Study setting
The school-based SBCC project was implemented from 2017 to 2019 in 75 selected rural schools of Jimma zone; the State of Oromia. Jimma is located 352 km away from Addis Ababa; the capital city of Ethiopia. Based on the projected 2007 Census conducted by the CSA, the total population of Jimma zone was 2,486,155 (50.3% male and 49.7% female) and the rural population accounts for more than 89%. The zone lies within an altitude ranging between 900 and 3500 meters above sea level.
The current intervention was conducted in five selected districts of Jimma Zone for intensive engagement on malaria communication. These districts were Limmu Kossa (population=209261), Shebe Sombo (population=146805), Nono Benja (population=77452), Chora Botor (population=74756), and Gera (population=147120). According to the Zonal health department report of 2016; an annual parasite incidence (API) rate in these districts was 16% in Chora Botor, 14.1% in Shebe Sombo, 10% in Nono Benja, 5.5% in Limmu Kossa, and 3.1% in Gera. The project was dedicated to benefiting schools and various community groups including vulnerable groups such as children less than five years, pregnant women, and school students.
Theoretical bases to inform the school-based SBCC interventions
To guide the SBCC content and interventions process, concepts, and principles drawn from some behavioural change theories or models were combined and applied. Accordingly, the Motivation Protection Theory (MPT) and the health belief model (HBM) were used to design the proposed SBCC elements [21]. These theories explain the cognitive mediation process of behavioural change in terms of threat and coping appraisal. According to the theories, the appraisal of the health threat (i.e. health risk due to malaria in this case) and the appraisal of the coping responses result in the intention to perform adaptive responses which are called protection motivation (i.e. using malaria preventive measures, such as ITNs, IRS, prompt care-seeking, and proper use of medications).
Thus, the theories propose that the intention to protect oneself or families from certain conditions (i.e. malaria) depends upon four factors: 1) the perceived severity of a threatening event (e.g., a malaria attack), 2) the perceived probability of the occurrence, or vulnerability ( e.g. perceived vulnerability of the individual to a malaria attack), 3) the efficacy of the recommended preventive behaviour (e.g. perceived effectiveness of recommended actions to prevent or remove the health risk, malaria in this case) and 4) the perceived self-efficacy (i.e., the level of confidence in one’s ability to undertake the recommended preventive behaviour, such as regularly sleeping under ITNs, prompt care-seeking behaviour and properly using medications) [21]. To this end, formative assessment was undertaken in the target districts based on the assumption and framework of the MPT and HBM. The results obtained from this formative assessment were used to guide the malaria communication activities and to monitor behavioural change progress indicators in schools and target villages.
The second theory, the theory of diffusion of innovation (DOI), was applied to complement the individual based theories. DOI is one of the most widely used communication theories. According to the DOI, the population can be broken down into five different segments, based on their propensity and time it takes them to adopt a specific behaviour (e.g. adoption of recommended malaria prevention measures in this case). These are innovators, early adopters, early majorities, late majorities, and laggards and people in each category have different needs, perceptions and require tailored interventions [22]. In this intervention, early adaptors were considered a role model for other group members after receiving basic malaria training.
Success stories and experiences of these role models were captured and used for educational purposes to motivate other students and family remembers. People in each category of adoption need different interventions ranging from sharing simple facts to implementing groups’ norms [22]. Thus, the due emphasis was given on promoting social and groups’ norms rather than just the health benefits of interventions and emphasizes the risks of being left behind for those who are late and laggards to adopt the behaviours. Thus, the group members reinforce each other, and households who do not practice the recommended behaviour begin to model a new behaviour and change themselves as a result of pressure from the group members and social networks. Household status and student behaviours were monitored and evaluated, and tailored education was provided accordingly.
Descriptions of the intervention
The current school engaged peer education combined with SBCC intervention was designed to facilitate behaviour changes on malaria prevention and control targeting various levels of personal, organizational, and community factors. Ultimately, it was intended to promote the five key malaria prevention and control practices both at schools and community levels. These were the use of insecticide nets (ITNs), appropriate & timely seeking care for malaria, appropriate use of quality anti-malarial drugs, acceptance of insecticide residual spray (IRS), and draining of potential breeding sources in the villages.
The interventions encompassed various capacity building and educational sessions that were implemented from 2017 to 2019 engaging 75 schools and respective villages. The programme was first initiated through participatory consultations of stakeholders or representatives of the community including key peoples from health offices, education offices, health extension workers (HEWs), and village leaders and schools. The formal supervisory committee was organized before the actual joint situation analysis that identified malaria situations, the interventions' needs, and strategies. Based on the need assessment results, joint planning (i.e. identifications of roles, developing goals/objectives, devising monitoring, and evaluation mechanisms). Finally, the plan was implemented over two and a half years through active engagement of the community, health institutions, and primary schools. A summary of the intervention process is presented in Fig.1. Furthermore, details about the intervention are provided in another publication [23].
Study design
The study employed a quasi-experimental evaluation design to collect post-intervention data from selected primary school students (i.e. grade 6th through grade 8th). A post-intervention quasi-experimental design was most widely and effectively used in impact evaluation of large scale interventions [24]. Students in intervention schools were considered exposed (intervention group) and those selected from non-intervention schools are comparison/control group. Controls were selected from adjacent schools to the project area of the same cluster.
Study populations and sample size
All grades 6th through 8th students in randomly selected schools from both the intervention and controls were considered to study population and included in the study. The sample size was calculated using two population proportion [25] given by where E= P1-P2 and r= the ratio of the two proportions. It was assumed to detect the effect or odds ratio two (OR=2 or greater) for 90% power (Zβ =0.96) and 5% level of significance (Zα/2=1.96) with an equal number of intervention and comparison groups. The P2=population prevalence of ITNs use among school children was 46% which was taken from the previous study [26]. This yields a sample size of 380 (190 each intervention and control groups) was calculated. Considering a factor of 2 for sampling variation or design effect and 5% for non-response rate, the final sample of 798 (399 each intervention and control groups) was drawn.
Sampling techniques
Seventy-five schools in five districts were addressed by the intervention. A total of 4 schools per each district (i.e. 4*5=20 schools) were randomly selected from the intervention village. Two corresponding schools from non-intervention village (i.e 2*5=10 schools), but in the same district were randomly selected. Stratification was further done to distribute a 399 sample to each school and grade levels through grades 6th to 8th assuming an equal number of students in all schools. This was done by dividing 399/20 for intervention and 399/10 for control, which gives 20 and 39 students per school respectively. Down stratifying to grades level, 6th through 8th=3, (i.e 20/3=7 and 39/3=13) for intervention and non-intervention schools respectively. Finally, 399 students each from 20 interventions and 10 non-intervention schools were interviewed (Fig. 2).
Data collection tools and methods
Malaria related data were collected using a questionnaire adapted from relevant literature such as malaria indicator survey and health, and demographic survey [27,28]. The questionnaire form covered socio-demographic factors, peer education experiences, behavioural outcomes such as ITN use, and psychographic outcomes such as knowledge, risk perceptions, self-efficacy, and attitude related to malaria and its preventive measures. The questionnaires were then translated into the local language, Afan Oromo, before data collection. Data was collected on an interviewer-administered basis. Qualified data collectors recruited ad three days of training were given to data collectors and supervisors about the purpose of the study, instruments, and data collection procedures. The data collection process was closely supervised by the research teams. Eligible students were called on by schools’ administrators to appropriate places for the interview which was conducted face-to-face.
Variables and measurements
The current study measured two outcomes of interest and these were measured at the end of the intervention. The primary outcome of interest was the ITNs utilization. The secondary outcomes of interest were the psychographic outcomes that include multidimensional knowledge, attitude, family supports, self-efficacy, and perceived malaria risk and severity. The psychographic outcomes or variables are conceptualized in this study as mental processes such as attitude, perceptual process, and beliefs about an individual’s behaviours or practices in the context of malaria preventive behaviours.
Knowledge: Multidimensional knowledge (MDK) questionnaire was used to measure comprehensive malaria knowledge related to the cause of malaria, signs/symptoms, vulnerable groups, preventive measures, and mosquito vectors biting behaviours. Questions encompassing 31 items were used [29]. For each item, the correct answer was assigned (1=Yes) and an incorrect answer as (0=No). Scores of correct responses were computed for subsequent analysis.
Attitude: This is evaluative beliefs or acceptance and benefits towards the ITN, indoor residual spray, vulnerable groups, and malaria situation in the study area. The items were scored on a five-point Likert scale ranging from strongly disagree (1) to strongly agree (5). The reliability statistics or internal consistency of the items were equals, α=0.72. The overall attitude score was computed by summing up the items after performing the reverse scoring for negatively worded sentences. A higher composite score indicated a more favourable attitude.
Perceived malaria risk: This was defined as individuals’ perception of vulnerability to malaria in the context of their daily experiences about the presence or absence of malaria, individuals who suffered from malaria in the villages. It was measured using seven items on a five-point Likert scale, (α=0.76). A higher composite score was interpreted as a higher perception of risk.
Perceived severity of malaria: This represented the perceptions of bad consequences of the malaria infection in causing pain, death, interruption with daily activities such as schooling, and its impact on their academic performances. Four items were used to tap this construct by using a five-point Likert scale, (α=0.84). Reverse scoring was done for both perceived risk and severity before conducting further analysis.
Self-efficacy: Self-efficacy is defined as an individual’s confidence or beliefs about the ability to prevent themselves from malaria infections under given circumstances by using ITNs or any other preventive measures. It addresses aspects of malaria prevention such as using ITN, application of IRS despite its perceived discomforts, ITN handling, early care for fevers and drug adherence, etc. Four items were used to measure it. Scoring and computing were all were done the same way. The items showed an acceptable level of internal consistency with α=0.89.
Perceived family support: This was conceptualized as perceptions about how much their family (parents and siblings) are motivated to use ITNs or IRS as preventive measures, advice and encourage the individuals on ITN access and use, and help them handle the ITNs for effective and sustained use. Ten items were used to measure this dimension using five-point Likert scales. The measure of internal consistency or reliability statistics was acceptable level, α=0.84. A higher score was interpreted as strong perceived family supports.
Utilization of ITNs: The access to, and use of ITNs was assessed by three items that include the presence of ITNs in the home, number of ITNs, and ITN utilization every night. It was coded as Yes=1 if the student used the net every night and otherwise No=0. Access to ITNs (ratio of ITNs) was defined as the presence of at least one ITN per two individuals in the household [8].
Social desirability bias (SDB): is defined as a tendency to portray oneself in a socially desirable manner [30]. In this study, it is believed that the SDB might exist especially among respondents in intervention schools as they were aware of the aim of the intervention. They could respond to the interview in a socially desirable manner; thus hiding reality. Consequently, it was interesting to measure the SDB to analytically adjust for its confounding effects on the outcomes. This was measured using the SDB for children scale that consists of 20 items constructed such that Yes=1 if the condition exists and No=0, otherwise [30]. The continuous score was computed from all 1’s (correct) responses. A high score was interpreted as the presence of a high SDB. This score was adjusted in every subsequent analysis and considered during the interpretation of the result.
Statistical data analysis
Data were analysed using statistical package for social sciences (SPSS) version 26 software for analysis. Means, standard deviations, frequencies, and proportions were calculated as descriptive analysis. The propensity score matching (PSM) technique was performed as a matching analysis to reduce the selection bias due to a lack of randomization in this study. The PSM was recommended in evaluation studies where observational data are used to adjust for the possible differences between intervention and control groups at baseline [31]. The method was found effective in reducing biases in observational and quasi-experimental studies [32]. A study conducted to evaluate the effectiveness of community-based SBCC interventions on the use of ITNs demonstrated that PSM approached was productive [33].
The propensity scores were predicted or computed based on the selected covariates or characteristics that may lead to the possible differences at baseline. Assuming that there are no unobserved differences between the groups (i.e. satisfying assumptions of ignobility), all socio-demographics and socio-economic predictors were included in the matching procedure that are known to be associated with both intervention assignment and the outcome [32]. Accordingly, ten covariates that include age, gender, grade point average (GPA), grade level, ethnicity, religion, schools, specific roles of the student in the class (e.g being leader, deputy leader or member), and participation in school wide-clubs (e.g sport club, health club, technology clubs) and specific roles in the clubs (e.g being leader, deputy leader or member) were selected to compute propensity scores.
Multivariable logistic regression modelling was used to estimate the propensity scores. The test of normality of the predicted scores was done to evaluate the balance of the matching result for the goodness of model fit. The matched sample was used in the subsequent analysis. Participants were matched using the one-to-one nearest neighbour algorithm by imposing tolerance level or calipers of width equal to 0.20 of the standard deviation of the estimated propensity scores [34]. Individuals scores not falling within this specified distance were excluded from the sample [31].
The average effect of the intervention on psychographic outcomes (including knowledge, attitude) was estimated using multivariate general linear modelling. The adjusted mean differences and effect size was calculated to examine the effects of the interventions on these psychographic outcomes. An odds ratio was calculated to estimate an average effect of the intervention for binary outcomes (e.g ITNs utilization). The analysis was adjusted for the confounding effects of the SDB and predicted propensity scores. The predicted propensity score was included in the subsequent analysis to adjust for the covariates it has been represented in matching analysis. Covariates that were included in the estimation of propensity score were excluded from this analysis. A p-value of less than 5% was considered to carefully signify the presence of association or difference.