2.1. Study locations and context
Ethiopia is located in the Northeastern part of Africa, known as the Horn of Africa, and occupies an area of 1.1 million square kilometers ranging from 4,620 meters above sea level at Ras Dashen Mountain to 148 meters below sea level at the Danakil depression (28). It possesses three major topographic-induced climatic zones, the hot lowlands (“Kolla”) located below 1500, the temperate (“Wayna Dega”) which range 1500-2400, and the cool temperate highlands (“Dega”) located above 2400 meter above sea level (28, 29). The mean annual temperature is around 15-20°C and 25-30°C for highlands and lowlands respectively (29).
This trial was conducted in a low-income rural community of the Mecha Health and Demographic Surveillance System (MHDSS) site. MHDSS site is a field research center established in 2013 by Bahir Dar University to conduct and support postgraduate level studies in the region. It is located 525 km away from the capital city of Ethiopia, Addis Ababa, towards Northwest and 40 km far away from the capital city of Amhara Regional State, Bahir Dar. According to the official population profile report of MHDSS, the study area comprises 132 clusters/ “Gots” with a total of 65, 086 populations within 20631 households at the end of 2016. Out of which children less than 4 years old were accounted for 13.3% of the total population. Biomass-fuel burning in open traditional cookstove was the major (94.5%) household energy source in the study locality (23, 30).
2.2. Trial design
With a longitudinal experimental design, a community-level cluster randomized controlled trial study with two arms of equal size was used to compare the effect of biomass-fuelled IBS intervention with TBS on childhood ALRI at an individual child level. Clusters were the small villages, termed as “Gots” in Amharic (both national & local language), are the lowest administrative units in Ethiopia, and used as the smallest unit of enumeration areas by the Ethiopian national census authority. Each cluster (“Got”) was comprised about 55 eligible children on average, and all eligible children in the selected clusters were enrolled as control or intervention for baseline and repeated follow-up visits approximately every 3 months for one year after receiving the intervention. With this design, childhood ALRI outcome was measured before installation of IBS, and again in the same households, 4 times after the intervention households received the IBS. The households with TBS method were served as a control arm.
2.3. Eligibility criteria
All clusters/ “Gots” under MHDSS site were eligible for participation in the trial; and to ensure a minimum of one year repeated visit for longitudinal data collection before the child’s 5th birthday, all households with at least one less than 4 years old child and who were exclusive users of TBS were eligible for participation in the trial. Merely households who did not have any enclosed baking quarter (kitchen) structure and children who were born during the course of the study were excluded.
2.4. Sample size determination
To estimate the effect of IBS intervention on child ALRI risk reduction over one year follow-up period, the sample size was calculated by applying the two-sample comparison of proportions formula using STATA. We calculated the required sample size by assuming childhood ALRI prevalence of 21% from a previous study in rural Ethiopia (31) and 0.05 two-tailed alpha to detect a 25% reduction in the prevalence of childhood ALRI with a power of 80%. Accordingly, the estimated sample size (n1) was 891 children per arm with equal numbers to both arms under individual randomization.
However, since this trial was randomized the intervention over clusters instead of individual households, the sample size was calculated using the standard formulae for unequal size cluster randomized controlled trials as:
nc = n1 [1+ (m’-1) ICC] (32).
Where, nc = sample size with cluster randomization, n1 = sample size under individual randomization, m’ = an average number of eligible children within each cluster, and ICC = intra-cluster correlation coefficient for cluster level ICC of childhood ALRI.
Thus, the required sample size (n) was 2,068 per arm under equal allocation by assuming an ICC value of 0.03; which is an estimated value for child pneumonia outcome in the cookstove intervention study (13), and an average number of eligible children of 55 within each cluster from the updated data of MHDSS.
The number of clusters (K) required in each arm for unequal cluster sizes was also determined using the formula:
K = n[1+((CoV2+1)m’-1) ICC]/m’] (32)
Where, CoV = coefficient of variation.
The number of clusters per arm became 40, and this was caused to increase the sample size to 4,400 (2,200 per arm) by considering an effective ICC value of 0.03 for cluster-level ICC value of childhood ALRI, a CoV value of 0.3 for cluster size and an average number of eligible children of 55 within each cluster.
As a final point, considering the nature of cluster sampling method, we added 25% of the sample to account for lost to follow-up (LTF) and any other unpredictable events in the field (33, 34), the required sample size (n) was increased to about 2,750 within 50 clusters per arm, which would provide a total of 100 clusters containing about 5,500 eligible children randomized in equal numbers to both arms.
2.5. Randomization and masking
Clusters were randomly allocated to intervention and control arms at a 1:1 ratio by an independent epidemiologist using a computer-generated randomization schedule, which was revealed after all baseline measurements had been completed as well as all study households recruited and assigned to their respective arm to ensure the allocation sequence was concealed from those assigning the arms. Both study participating households and data collectors were blinded to intervention status during study enrollment and baseline data collection. In addition, all eligible children/households within the clusters were included in the study to minimize the risk of selection bias; however, because of the distinctive feature of cluster design and nature of the intervention under study, blinding of the households receiving the IBS intervention was not possible.
2.6. Sampling method and recruitment of participants
The cluster sampling method was used to randomly select 100 clusters (50 clusters per arm) among the total 132 clusters in the MHDSS site, and all eligible households were included within the selected cluster (complete enumeration). The sampling frame, the list of clusters and the potential eligible households with less than 4 years old child, was established from the MHDSS record. Then, the eligible households were chosen from the record and a child less than 4 years old was recruited from each household. In situations where there were two or more children less than 4 years old living in the same household, only the youngest child was included in the study.
The selected households were identified using the permanent MHDSS site house number and the actual participants were recruited at the household level by field workers during the baseline survey after ensuring whether the households met the eligibility criteria. A screening questionnaire was used by field data collectors upon their first visit to each household to ensure that the household was appropriate and willing to participate. When the household meets the eligibility criteria, the study was explained to the index child parents, and they were asked whether the household is willing to participate in the study and use improved baking stove technology for at least 12 months.
Then, when the parents of the household agreed to be involved in the study, the field staff administered a written consent form at that time and the consent procedure was conducted in Amharic (both national & local language). It was also explained that the allocation to intervention or the control group was based on the concept of a “lottery” method, and several special efforts were applied during recruitment to facilitate the enrolment process as described within our earlier research report (23), and there was no household unwilling to participate in the study.
2.7. Intervention
2.7.1. Trial descriptions
Replacing the open burning TBS with “Mirt” IBS, the well-known commercially distributed type of IBS in Ethiopia, was the intervention for this study. The “Mirt” IBS intervention can save a considerable quantity of fuel-wood compared to the TBS method (35), and it is a biomass-fuelled without chimney stove designed by the Ethiopian Energy Studies Research Center for cooking the staple food of Ethiopia called “Injera” (i.e., a unique type of yeast-risen flatbread, consumed widely in Ethiopia) (21, 22).
All households who were randomized to the intervention arm were identified using the permanent MHDSS house number for a convenient appointment date, and the intervention was delivered to all eligible households in 50 randomly allocated clusters at the beginning of the study period, and the control households were continued to use the customary open burning TBS method equally in 50 randomly allocated clusters. All the trial cookstoves were manufactured by a local licensed firm and installed on-site by the installation teams.
Demonstration in the use of the IBS was also provided to each household during the time of installation, and the IBS intervention was promoted regularly throughout the follow-up period by the local energy experts’ team of IBS monitors. Concerning study duration, since the life span of “Mirt” IBS is about 5 years (21, 22), the length of both the intervention and the follow-up period was one year safely to account for seasonal factors that have a major effect on the magnitude of HAP in Ethiopia (36), and to maintain a sufficiently short follow-up period to decrease attrition.
2.7.2. Trial adherence and compliance monitoring
Participants’ adherence to the intervention assigned to them were assessed through self-report and direct observation by trained field enumerators along with local energy experts’ team in both arms. At each follow-up visit, the enumerators observed and recorded the type and condition of baking stove currently being used (i.e., no observed breakage resulting in no use). Additionally, the primary cook was asked whether the baking stove intervention was in good working order (i.e., no reported breakage resulting in no use).
Accordingly, timely response to trial-related difficulties such as rapid maintenance or replacement of defective baking stove were accomplished by the installation teams as needed to improve intervention protocol adherence and avoid the potential detrimental effects of non-adherence. In addition, trial protocol compliance was checked by the local energy experts’ team of stove monitors’ through unannounced visual inspection visits in homes of both arms to enhance data validity.
2.7.3. Participant retention strategies
Once the households were enrolled, reasonable efforts were made to promote participant retention and complete follow-up for the entire study period by working on active community engagement through the Ethiopian health extension program as well as through the local health development army team structure to prevent missing data and avoid the associated complexities in analysis and interpretation. In this regard, interest in the study was maintained through periodic communications about the intervention protocol adherence during the regular local health development army team meetings as well as during home visits by field workers, health extension workers, and local energy experts.
Furthermore, home visits were scheduled to limit the participants’ burden related to follow-up visits, and at the start of the trial, control households were informed that they would receive the IBS intervention at the end of the study period to maintain justice and achieve a high level of post-recruitment participant retention.
2.7.4. Trial safety monitoring
Even though the “Mirt” IBS intervention (21, 22), which was tested by this trial, was not involved any drug or medical procedure; and not known to increase the risk of any adverse event, masked interim analyses were included in the protocol for safety and efficacy monitoring. Since the standard “Mirt” IBS intervention (21, 22) is expected to reduce HAP related health effects (22), it was likely to be safer than the open burning TBS method. Any adverse events data deemed related to the trial intervention (i.e., such as cooking-related burn events data) were collected, and reported immediately for appropriate medical action and for further assessment of seriousness and expectedness to inform the conduct of the ongoing trial.
The collected data were reviewed for safety in December 2018 by an independent data Safety, and Monitoring Board to determine whether there were grounds to stop the trial for adverse events. Nevertheless, the Board found no grounds to stop the trial early due to adverse events.
2.8. Outcome assessment
To identify the occurrence of childhood ALRI, we used the definition of Integrated Management of Childhood Illnesses (IMCI) pneumonia algorithm developed by the World Health Organization (WHO) (37, 38). Since nurses can effectively diagnose childhood pneumonia, and the IMCI-pneumonia assessment protocol is a common diagnostic criterion at health facilities in Ethiopia (39), the outcome variable was assessed in the same manner in both arms by field nurses, who were trained in IMCI-pneumonia algorithm, and not involved in either randomization or intervention delivery. In this trial, the term childhood ALRI was used as a synonym for the IMCI-pneumonia (37, 38), because it is the preferred term for childhood pneumonia in peculiar to the developing countries (40).
2.9. Data collection methods
Baseline data were collected in each household before intervention delivery, and a total of four follow-up visits were carried out, immediately after the delivery of the intervention, in the same week at approximately 3-month intervals for longitudinal data collection by trained local field nurses. The duration of the follow-up period was one year to cover the major Ethiopian periods of annual weather changes and to account for seasonal factors that might have a major effect on the magnitude of both ALRI and HAP in Ethiopia (41).
2.10. Data quality assurance
Various appropriate measures were taken to address the validity and reliability of the study by ensuring the quality of data. To start with the outcome variable, childhood ALRI was assessed in the same manner in both arms by field nurses who were trained in the standard IMCI-pneumonia algorithm. The study team was in regular contact with the data collection team with scheduled meetings and additional communications as needed for quality control. About 5% of randomly selected home visits were done in duplicate by supervisors as cross-checking mechanism to ensure the validity of the collected data, and feedback sessions were done on about a three-month basis.
To minimize the risk of bias, clusters were randomly allocated to intervention and control arms, all eligible households within the clusters were included in the study, the allocation sequence was concealed from those assigning participants to groups, primary outcome assessors were blind to the intervention and outcome measurement was performed in the same manner in both arms. In addition, the entire trial stoves were manufactured by a single licensed firm and the same installation teams were administered the intervention in both arms.
Furthermore, all initially randomized participants were analyzed in the groups they were assigned to (i.e., intention-to-treat/ ITT analysis principle). Furthermore, the methodological soundness such as large sample size, longitudinal study design, and baseline data collection on the primary outcome and risk factors to be adjusted through GEE modeling can help us to achieve an effective balance of confounders in both arms. Finally, this manuscript was reported following both the guidelines of Consolidated Standards of Reporting Trials (CONSORT) 2010 statement extension to cluster randomized trials (42) to address the essential study design components of this trial report; and Template for Intervention Description and Replication (TIDieR) checklist (43) for better reporting of the intervention aspects.
2.11. Statistical analysis methods
Data were analyzed using the Statistical Package for Social Sciences (SPSS), version 22 and all statistical tests were two-sided with p-value < 0.05 considered statistically significant. To quantify the magnitude of clustering for childhood ALRI outcome, cluster-level ICC value was calculated using the standard formulae as:
ICC =Var(Uo)/Var(Uo) + Π2/3 (44, 45).
Where, Var(Uo) is the random intercept variance, that is, the estimated cluster-level (i.e., level-two) variance component, and Π2/3 refers to the individual-level (i.e., level-one) variance component which is fixed to be 3.29 (45). We took this fixed value of 3.29 for the individual level variance component, as there is no individual-level (level-one) residual for binary (discrete) outcome variable in the multilevel logistic analysis model (44, 45).
Using this formula, the estimated magnitude of cluster-level ICC value for childhood ALRI outcome was found to be not different from zero, which indicates that only well below 1% of the total variability in the chance of acquiring childhood ALRI is explained by the between cluster-level variation. Therefore, we safely treated individual children as the sole unit of analysis and interpretation in determining the effect of IBS intervention on childhood ALRI (44) compared with the continuation of the open burning TBS.
The effect of IBS intervention on the repeated response of childhood ALRI between the two arms was estimated using adjusted Odds Ratios (AORs) , with the respective 95% CIs, as measures of effect following a Generalized Estimating Equations (GEE) modeling approach among the ITT population by considering the underlying design of the study and nature of the outcome variable under investigation. The GEE analysis method is the preferred method for longitudinal data analysis due to its computational simplicity and robustness to misspecification of the repeated measures correlation structure (46). Besides, this analysis technique can take both the village-level clustering of childhood ALRI outcome and the longitudinal sampling method into account, and it can handle missing data values without the need for explicit imputation by considering all participants with at least one follow-up visit through automatically excluding the missing values by the GEE analysis model (46).
Concerning model fitness, even though the GEE method is understood to be robust against a wrong choice of working correlation structure (WCS), the best WCS of the outcome variable was chosen as autoregressive (AR1) by means of a critical examination of the working correlation matrix of the observed correlations between subsequent measurements (46) to uphold the goodness of model fitness, and hence to get a more precise estimation of the IBS intervention effect.
As a final point, our GEE analysis model has included a binary outcome variable of repeatedly measured childhood ALRI and a binary indicator of treatment allocation (i.e., control versus intervention) as well as other indicator variables measured at baseline such as gender, age, childhood ALRI, location of cooking quarter, secondary cookstove type used for other cooking/boiling water purposes and frequency of “Injera” baking event; and the results are presented next in texts and tables.