Long-lasting insecticide-treated bed net ownership, utilization and associated factors among school-age children in Dara Mallo and Uba Debretsehay districts, Southern Ethiopia

The study was conducted in Dara Mallo and Uba Debretsehay districts in Gamo and Gofa Zones. Both zones are found in Southern Nations, Nationalities and Peoples Region, which is the third populous region in Ethiopia. These districts are found in the western part of Arba Minch town, which is the capital of the former Gamo Gofa Zone (Fig. 1). Based on the 2007 national census, 150,145 people were living in the two districts, of whom 76,550 (51%) were male [25]. According to the recent update of the population by the respective districts, there were a total of 94,396 people in Uba Debretsehay district and 110,207 people in Dara Mallo district.

A cross-sectional study was undertaken by recruiting 2261 children from 32 schools in the study area. The number of participants involved in this study was about 98.1% of the estimated sample size. The sample size estimated was 2304 SAC and their households. The estimated sample size is used to evaluate the effect of malaria prevention education on malaria, anaemia and cognitive development among SAC in the study area. This sample size was estimated by using formula recommended to calculate sample size for cluster randomized controlled trials [26, 27] for each of the two parallel arms. The assumptions used were 11.04% and 18.4% prevalence of malaria with and without the intervention, respectively; 95% level of confidence, 80% power, 72 cluster size and 0.025 intraclass correlation coefficient (ICC) with the calculated design effect of 2.775 and 15% loss to follow-up or non-response rate.

A total of 3204 children attending primary education were approached at 32 primary schools. It had been planned to recruit 32 schools by using simple random sampling technique, but only 32 primary schools in malarious areas in the two districts could be recruited. Seventy-two children from grade 1 to 3 were selected using systematic random sampling technique from eligible children in each section of the students with a class roster as the sampling frame. All participants involved in the trial were included to address the research objective. The number of participants in the trial from each grade level (grade 1 to 3) was determined by the relative contribution of each grade level to the total enrolment of students in each school. Children who were attending their education in the schools during the data collection period were included in the study while those mentally not fit to respond to questions directed to them and with physical problems to measure their height were excluded from the study. Children enrolled to the study were used to trace their households. Participants’ households were approached by trained data collectors for interview. A pretested, structured questionnaire was used to collect data on demographics, water source, toilet structure, household assets and bed nets. The questions were adapted to local context from the Demographic and Health Surveys (DHS) malaria indicator survey household questionnaire [28]. The questionnaire was uploaded to tablets in Open Data Kit (ODK) data collection tool. The data collectors were trained on how to use the data collection tool and ethical procedures. They interviewed mothers or caregivers in conditions when the selected child mother was not alive, observed the toilet structure and the number and position (hung over sleeping place or not) of bed nets in the households.

Four indicators of malaria prevention measures by bed net utilization in the current study were used, but ownership and utilization of bed nets by SAC conditional to presence of at least one bed net in the household were dependent variables.

Indicator I: ownership, which is defined as SAC households those had at least one LLIN among SAC households participated in the survey.

Indicator II: utilization, which is defined as proportion of SAC passed the previous night under LLINs irrespective of ownership of LLINs.

Indicator III: conditional utilization, which is defined as proportion of SAC passed the previous night under LLINs conditional to presence of at least one LLIN in the household.

Indicator IV: access, which is defined as proportion of SAC households those had at least one bed nets for two household members among those owning at least one LLINs.

Indicator V: utilization conditional to access, which is defined as proportion of SAC passed the previous night under LLINs conditional to bed net access in the household [29].

In addition to SAC related indicators, any person from the household passed the previous night under bed net irrespective of presence or conditional to presence of at least one LLIN in the household were also assessed.

Indicator VI: any membered utilized, which is defined as any member of SAC household that passed the previous night under bed net among all surveyed SAC households.

Indicator VIII: conditional any membered utilized, which is defined as any member of SAC household that passed the previous night under bed net conditional to presence of at least one LLIN in the household.

The data collected by using ODK data collection tool with tablet apparatus were converted to Microsoft Excel using ODK Briefcase. These data are imported to RStudio version 4.0.0 statistical software and STATA version 14 to predict the principal components (PCA). The variables used in predicting the wealth index of the household were the type of toilet used by the household; the source of drinking water for the household; the presence of sponge mattress in the household, the floor of which the house is constructed; the outer cover of the house, the total number of milk cows in the house; the number of other cows in the house; the number of goats and the number of poultry in the households. The predicted index from PCA is categorized into 4 quartiles to determine the wealth index of the households. Cleaned data were subjected to descriptive statistical analysis, such as proportion for categorical variables and plots and descriptive measures for continuous data. The survey data used in this study indicate a hierarchical nature in which children included were nested within the schools. Children within the same school will be similar to each other than students in another school. Thus, mixed effects multilevel logistic regression with random intercepts by schools for the outcome variables was used to assess the individual level fixed effect of correlates on the outcome variables. In the analysis of factors influencing ownership and use of bed net glmer command in lme4 R package was used. In the analysis of factors influencing bed net utilization, children with at least one bed net in their household were included. Odds ratio (OR) and corresponding 95% level of confidence (CI) were used to assess the strength of association between bed-net ownership or utilization and potential predictor variables. Variables that are significantly associated with bed net ownership and utilization in univarite multilevel logistic regression were taken to multivariate multilevel logistic regression. The fit of the mode in predicting the outcome variables were checked by Akaike Information Criterion (AIC). The averages of the household coordinates taken by global positioning system (GPS) were used to represent the coordinates of each school. The geospatial data was added to Quantum Geographical Information System (QGIS) to present the ownership and utilization percentages at school level.

The trial was approved with written consent procedure to be followed by the Institutional Research Ethics review Board (IRB) in College of Medicine and Health Sciences, Arba Minch University with the reference number of IRB/154/12. Official permission letter to conduct the research was submitted to district health offices and education offices in Dara Mallo and Uba Debretsehay districts. Support letters written by the respective education offices were given to each participating school and written consent was obtained from school head teachers. Parents of the selected SAC were invited to come to the schools and written consent was obtained before data collection at school and household level, which was documented in College of Medicine and Health Sciences, Arba Minch University.

Ownership of at least one LLIN at household level was 19.3% with 95% CI of 17.7–21.0%. The mean number of bed nets in a household owning bed nets was 1.8 (SD = 0.9), while 45 of the participated SAC households had adequate numbers of bed nets (at least one bed net for 2 household members). Overall ownership of at least one bed net and adequate access per district were indicated in Table2. There was wide difference in ownership of bed nets in households per school where children attend education (Fig. 2).

In univarite multilevel logistic regression analysis, the ownership of bed nets was affected by characteristics related to place of residence, household head occupation, maternal (caretaker) occupation, household head education, maternal (caretaker) education, maternal (caretaker) education level and socio-economic status. SAC households living in semi-urban area, literate educational status of the household head, literate educational status of mother (caretaker), education level of mother (caretaker) higher than grade 6 and being employee occupational status of the mother (caretaker), being employee or other occupational status of the household head and being in the 2nd or 4th quartile of the household wealth index positively and statistically significantly affected ownership of LLINs by SAC households. However, after fitting the model to multivariable multilevel logistic regression, none of the above factors significantly influenced ownership of LLINs but being in semi-urban area (AOR = 2.6; 95% CI 1.0–6.9) and employee as occupation of the household head (AOR = 1.8; 95% CI 0.9–3.7) marginally significantly affected ownership of LLINs by SAC households. The crude and adjusted ORs and the corresponding 95% CI for each of variables were presented in Table 3.

The percentage of any person passing the previous night under a bed net from the total participating households and households conditional to bed net ownership was 12.0% (95% CI 10.7–13.4%) and 62.2% (95% CI 57.4–66.7%) respectively. About 40.3% (95% CI 35.8–45.1%) of participating SAC in the households owned at least one bed nets and 7.8% (6.7–10.0%) of all participating SAC passed the previous night under a bed net irrespective of bed net ownership. The percentage of bed net utilization among SAC conditional to adequate access of bed net in their household was 66.7% (95% CI 50.9–79.5%). The frequency, percentage and 95% CI of bed net use to prevent malaria and other indicators per the study districts was depicted in Table 2 while percentage of bed net utilization conditional to at least one bed net in the household per study schools was depicted (Fig. 3).

Passing the previous night under a bed net by SAC in univariable multilevel logistic regression analysis was affected by place of residence (COR = 4.1; 95% CI 1.5–12.4), employee being occupation of the child mother (caretaker) (COR = 3.4; 95% CI 1.4–8.1), education level above grade 6 for the child mother (caretaker) (COR = 3.9; 95% CI 1.8–9.4) and household size to bed net ratio less than or equal to 2 (COR = 4.0; 95% CI 1.8–9.8). However, after fitting the model to multivariable multilevel logistic regression, household size to bed net ratio less than or equal to 2 (AOR = 20.7; 95% CI 4.7–132.5) and child mother education level above grade 6 (caretaker) (AOR = 3.4; 95% CI 1.3–9.3) remained statistically significant. The crude and adjusted ORs and the corresponding 95% CI for each of variables assessed for bed net utilization of SAC was presented in Table 4.