An exploratory study of community factors relevant for participatory malaria control on Rusinga Island, western Kenya

Rusinga Island (0°35'–0°44' South; 34°11'–34°22' East; altitude 1,100 m) is 42 km² in area and is the second largest island in Lake Victoria. Due to its close proximity to the mainland a 200 m long causeway was constructed in 1983 to link the island with Mbita township, the major trading and the administrative centre of the district. The terrain is extensively deforested and generally rocky and hilly with limited vegetation cover. There are a number of seasonal rivers which contain water only during the rainy season and the lake provides the main water source for the population. Two rainy seasons are typical for the area, the 'long rains' between March and June and the 'short rains' between October and November, but these seasons are unreliable with some years characterised by prolonged dry periods. Malaria transmission fluctuates seasonally but is sustained all year round by the three primary malaria vectors: An. gambiae,, Anopheles arabiensis and, to a lesser extent, Anopheles funestus [8, 54‐56]. As per a census implemented by the end of 2006 during the establishment of a Demographic Surveillance System (DSS), Rusinga island had 24,078 inhabitants which formed 5,425 households and 21% of the population were children below the age of five (S. Kaneko et al., Nagasaki University, unpublished data). The predominant language spoken is Dholuo.

People living on Rusinga face a multitude of problems. The island has suffered enormous environmental degradation, soil erosion and extended drought conditions in recent years leaving little productive land and few opportunities to make money other than through fishing. Furthermore, construction activities, deforestation, vegetation clearance and poorly planned infrastructure development has led to an increased abundance of mosquito larval habitats [8], notably those suitable for malaria-transmitting Anopheles. The high prevalence of both malaria (50%) and HIV/AIDS (30%) on the island has been a major impediment to socio-economic development [57, 58].

Two government health centres serve Rusinga's population; one in the north-eastern part of the Island and one in Mbita township. Additionally, there are three registered and many unregistered private facilities on the island. Due to the bad condition of the roads public transport is rare especially in the rainy season and it is difficult to reach the health facilities.

The RMP is implemented through CCF-Kenya's community-based organization RICFP. The CBO has been carrying out development activities on Rusinga for the past 18 years. CCF is an international NGO with the mission to promote the well-being of children by facilitating quality programmes in education, early childhood development, youth, health and nutrition, all including aspects of environmental health and healthy homes [59]. CCF-Kenya is supported by sponsors for 44,000 individual children from around the world and operates in 30 districts of Kenya. The project involves all families of enrolled children in ongoing activities and decision-making processes. CCF-K affiliate CBOs are owned and managed by parent-committees, selected by neighbourhood groups (jirani) of enrolled parents. The project has divided the island in eight administrative zones. Each jirani group selects a representative to serve in the parent's committee. Using this system, parents identify their own needs, prioritize them, plan, budget and also implement activities. RICFP reaches a total of 4,352 people in 750 families of which 869 are enrolled children aged below 1 to 17 years through direct sponsorships. RICFP intends to reach the entire Rusinga community with their educational programmes.

Prior to the community-wide KAP survey a series of focus group discussions (FGDs) were conducted in all eight administrative zones with eight to 10 community members per zone, selected by RICFP parent committees. Those focus groups contained of equal numbers of members from CCF-enrolled and non-enrolled families and were gender balanced. The topics that were discussed were signs and symptoms of malaria, causes and mode of transmission and prevention of malaria. The information from these discussions was used to guide formulating the questions for a cross-sectional KAP survey which was conducted from April to July 2004. The study adopted the RMP administrative units of zones to implement a stratified random sampling scheme. A two-stage cluster procedure with zones and households within them as the two levels sampling units were adopted. The households were divided into two categories; the CCF-enrolled and the non-CCF-enrolled households. The sample size depended on the number of CCF-enrolled households as the intention was to interview all these households (n = 750). A matching household sample size from the same community was randomly selected from non-CCF-enrolled households. Preference for interview was given to households that had at least one child under five years of age. Therefore, a census was implemented in all the eight zones, recording all families with children under five years of age. From this list, households were randomly selected to match approximately the number of CCF-enrolled households in each zone. A total of 1,500 households were selected for interview.

The study largely used quantitative approaches in data collection through a semi-structured questionnaire consisting of 69 questions, both open ended and closed. While the questionnaire was written in both English and Dholuo, the interviews were conducted in the latter ethnic language.

Pre-tests of the questionnaire were conducted in 14 households and adjustments made accordingly. The interviewers were trained residents of Rusinga Island who were fully familiar with the local language, culture and sensitivities. One questionnaire was administered to one household per compound. Households were defined as a group of individuals sleeping in the same house and depending on the same budget [60]. Because many men on Rusinga are polygamous and live in extended family structures several households usually form a compound and often share in a common pool of resources. For interview, preference was give to the female head of the household because she is typically the care taker of the children and responsible for all household duties. In the event that she was absent at the time of interview either the male head of the household or any child above the age of 12 years was questioned. Interviews were conducted in private to reduce influence of relations. Where occupants were not found on the first visit, two other attempts were made to trace them.

Being part of an integrated development programme at community level, the CCF-enrolled families have been involved in various health training activities including malaria prevention. On the other hand the non-CCF-enrolled families have not been exposed to this type of training and health care. The objective here was to investigate whether there was a measurable difference in knowledge, attitude and practice concerning malaria in families that have been exposed to these activities and been embedded in a social network by an established NGO. In general, the study aimed to investigate whether household socio-economic status or demographic characteristics affect malaria related knowledge, prevention and treatment behaviour and the willingness to participate in a community-based programme. The questionnaire was structured into the following topics: (1) demographics of respondent and household; (2) socio-economic status of the household; (3) knowledge and beliefs about malaria (symptoms, prevention methods, mosquito life cycle) (4) typical practices toward malaria prevention; (5) the treatment-seeking behaviour and household expenditure for malaria treatment; and (6) the willingness to prepare and implement community-based vector control.

The institutional ethical clearance was granted by the joint University of Nairobi – Kenyatta National Hospital ethical review committee (protocol approval number P102/7/2004). In addition, permission was obtained from the district authorities and community leaders. Individual interviews were only started after the purpose of the study had been clearly explained to the participant and an informed consent form was read and signed.

The semi-structured part of the questionnaire was coded after completion of the survey. All data were entered and analysed using the statistical software package for social sciences (SPSS) Version 14.0. Analyses of the outcome of variables were performed excluding non-responders or missing data points, therefore only valid percentages of the responses were accepted which leads to the fact that the total number of respondents (n) varied between questions. A socio-economic index [61] based on household assets was created using Principal Component Analysis (PCA) [62]. The final PCA was based on 13 asset variables (sofa set, bicycle, radio, television set, solar panel, generator, car battery, mobile phone, boat, fishing net, number of cows, goats and chicken) that were identified by community members during FGDs and explained 27% of the variability in the 13 variables. Each interviewed household was assigned to a socio-economic quintile (most poor; very poor; poor; less poor; least poor) according to PCA. Chi-squared tests (χ²) were used to examine whether the distribution of individuals/households among the categories of one variable is independent of their distribution among the categories of the other. Logistic regression analyses (backward stepwise) were used to explain variations in responses to questions about knowledge and behaviour towards malaria and its control (e.g. bednet ownership and bednet use, knowledge of mosquitoes as malaria vector) based on socio-economic and demographic indices.

Of the 1,500 households selected for the survey residents from 1,451 households were interviewed (97% response rate), 701 households being CCF-enrolled and 750 households being non-CCF-enrolled. Interviewed households were equally distributed over the island with an average of 12.5% (95% C.I. = 10.6–14.4) of all respondents interviewed in each zone. There were 1,054 female and 397 male respondents.

The mean age of respondents was 34.5 years (95% C.I. = 33.9–35.2) and 70% were within the age range of 21.5 to 47.5 years. The average household on Rusinga had 6.2 household members (95% C.I. = 6.1–6.3) and 4 children (95% C.I. = 3.9–4.2); 1.7 children below the age of five years (95% C.I. = 1.7–1.8). Nearly all respondents (98%) had lived on Rusinga for most of their life. In 32% of all households, the household head was polygamous with two or more wives. 2% of wives were 18 years or less (n = 23), some as young as 14 years old.

Table 1 shows how some of the demographic and economic variables are distributed in households of different socio-economic level. Notably, the majority of the female headed households were in the poorest socio-economic quintiles. This distribution differs significantly from the male-headed households. Of all respondents, only 21% were educated beyond primary school, among whom 1.6% attained tertiary-level education (Table 2). There was a significant difference in the educational level between women and men, with more men educated beyond primary level and a higher percentage of women without any formal education (Table 2). Unsurprisingly, more highly educated respondents were found in wealthier households (Table 1). CCF-enrolled households belonged predominantly to the lower socio-economic levels including 76% of all respondents without formal education (68 out of 90) and only 4% of respondents had a college degree (one out of 23), indicating that the CCF programme's targeting strategy is well implemented. Interestingly, households of higher socio-economic status were also characterized by a higher number of household members, children and wives. Fishing and small-scale businesses like fish mongering and vegetable sales were the most common income generating activities undertaken by the residents of Rusinga; fishing primarily done by men (62% of male respondents) and small-scale businesses by women (48% of female respondents). A sizeable proportion of women (32%) did not work outside the home (housewives) and depended on the income of their husbands (Table 2). In 30% of households fishing (n = 472) and small-scale businesses (n = 426) were mentioned as main sources of income, respectively. 12% of households (n = 179) received their main income from larger businesses and notably 8% (n = 116) of the households depended primarily on remittances from relatives outside Rusinga (Table 1). Formal employment as major income source was primarily found in households of the highest socio-economic status while most of the labourers for small wage income were found to belong to the poorest households (Table 2). The average monthly budget per household was reported to be Kenya Shillings (KShs) 5,360 (95% C.I. = 5,221–5,505) which equals US Dollars ($) 72.5 (KShs 72 = $ 1) but only 8% of all households (n = 1,312) were able to meet their monthly budget in cash. Notably, the households that were unable to meet their budget were equally distributed over all the socio-economic levels (χ² = 9.8; d.f. = 4; p = 0.056) indicating that higher socio-economic standard induces higher demand and expectations [60]. The average expenditure that could actually be met by households in terms of cash available was KShs 2,960 (95% C.I. = 2,848–3,068; S.D.: 2,007); 15% of all households lived below the poverty line [6] having less than KShs 1,500 ($ 21) per months. Households of higher socio-economic status met a significantly higher amount of cash per month (Table 1).

The respondents were asked about the perceived threats for life on Rusinga, most seriously felt diseases, malaria symptoms, most vulnerable groups to malaria infection, mode of parasite transmission and malaria prevention methods known. In FGDs, community members identified five major threats to life on Rusinga which were: droughts, diseases, no access to safe water, witchcraft and dangerous animals like snakes. For the diseases HIV/AIDS, malaria, diarrhoeal diseases, typhoid and tuberculosis were noted to be most prevalent. In the individual interviews, respondents were requested to rank these threats and diseases in descending order of importance and to name others that might be felt more important (Figure 1 and 2).

Most respondents felt that diseases are the most important threat to life on Rusinga, followed by drought, lack of access to safe water, witchcraft and dangerous animals; 10% of all respondents added famine and 5% added poverty as major threats to the list. Poverty featured in average on rank 5 while famine was ranked 1–3.

Diseases were ranked the most serious problem with 75% of the respondents ranking it first (Figure 1). Among the diseases, 67% of respondents ranked malaria as the most dangerous threat (Figure 2). The ranking of HIV/AIDS, typhoid, diarrhoeal diseases and tuberculosis was less consistent. Notably, there was a very indifferent view about the ranking of HIV/AIDS; only 21% of all respondents ranked it first and second, respectively; 29% of all respondents ranked HIV/AIDS the least important disease threat in comparison to the others. The ranking of life threats and disease importance did not differ with sex, age, CCF-enrolment status or educational level of the respondent.

Over 95% of respondents correctly identified headache, sweating, shivering body, high fever, joint pains, loss of appetite and vomiting as malaria symptoms but 88% of respondents associated malaria also with a running nose, 47% with body rashes and 32% with blood in sputum. Only 9% of all respondents mentioned all malaria symptoms correctly.

Regression analysis revealed that correct knowledge of malaria symptoms was dependent on the education level of the respondent but not the age, sex or CCF-enrolment. Respondents educated beyond primary school level were far more likely to correctly list the malaria symptoms (Table 3).

Among respondents, 96% knew young children to be at highest risk of suffering from severe malaria and 76% also knew of pregnant women being at increased risk. But 60% and 24% of the respondents also believed that young adults and old people were at high risk of suffering severe malaria, respectively. Other categories of people mentioned were fishermen, those who do not use bednets, tourists, dirty people and those who do not eat a balanced diet.

Of all respondents (n = 1,445), 5% did not know what caused malaria. Of those who indicated they knew (n = 1,378), 91% mentioned mosquito bites but more importantly only 47% correctly knew that was the only route for malaria transmission. A large proportion (44%) of the community believed in a number of causes in addition to mosquito bites and 9% of all respondents did not mention mosquitoes at all. Other major reasons believed to be responsible for 'catching malaria' (Table 4) were unfavourable weather conditions (cold temperatures, change of weather from cold to hot or vice versa, when one is rained on, sitting in the sun for too long, at times of new moon), lack of hygiene (drinking of dirty water, walking barefoot in dirty environment, unhygienic conditions at home, badly ventilated house, dirty utensils, dust, lack of a latrine or rubbish pit), a bushy compound (planting crops next to the house, bushes and high grass on the compound), food (raw, cold, contaminated or processed food) and body exhaustion (hard labour, no sleep, starving, fever).

Logistic regression analyses were used to identify the variables impacting correct knowledge and misconceptions/traditional beliefs (Table 5). Education beyond primary school level increased the probability of respondents knowing mosquito bites as the sole cause of malaria by 3–4 times.

Conversely, community members educated beyond primary school level were less likely to believe that unfavourable weather conditions was responsible for catching malaria. Older age groups were more likely to believe that malaria is caused by unhygienic living conditions and were also more likely not to mention mosquitoes involved in malaria transmission at all. Non-CCF-enrolled community members were less likely to mention mosquitoes and more likely to believe in weather as a cause of malaria than CCF-affiliated community members. There were no differences in knowledge and beliefs between men and women.

Table 6 illustrates how, according to community members, malaria parasites enter the human body. A large proportion (61.8%) of the respondents (n = 1,439) stated mosquito bites only; 7.6% believed in other ways in addition to mosquito bites, while 10.4% thought that mosquitoes had no role in the parasite transmission.

Notably, 20.2% of all respondents declared they do not know; this included respondents that mentioned mosquitoes as cause of malaria earlier. Other beliefs on how one could get infected with malaria parasites included consuming dirty water and food, or through cuts in the skin, and through ears and mouth. Of all respondent that knew that the malaria parasites can only enter the human body through mosquito bites (n = 890), 50% stated unfavourable hygienic and weather conditions as causes of malaria earlier. More surprisingly, more than half of the respondents that did not cite mosquitoes as cause of malaria earlier identified mosquito bites as the only way that malaria parasites can enter the human body indicating that despite the fact that there is a lot of knowledge in the community, this knowledge is distorted and biomedical relationships not comprehended.

Although over 60% of the interviewees correctly stated that mosquito bites are responsible for injecting the malaria parasite into the human body, none of the respondents was able to explain correctly where this parasite has been picked up by the mosquito. A common belief was that the parasite comes from various sources of water, bushes, dirty environments, food and air indicating that the cycle of malaria transmission via mosquitoes from a sick person to a healthy one has not been comprehended at all.

Evaluating the community knowledge of the mosquito life cycle 87% of the respondents (1,254/1,451) said they knew where mosquitoes lay their eggs. Places mentioned included stagnant water (79%), bushes and grass (6%), humid places (4%) and dark corners (3%). The majority (65%) of respondents did not know what mosquito larvae look like and those (35%) who attempted to define them talked of worm-like invertebrates, small toads, small insects, animals with big heads and small abdomens as well as twinkling reflections in water.

Knowledge of malaria prevention methods differed only slightly between households. Most respondents mentioned more than one method known to them as shown in Table 7; only 4% of all respondents stated not to know at all how to prevent malaria.

The majority of interviewees had a good theoretical knowledge of how to prevent malaria, with bednets most frequently mentioned. A summary of all methods which community members believed to be useful for preventing malaria is given in Table 7. Notably, clearing vegetation was the second most common method which community members believed to be useful to prevent malaria despite the fact that it is established knowledge in the scientific community that clearing of vegetation is of no benefit but might even worsen the malaria situation [63‐67].

Of those that responded to know how to prevent malaria (n = 1,398) only 34% listed solely biomedical correct methods to target either the mosquito adults (bednets, repellents), larvae (destruction of holes with stagnant water) or malaria parasite (drugs); but 72% listed only correct prevention methods with the one addition of bush clearing. Another 25% mixed correct knowledge and beliefs e.g. that increased hygiene, disposal of rubbish or keeping warm would help to prevent malaria. Only 3% of all respondents did not list any scientifically correct malaria prevention measure.

The probability of knowing that bednets could be used for malaria prevention was significantly higher in CCF-enrolled families than non-CCF families and also increased with increased levels of education. Moreover, people more than 45 years old were less likely to mention bednets than younger people (Table 8).

Source reduction was less likely to be mentioned by non-CCF affiliated community members and women compared to CCF-enrolled members and men, respectively, furthermore it was 3–7 times more commonly mentioned by respondents educated beyond primary level.

Whether respondents mentioned repellents and insecticide use for malaria prevention was confounded by their CCF enrolment status and age. The methods were less likely to be mentioned by respondents not affiliated to CCF and older age groups. Interestingly, the belief that bush clearing can prevent malaria was twice as high in families that were enrolled in CCF compared to non-CCF respondents and more importantly increased significantly with the higher the level of education. Women were less likely to mention bush clearing than men. The probability of believing in measures of general hygiene for malaria prevention increased with being a non-CCF-enrolled respondent and with older age but decreased with education beyond primary level.

The actual malaria prevention behaviour practiced by community members differed significantly from the methods known to respondents (Table 7). Despite the fact that 88% of all interviewees knew that bednets protect from malaria only 58% used them and in only 48% (692/1,451) of the interviewed households could the ownership of one or more nets actually be confirmed. Furthermore, only 37% of respondents (535/1,451) slept under a bednet the night before the interview, which was held during the main malaria transmission season. Of those, most (94%) said they sleep under a net throughout the year, while a few only used a net when mosquitoes are abundant. Children were the main beneficiaries of bednets: In 88% of the net-owning households children slept either alone or with their parents under the bednet. In total, 1,073 bednets were found in 692 households serving 4,419 people. This accounts for 1.5 bednets per household or 0.2 bednets per person in bednet-owning households alone.

Consequently, the average bednet coverage for the entire community was 0.7 bednets per household or 0.1 bednet per person which is not enough for a community-level effect [68]. Remarkably, of those respondents that correctly identified mosquito bites as the only means of malaria transmission (n = 642), 48% did not own a mosquito net while, in contrast, 40% of those that did not associate malaria with mosquito bites (n = 128) owned a net indicating that net ownership does depend on socio-demographic factors other than knowledge alone.

Bush clearing was the second most common method that community members practiced (492/1,451); 51% defined bush clearing as cutting down all the vegetation on the compound and burning it, 46% only slashed grass and 3% characterised it as collecting vegetation and empty containers from the surroundings. 56% of respondents said they cleared the bush in the last month. The activity was predominantly implemented by men. Of the 492 respondents that practiced bush clearing, 20% believed that bushes and other vegetation served as larval habitats for mosquitoes and 80% believed that mosquitoes hide in vegetation and can be prevented from entering the compound by removing it. Most (70%) said they learned that clearing vegetation prevents malaria in school. Another important source of this information was national and international health care organizations like local extension workers of the Ministry of Health and NGOs, including CCF and UNICEF. Equally worrying, radio and newspaper announcements on malaria control were also cited.

Despite that there was a very good knowledge in the community that the removal of water containing borrow pits protects from malaria, a fact recently highlighted with particular strength in nearby areas [69] nobody actually practiced this. Notably, 13% of all respondents did nothing to prevent malaria.

While knowledge of prevention measures might be confounded by demographic variables like the age and education level of the respondent, the actual usage of various methods might be decided by the household head and depend on the socio-economic status of the household [70]. The results of logistic regression analyses to investigate this are summarised in Table 9. Socio-economic status had significant impact on the probability of a household owning a bednet, boiling drinking water or not practice any prevention behaviour. Wealthier households were more likely to practice malaria prevention than poorer ones, thus they were more likely to own a bednet and buy fuel for boiling water. Insecticide use, like burning mosquito coils, was independent of the socio-economic status but women-headed households were more likely to practice this. Proper disposal of tins and rubbish was more likely to be practiced by men-headed households. Clearing vegetation in and around the compound was expectedly independent of the socio-economic status of the household but was more likely to be practiced by CCF-enrolled than non-enrolled families and men-headed households than women-headed households. Use of anti-malaria drugs was independent of CCF enrolment status, sex of household head and socio-economic status of the household.

Of all interviewed households (n = 1,451), 68% reported a child below the age of five years being sick in the last two weeks. The total expenditure for treating sick children ranged from KShs 2 to 10,000 with a median of KShs 250 (IQR 400). Of the occurring costs, 92% were spent on drugs and only 5.5% on laboratory and doctors' fees and 2% on transport indicating an even lower frequentation of health facilities in practice compared to theory. The remaining 0.5% was spent on herbalists and traditional herbs. 82% of respondents with a sick child felt that the disease had negative impacts on their daily life. Women were specifically concerned that they could not run their small-scale businesses and had to neglect farming and domestic work in order to take care of the child. There was no difference in health-seeking behaviour in relation to CCF-enrolment status, education level of respondents or socio-economic status of the households.

Most interviewees (939/1,430; 64%) said that malaria could be controlled on Rusinga island, while 34% (491/1,430) did not believe so, with 2% were undecided. Table 10 summarizes the reasons given for not being able to control malaria. Most of these were based on misconceptions or a fatalistic attitude. With increasing education level respondents were more likely to believe that malaria can be successfully controlled on Rusinga (Table 11).

A number of suggestions were made by respondents who believed that malaria control could be successful on Rusinga (n = 939) as to what needs to be done (Table 12). People most frequently mentioned that everybody needed to sleep under a mosquito net. This was closely followed by the belief that for successful control everybody on Rusinga would need to clear the vegetation from the compounds and thirdly the drainage of stagnant water and the treatment of larval habitats was mentioned.

Although a large number of community members thought that malaria can be controlled, there seemed to be little that the community felt they could contribute personally and without outside assistance. Moreover, the majority of contributions that were suggested would not even target malaria control (e.g. clearing vegetation, teaching to boil water, collect rubbish, constructing latrines; Table 13). A total of 12% (n = 243) felt they can not personally contribute at all. The most frequently associated benefit with malaria control was that the family would be happier because no-one would be sick and that time and money could be saved for other 'family projects'.

Community members expressed the need to be properly trained on how to prevent malaria during the FGDs and the individual interviews. Only 12.3% of all interviewees remember receiving malaria-related health education in the past, with CCF being the most frequently-mentioned training partner. In fact trained respondents were twice as likely to have been CCF-enrolled than non-enrolled community members (Odds ratio = 0.493 (for non-CCF), 95% C.I. = 0.356–0.682, d.f. = 1, p < 0.001). Other training partners mentioned were the Ministry of Health, churches, schools, the media and local CBOs. Major training components were general cleanliness, collecting tins, clearing vegetation and the use of bednets. Some of those trained said the training had helped them in different ways by encouraging them to drain stagnant water, buy mosquito nets and self-medicate. Others however, said they had financial constraints and had not implemented what they were taught. 72% of respondents were aware of the malaria project initiated by the RICFP but did not exactly know its goals; the majority demanded to be given free bednets and malaria medication through the project.