A sample size of 497 households was calculated using Cochran formula (13). A non-response rate of 10% and a proportion of 30% with knowledge of mosquito LSM methods were assumed based on a similar study in rural Kenya (14) A design effect of 1.4 was calculated based on a cluster size of 15 and a rate of homogeneity of 0.025 (15). Level of precision was set at 0.05. Sample size was adjusted to 500 to maintain same number of households per cluster (20 households/cluster * 25 clusters). However, 479 households were interviewed because 21 were not available. A two-stage cluster random sampling procedure was used. In Stage 1, twenty five (25) clusters were randomly selected from 43 villages that form Nthache area using probability proportional to population size. Each selected village formed a cluster or two clusters depending on population size. Prior to random selection of clusters, a sampling interval (SI) was calculated after dividing the total population in study area (33,870) with the 25 clusters. Cumulative population sizes for the villages were calculated. These were sizes of the population for each village plus the sum of all villages which came before it on the list. A range for each village according to its cumulated population was developed. To determine clusters, a number 52 was randomly selected between 1 and SI (1,354.8). The village with a range that 52 fell was identified as the first cluster. Then, the SI was added to 52 to determine the second cluster. This process continued as SI was added to the immediate calculated result until 25 clusters were identified (13). In stage 2, village registers from village heads were used to assign identification numbers to households in clusters. The identification numbers were put/mixed in a pot. The village heads or their representatives randomly selected 20 households per cluster.
A validated household questionnaire was used to collect data. Respondents were asked questions on socio-demographic characteristics, knowledge, perceptions and practices related to specific mosquito LSM methods. An adult member in the household was interviewed (> 18 years-old). In child-headed households (< 18 years-old) the heads were interviewed. Household heads or an adult from the sampled household who had consented to respond to survey questionnaires were included.
Sex of the respondent was classified and coded as female, 1 and male, 2. Data on age was collected as a continuous variable and categorized as well as coded into a binary variable of ≤ 35 years-old, 1 and > 35 years-old, 2. Education status was classified and coded as primary, 1, secondary, 2, none, 3, and informal/pre-primary, 4. Marital status was classified and coded as married, 1, single, 2, widowed, 3 and divorced/separated, 4. Pregnancy status was classified and coded as being not-pregnant, 1 and pregnant, 2. Occupation was classified and coded as crop farming, 1, mixed-farming/pastoralist, 2, business/self-employed, 3, unemployed/student, 4, employed, 5, housewife, 6 and other, 7. Household ownership was classified and coded as owned, 1 and rented, 2 whereas household floor was categorized and coded as natural/earth, 1 and cement/tiles, 2. Household roof was classified and coded as grass/thatch, 1 and ironsheets, 2 whereas wall type was classified and coded as brickwall, 1 and mudwall, 2. Energy used was classified and coded as firewood, 1 and charcoal, 2.
Level of knowledge was measured as a binary variable (1 - high and 0 - low). A scoring system known as knowledge score was developed to assess the level of knowledge. To score full points, respondents had to mention four methods: draining stagnant water (1 point), larviciding (1 point), clearing grass/bushes (1 point) and clean environment (1 point). The total score was 4 points. A mean score of 1.7 was calculated. Respondents with scores above this were deemed to have high-knowledge whereas those below it were deemed to have low-knowledge (16). Respondents were asked to mention specific mosquito LSM methods. They could provide multiple responses from this list: draining stagnant water, larviciding, clearing grass/bushes and clean environment. Those who stated correct answers were deemed as having knowledge of those specific methods and the remaining as not having knowledge. The responses were coded as 1 - “having knowledge” and 0 - “not having knowledge.” Respondents were asked to mention one specific mosquito LSM method they perceived as the most effective for malaria control among those they initially expressed knowledge of. The responses were coded as 1 - “positive-perceptions” and 0 - “not positive-perceptions.” Respondents were asked to mention specific mosquito LSM methods they practice for malaria control they initially expressed knowledge of. The responses were coded as 1 - “practiced” and 0 - “not practiced.”