, in order to receive authorization to work in these areas, and in respect of the health and administrative hierarchy in place in Madagascar, courtesy visits were made to the administrative authorities (mayors) and sanitary authorities (medical inspectors) at Regional Health Centers. On the local level, the fokontany presidents informed the community of our presence during village meetings. Along with these efforts, we also presented ourselves to traditional leaders to facilitate access to households.

The number of interviews was determined by the data saturation criterion and the time limit for each mission on the field. The objective was to include the same number of men as women, divided according to different age groups (under 20 years, 20–30 years old, 30–40 years old and more than 40 years old). Our choice of interlocutors was dependent upon availability at the time of our investigation. One in two eligible households was interviewed. In all areas, all doctors, nurses, traditional healers were interviewed.

All the interviews were fully transcribed and translated in French by a team of five bilingual translators. We analyzed the interviews according to a thematic analysis chart that was designed to identify the themes cited in each interview and to make comparisons in order to highlight differences and recurrences in the discourses.

The Malagasy word for malaria is tazomoka which literally means "mosquito fever". Tazomoka is the word used by the biomedical establishment for malaria. However, this term was almost never used by interviewees, and when it was used, it was onlyafter a diagnosis had been done by biomedical health workers.

Several channels disseminate information on tazomoka: radio, posters, mass awareness campaigns during village meetings and consultations with health professionals. Interviewees say they appreciate the information transmitted about tazomoka on the radio. However, such information is sometimes misunderstood. For example, one man said he heard on the radio that sickle cell anemia is a new disease transmissible through mosquitos. A woman claims to have heard that tazomoka is a "problem in Tana".

Village meetings were places of socialization and opportunities to exchange with their neighbors. Most interviewees were not able to recall all of the information given during these meetings. For example, in all areas they retained a specific message "you have to sleep under a mosquito net" but did not retain information on maintenance of the nets or duration of treatment effectiveness. At the time of the outbreak in Antsohihy, official messages seemed not to have had the expected impact on population: for instance, one of the messages transmitted by the health staff consisted in saying that the vectors of tazomoka were only female mosquitoes, which could be identified by their long legs.

National organizations and various NGOs disseminate many messages related to health (e.g. family planning, prevention against diarrhea, tuberculosis). These messages were sometimes competing, and confused the information that was received. As a result, for instance, some interviewees came to believe that the cause of the malaria outbreak in Antsohihy was related to a hygiene issue and they cleaned their hands as often as possible to protect themselves.

According to both caregivers and patients, given the lack of time available for consultation with biomedical health professionals, only minimal information was transmitted at this occasion.

When the question was asked, interviewees couldn’t tell the name of the drug delivered at the health center during consultation for malaria related symptoms. Likewise, most women didn’t know the names and the purpose of the drugs received during pre-natal consultation (including ITPs). On this particular issue, many caregivers said they did not disclose the name of ITP to pregnant women in order to avoid self-medication. Biomedical health professionals acknowledged that during consultation they only gave the information on the drug dosage. Limited time available for consultation was an explanation to this lack of communication. Another explanation advanced was that in the countryside, the patient is seen as illiterate, and not able to understand such details. This perception was largely shared among health staff. At the same time, we observed a lack of interest of patients for the biomedical names of diseases and details about treatments prescribed. Interviewees reported that they stop treatment when they feel better, without taking into account the treatment period prescribed by the caregiver. Consultation is not perceived as a place where questions can be asked.

We observed that most information retained about malaria was spread through informal information circulation channels such as rumors, by word of mouth or discussions between closest friends/relatives.

The symptoms related to malaria (fever, seizures) are not interpreted by the population as necessarily being linked to tazomoka, (i.e. the infectious disease that occurs due to the bite of a mosquito). Two popular nosological entities were frequently used by interviewees to refer to malaria: Tazo (“fever”) which referred to the sensation of warm body and simple fever and tazomahery (“strong fever”), when high fever was accompanied by seizures. Tazo in everyday language means “fever” and may be accompanied by more symptoms defined as frequent and benign: flu, cough, and diarrhea. The symptoms of tazomahery are similar to the biomedical category of severe malaria.

Tazo was perceived as simple to treat through self-medication, the habitual care seeking behavior in case of illness. For the biomedical health professionals, the difficulty in the differential diagnosis of malaria, coupled with frequent shortages of Rapid Diagnostic Tests (RDT), has been frequently cited by caregivers as a problem in the management of the disease.

According to local perceptions in all areas, tazo or tazomahery were not caused by mosquito. Tazo was mainly caused by an imbalance between the individual and the environment, such as a weather change. In Antsohihy, tazomahery was perceived as a supernatural disease brought about by ancestors’ as a sanction due to the violation of an ancestral taboo (e.g. bathing in a sacred lake). In Mananjary, it was mainly interpreted within a social dimension through the power of witchcraft. In local perceptions, from the time when convulsions appear, the disease was considered as serious and worrying. But these symptoms were not interpreted as belonging to tazomoka.

The outbreak attributed to malaria by health authorities in February of 2012 marked the memory of Antsohihy’s inhabitants. A health worker in Antsohihy declared that during the rainy season of 2012, 70% of consultations were related to malaria. The villagers gave various reasons to the outbreak, but only few people recognized tazomoka to be the cause, because of four main reasons. The first reason was the low density of mosquitoes observed during the outbreak. The second was that so far, mosquitoes and tazomoka had never been fatal in the view of the respondents. In certain circumstances before the outbreak, villagers slept outside of their houses without mosquito nets, e.g. when the cattle farmers annually migrated to seek water and pasture land for the herds of zebus, or when they harvested rice. According to our speakers, in spite of lack of protection, these age-old practices had never brought about death. The third reason for which tazomoka was not incriminated was that during the outbreak, the usual tazo management practices, i.e. self-medication, did not give the expected results, reduction of fever and recovery. The last reason that confused the population was that the symptoms during the outbreak (diarrhea, vomiting, rapid loss of consciousness after the first signs of fever) were different from the usual symptoms of tazo. For all these reasons, for the villagers, the cause of the outbreak couldn’t be tazomoka.

The populations looked for other causes of the disease. Since mainly children suffered and died from the outbreak, some blamed the last tetanus vaccination campaign, which had taken place a few weeks before and targeted children. Others, who had received various sensitization messages in the previous months, identified dirt and lack of hygiene as causes of the outbreak. Some interviewees, puzzled by the immediate consequences of the disease, gave a supernatural explanation. The outbreak was then interpreted as a sign coming from “up above” of the ancestors’ anger because of the lack of respect for a sacred place.

In agreement with the numerous interpretations of the origins of the outbreak, many prevention attitudes came up: refusal to participate to the vaccination campaigns that followed the outbreak, increased caution as to hygiene and food eaten, construction of latrines, wearing amulets and traditional objects of protection. For the minority of individuals who had assimilated the association between outbreak and mosquito, more regular use of bed nets.

In all sites, most people affected by fever declared resorting to self-medication based on plants and pharmaceuticals. Most individuals bought drugs from the informal sector: grocers, street vendors, individuals who have drugs for sale at home, or itinerant hawkers.

Self-medication practices were also observed in villages where there was a public health center and its drugstore (Fianarantsoa and Antsohihy). In all study sites, reasons mentioned for using the informal sector included convenience (availability of drugs day and night, social proximity of the vendors), geographical and economic affordability (retail at lower costs and/or on credit, availability of drugs within the isolated villages).

When resorting to these retailers, the process was always the same: either the patient described his symptoms and the seller offered drug(s) that he thought adequate, or the patient had already made his choice based on a previous pathological experience. It was observed in all areas that the close circle (neighbors, relatives) and word of mouth played an essential role in the choice of drugs and in care seeking practices. Previous episodes of fever, for which a doctor's visit had been made, guided the choice of drug when new episode of fever appeared. One would then rely on an old prescription to buy drugs.

Only when the patient’s health got worse, or when the available community resources were exhausted, did he resort to a health facility. As a consequence, he would get a medical diagnosis and received proper treatment only at the end of the care seeking process. In the case of malaria, we observed that there were about three days between the moment fever appeared and the consultation at a health unit.

People declared that, once they arrived at the health center, there was a long queue before getting to the consultation, and that health staff only spent a few minutes on consultation. Some patients had to pay for the ACT treatment, as not all interviewees were informed of the gratuity of the ACT treatment.

Along with self-medication based on industrial pharmaceutical drugs, the use of traditional herbal remedies was also mentioned in fever case management, especially when the cause of fever was believed to be caused by a magic spell.

In all areas, the most frequent reason for bed nets use was protection from inconvenience caused by insects, e.g. stings from fleas and mosquitoes or noise nuisance from mosquitoes.

Beyond protection against insects, other factors promoting use of bed nets were mentioned such as protection of the intimacy of the parents. Bed net usage was reported to increase when mosquitoes were present in the environment. According to the interviewees, Mananjary was an area where mosquitoes were present in swarms. In that area, the bed net was part of families’ daily lives, and in the rainy season, when mosquitoes were more numerous, the speakers were more careful in the use of it. The bed net was part of the birth and wedding kit; it was part of the equipment the couple must have to live together. According to interviews with men, it was perceived more as the sign of a man’s capacity to support a family than a means of protection. In the three other areas, due to the perceived sparse density of mosquitoes during the year, bed nets were used to protect against the cold in the winter, or as decoration, as it was seen in Antsohihy. With its colored ribbons and its openings on the four sides, the “four doors bed net” was quite popular in Antsohihy. These bed nets cost between 20 000 and 100 000 ariary (approx. 7.00–35.00 euros), on the contrary, LLINs were free of charge during distribution campaigns and 3 000–4 000 ariary (approx. 1.00–1.30 euros) in shops and drugstores. The richer families that could choose between a free LLINs or a “four doors bed net” would buy the latter, or would make a “four door bed net” with several LLINs. Poorer families would use free LLINs and, in case of shortage, they would not use their resources to buy additional mosquito nets to accommodate all family members.

In all study sites, the prioritization in the use of nets was largely influenced by messages that emphasized on the vulnerability of the target beneficiaries—pregnant and lactating women, and children less than 5 years- leading to the idea that other family members would be less exposed to tazomoka. In two out of four study sites, men said they were not at risk to get tazomoka.

Right after these vulnerable groups, parents were prioritized in the use of nets, since it preserved the couple’s privacy. In three out of four study sites, it was mentioned that children of the same gender over ten years of age could not share a same bed because of traditional taboos (fady).

Some reasons mentioned for the non-use of bed nets were common in all study sites: the feeling of suffocation especially with high temperature even during the night, irritation of the skin in contact with the net, loss of light in the room, and difficulties related to the fixing of the nets in tight spaces. Several local prevention measures to avoid having mosquitoes into the houses were cited: insecticide coils, burning leaves (cypress, orange or mango), or diverting smoke from charcoal and wood cook stoves into houses.

Observations were made of alternative usage of nets: fishing or crop protection. But observations implicated that such nets were already patched and stitched together and could not be used as protection anymore. In case of distribution of new bed nets, the pervious nets were used until they were useless. This was especially true if the new nets distributed were not appreciated because of the color, type of net veil, or type of mesh.

With the aim of understanding the barriers to adherence to malaria control strategies, we mainly analyzed local perspectives, through the concepts of risk perception and gravity of malaria. One limitation of this study lies in the overly specific focus on malaria without a more local understanding of the body, health, and well-being.