Setting
Chano Mille village is one of the rural, malarious areas near Arba Minch town, 492 km south-west of Addis Ababa. The altitude is 1,206 m above sea level. The village was selected purposively to study malaria epidemiology in detail in the presence of favourable malaria vector breeding site. The presence of Lake Abaya to the south-east of the village resulted in intense malaria transmission since the shore of the lake favoured malaria vector breeding. The incidence rates of falciparum and vivax malaria in the village were 22.9 and 22.2 per 1,000 persons per year, respectively. The distance of the household from the shore of the lake, wealth index, age and gender were found to be significant predictors of malaria infection. In the two-year study period, the government had undertaken indoor residual spraying with insecticides (twice) and mass free ITN distribution (once) as prevention and control measures [16, 17].
Study design and data
This was a prospective cohort study that involved all residents of the village. The total number of households was 1,388 and the total number of individuals followed for 101 weeks (from April 2009 to April 2011) was 8,121. Every week, individuals were asked whether they slept under an ITN the night before the interview; and if they did not use the ITN, open-ended question was used to ask the reason why. To maintain a gap of six days between the visits, households were visited on the same day each week. A census was carried out three times to update the denominator: at the start of the study, on week 50, and at the end of the study. For the first four weeks, ITN use data were collected considering vulnerable groups, including children under five years and pregnant women. After week 5, all residents of the village were considered and the name of the individual who slept under an ITN was recorded. Weekly ITN use fraction was calculated by taking the number of individuals who slept under an ITN as numerator and the total population of the week as denominator; this was done for different gender and age categories as well. The number of weekly follow-ups in which ITN use was reported was calculated for each of 8,121 individuals. The number of bed nets available at each household was recorded at the beginning. In addition, after free mass distribution of ITNs, which was carried out by the government on week 48, ITN coverage survey was carried out on week 50. During the ITN coverage survey, the households were asked if they had usable ITNs in addition to the new ones; and when available, these ITNs were considered old-functional.
The data collectors were recruited from the village having college level diploma.
Data analysis
Summary statistics were used to report the number of bed nets (new and old-functional) available in each household, and the proportion of children aged less than five years, and pregnant women that did not use ITNs. Likewise, the median number of weekly follow-ups in which ITN use was reported was calculated for different population sub-groups. A summary was provided on the reported reasons for not using ITN.
The count data on the number of weekly follow-ups, in which ITN use was reported, was over dispersed while Poisson regression was fitted. The ratio of the deviance over the degree of freedom was 24.4. This value became 1.6 with a negative binomial probability distribution model. As the later model handled the overdispersion problem (since the value was very close to 1), a negative binomial regression model was fitted to the data. The number of weeks an individual was observed was set as a scale weight variable. A fixed value of 1 was used as a scale parameter method and robust estimator was used for the covariance matrix. Exponential parameter estimates were interpreted as incidence rate ratios (IRR). The 95% confidence intervals (CI) for the IRR were also reported. Gender, age, education of the household head, wealth tertiles and distance (in km) from vector breeding site were considered as determinants for ITN use. To construct wealth index, principal component analysis (PCA) was used. The variables included were presence of electricity, watch, TV, radio, mobile phone, refrigerator, separate room used for kitchen, bicycle, agricultural land, livestock, account in bank or credit association and latrine facility. In addition, the main materials of the floor, wall and roof were considered. The details of wealth index construction are reported elsewhere [16]. Distance of each household (in km) from the identified vector breeding site was calculated using proximity analysis tool of ESRI ArcMap 9.3 (Redlands, CA, USA). Statistically significant independent variables during bivariate analyses were used to construct the multivariate model. Pairwise comparison was done for age categories using sequential Sidak as adjustment for multiple comparisons. PASW 18.0 (Chicago, IL, USA) was used for analysis.
Ethical clearance
The Southern Nations and Nationalities Regional Health Bureau Ethical Review Committee approved this study. Permission and support letters were obtained from relevant administrative bodies of the area. Informed verbal consent was obtained from each household.