The study is implemented in Chiradzulu District, Malawi (Figure 1). Chiradzulu District is situated in the southern region of Malawi and is subdivided into 10 administrative regions, or traditional authorities (TA). The Malawi 2015-2016 Demographic and Health Survey (DHS) indicated that 52% of the population has access to improved sanitation. According to the National Statistical Office, in 2019, 93% and 10% of the households in Chiradzulu District had access to safe water and improved sanitation facilities, respectively. OD rates in Malawi and Chiradzulu District are 6% and 7%, respectively [25]. Given this low level of coverage, Chiradzulu District is the target of a 3-year (2021-2024) district-wide water, sanitation, and hygiene program, known as WASH for Everyone (W4E), implemented by World Vision and Water for People, alongside which this trial is embedded.

The study design is a controlled before-and-after (CBA) trial with 2 treatment arms (each with 20 villages or clusters) and a control group (30 villages; Figure 2; Multimedia Appendix 1). CBA intervention designs are a nonrandomized approach used to evaluate the impact of interventions [26]. The advantages and disadvantages of the CBA study designs, also known as nonrandomized cluster-controlled trials, have been discussed elsewhere [27]. In this study, TAs are the unit of intervention assignment and villages (clusters) are the unit of analysis. [28] In our study, we selected 3 TAs (Figure 1) in Chiradzulu District, Malawi. TAs are 4th-level administrative units, with an average population of 34,000 in Chiradzulu [29]. Details on the program intervention are described below. For our trial, we selected 2 out of the 5 TAs scheduled to receive the full W4E intervention during the second year of program implementation to align with our implementation. We selected the 2 TAs with the closest match in population and estimated sanitation coverage from previous surveys. In discussion with program partners, one TA was assigned to receive the CLTS intervention, and one TA was assigned to receive the CLTS+Care Group intervention. We selected a third TA from the 3 TAs scheduled to receive the WASH for Everyone intervention in the third year of program activities as a control group for our study (Figure 2). The control TA was selected based on generally similar population and sanitation coverage measures and lack of contiguous borders with the 2 intervention TAs. TAs are the unit of intervention assignment for our study. Within each TA, villages serve as the specific study clusters which are the primary sampling units for data collection.

Sample size calculations were based on the minimum detectable effect (MDE) to assess changes in sanitation coverage, the primary outcome measure, at the project end line. Data for all parameters used in sample size calculations were drawn from the baseline survey conducted by World Vision before the implementation of the W4E project between January and April 2021. For the MDE calculation, we assumed a 2-tailed test with a significance level of 0.05 and a power of 0.8. We assumed a conservative intracluster coefficient of 0.1 [30]. As baseline data indicates that the average percentage of households with at least unimproved sanitation coverage across Chiradzulu District was 67%, power calculations assumed a total enrollment of 20 intervention clusters in each TA receiving interventions and 30 clusters selected from the control arm of the study. Assuming an average of 20 households surveyed per selected cluster results in an MDE of 0.136 or an average difference of 13.6 percentage points in sanitation coverage (relative change in coverage of 19%) between either arm of intervention clusters and the control cluster. Therefore, the study aims to recruit a total of 1400 households (CLTS+CG: n=400; CLTS only: n=400; Control: n=600; Figure 2) at each data collection round.

Households in TA Ntchema (CLTS only) will receive standard CLTS intervention plus government-delivered hygiene promotion. We refer to this arm as the CLTS group. Households in TA Chitera (CLTS+CG group) will receive the same combination of interventions as the CLTS group in addition to community-based CGs. We refer to this as the CLTS+CG group.

CLTS is the main sanitation intervention component for W4E and is implemented in line with the Malawi National Sanitation and Hygiene Strategy [29]. As part of W4E, relevant district technical officers and community leaders (also known as Natural Leaders) are responsible for implementing CLTS. Trained CLTS facilitators will conduct the triggering sessions that include participatory activities, such as the walk of shame, shit calculation, and community sanitation mapping. The purpose of these activities is to trigger behavioral emotions, such as shame and disgust so that community members understand the consequences of open defecation. If successfully implemented, triggering sessions have the potential to stimulate community members to stop open defecation and adopt improved sanitation practices, including the construction and effective use of latrine facilities. In addition to sanitation promotion, the project partners delivered hygiene promotion campaigns using pre-established sanitation and hygiene messages from the Ministry of Health in marketplaces and villages. This was achieved through 4 separate campaigns, each lasting 5 days, using a mobile van for message dissemination and distributing information leaflets.

In TA Chitera, in addition to the activities outlined above, households will receive visits from local CGs. CG leaders and cluster leaders support intervention delivery and local triggering sessions, specifically by facilitating CG meetings with CG households and conducting household follow-up visits. CGs are intended to extend the reach of CLTS behavior change messaging, providing additional points of contact with program households. CGs also participate in posttriggering follow-up visits to households to assess sanitation coverage and use.

In line with CLTS values, the W4E project does not intend to provide any latrine construction or hygiene facility materials or financial subsidies to households. It is the responsibility of the household owners to support themselves throughout the latrine and hygiene facility construction process.

The primary sampling units (clusters) for the study are clusters (also called villages) and represent the units of W4E delivery for CLTS. Communities from the 3 participating TAs were randomly selected from a list of communities obtained from the Chiradzulu District Health Office. Inclusion criteria for the selected communities include that the community is in one of the 3 selected TAs that are part of the study area and have not yet received any exposure to the W4E project. Communities that were exposed to CLTS-related activities in the past 12 months and are outside the area of the 3 selected TAs are excluded from the study.

In early 2023, a list of all clusters in each study TA and their associated number of households was provided by the District Health Office. TA-specific median village size was calculated, and villages were categorized as either above or below the TA-specific median. Villages in each TA-specific above and below median list were ranked ordered according to a random number generated in Microsoft Excel and villages were enrolled sequentially until the necessary number of villages were enrolled. If a cluster could not be located or if the village chief did not provide permission for data collection, the next cluster on the rank order list was enrolled.

Households are the secondary sampling unit and individuals living in households in the study area are the primary study population. Twenty communities will be selected from TA Chitera (CLTS+CG), 20 from TA Ntchema (CLTS), and 30 from TA Nkalo (control). Among selected communities, a systematic sampling procedure (selecting every kth household) was used to select 20 households from each cluster included in the study. For each cluster, the population of households obtained from the District Health Office was divided by 20 (cluster sample size) to find the sampling interval (kth number). Then, the kth household was selected and included in the survey from any starting household chosen by an enumerator within the cluster.

A baseline survey was conducted in all 3 TAs between May and June 2023 before intervention implementation in TA Chitera (CLTS+CG group) and TA Ntchema (CLTS group). An end-line survey was conducted in all 3 TAs between March and April 2024. TA Nkalo (control group) received the CLTS intervention after the end-line survey. A structured questionnaire with closed-ended questions with precoded responses on mobile devices on the KOBO collect platform (KoboToolbox) was used to collect data on household membership, wealth index, sanitation and hygiene facilities, and child health (Multimedia Appendix 2). Further, the enumerators conducted spot checks and recorded hygiene proxy measures, such as the presence and state of a latrine, presence, location, and type of handwashing facility (including the availability of soap and water). The interviews were conducted in Chichewa, the local language of Chiradzulu District. Chichewa-speaking enumerators with extensive training and expertise administered the structured questionnaire.

To confirm the availability of WASH infrastructure and hygiene practices, we conducted structured observations in 350 selected households (ie, 100 households from each intervention arm and 150 households from the control group) among the recruited 1400 households. A line listing of the 1400 households was created and a Microsoft Excel random generation number was run to identify the 350 households for observations. Specifically, the observations were intended to capture practices pertaining to the presence of latrine and handwashing facility with soap, handwashing practice at critical times (ie, before eating, before food preparation, after changing the child's nappy, and after latrine use), and child feces disposal. One observer was placed at each selected household. As behaviors of interest mostly occur in the morning hours to noon [31], observations lasted 4 hours from 8 AM to noon. The observer first conducted the structured observations after which he or she proceeded to administer the structured questionnaire.

Data collected using mobile devices was uploaded directly on the web to KOBO daily. Only the PI, coinvestigators, and study personnel with authorized access have access to web-based data. At the end of the data collection, the full dataset was uploaded into STATA (version 18; StataCorp) for analysis. Although no identifiable information was recorded during the surveys, study personnel will review the final database and permanently delete any identifiable information inadvertently collected during surveys. All data is stored on encrypted, password-protected servers.

The primary outcome is sanitation coverage. Secondary outcomes include basic sanitation coverage, sanitation use, sanitation-related quality of life, latrine quality, presence of a handwashing facility, handwashing behavior, and safe disposal of child feces (Multimedia Appendix 3).

Statistical analysis will be carried out at the individual or household levels with appropriate adjustments for clustering within villages (and within households for individual-level outcomes). Data will be analyzed according to the TA and intervention assignment irrespective of whether the intervention was taken up fully, partly, or not at all. Since this is an effective study, only an intention-to-treat dataset will be maintained.

The primary estimates of the effectiveness of the intervention on primary and secondary outcomes will be based on a difference-in-difference analysis and models adjusted for design variables alone (TA, village population above or below TA-specific median); a fixed effect for village size (above or below the median) will be included in the model. All models will include a dummy variable for the treatment arm (ie, CLTS, CLTS+CG, or control) and data collection round (baseline and end line) as well as the interaction term of those 2 variables.

To account for the clustering of observations, hierarchical mixed-effects models will be used for all analyses. For household-level outcomes or outcomes where there is only one respondent per household, mixed effects models will account for clustering by including a random effect at the village level. For outcomes with multiple respondents or data points per household (specifically sanitation use, and hand washing behavior), an additional random effect will be added at the household level.

All analysis of primary and secondary outcomes will be based on the difference-in-difference approach. The regression coefficient of the interaction term between the treatment arm and the data collection round will be used as the primary effect measure. For binary outcomes measure, we will use melogit with robust SEs and regression coefficients exponentiated to estimate an odds ratio (OR). For continuous outcomes, we will use meglm, with identity link and Gaussian family.

The results will be presented with 95% CIs. Analyses will report on ORs and 95% CIs of each treatment arm compared to the control as well as the CLTS alone compared to CLTS+CG. Secondary outcomes include the following:

Results will present 2 sets of outcome measures. First, we will report on all outcome measures adjusted for design variables (village above or below TA-specific median) and models also adjusted for a priori-defined covariates and design variables. We will explore differences between the design-adjusted and covariate-adjusted models but will consider the covariate-adjusted models as the primary effect estimates.

Covariates have been selected based on hypothesized relationships that could confound the relationship between intervention exposure and primary and secondary outcome measures.

At the respondent level, gender and primary education (less than vs completed primary) will be included. The following covariates will be included at household-level: household size, household economic status based on PCA of household assets, any member of the household experiencing a disability as defined by the Washington Group (any functional disability for a member of the household greater than 2 on the functional disability assessment) and household reports a water source that is located on site or on plot [33]. At the community level, we will look at the village having an improved road as reported by the Village Chief at the time of baseline enrollment. The improved road is used as a proxy measure of village accessibility to both the intervention and markets.

It is likely that some missing outcome data will be encountered, especially for individual-level self-reported outcomes. The patterns of missingness of variables will be tabulated to describe and compare the extent of missingness of any affected variable between study arms. No adjustment will be made for missing data.

Unusual values and potential outliers will be flagged and queried. Unlikely values will be dropped and treated as missing data in the main analysis. A sensitivity analysis will be conducted which includes potential outliers (but not unlikely values).

The number of primary outcomes that will be tested for significant differences between arms is small; thus, no formal adjustment for multiple comparisons will be made.

The study protocol, which includes data collection tools, participant information sheets, and consent forms, has been approved by the National Commission for Science and Technology (P01/23/718) in Malawi and the London School of Hygiene and Tropical Medicine (28249). Further, consent was obtained from the Chiradzulu District Council and community leaders. Informed written consent was obtained from all study participants recruited into the study before data collection. The trial was registered in ClinicalTrials.gov (NCT05808218).

Given that CLTS alone is reportedly not effective at ensuring sustained uptake of sanitation and hygiene behaviors, we test the hypothesis that the CLTS+CG intervention will be more effective at improving sanitation coverage in a rural area of Malawi than the CLTS intervention alone, and also relative to the no intervention arm. Studies have shown the effectiveness of using the CG model in promoting community health interventions, mainly nutrition programs [19,34-37]. However, there is limited evidence on the impact of interventions integrating CLTS and the CG model to address this gap, we tested an intervention that integrates the CG model into the standard CLTS approach. The CG model is a community-level government-recognized approach that can potentially support the already existing informal natural leaders in the delivery of CLTS, both pre- and post-ODF status attainment. This could strengthen the sustainability of CLTS as CG members are trained to continuously monitor groups of households post-ODF attainment, whereas Natural Leaders only follow up with households until the community has achieved ODF status [9,11,38,39]. Results from this trial will provide evidence on the extent to which the CLTS+CG approach is effective at improving sanitation and hygiene practices in the W4E program area compared to CLTS alone and no intervention, as well as inform implementing partners on future interventions in Chiradzulu District, Malawi.

Our baseline results indicate that the study is conducted in a rural area where the literacy level is moderate and the quality of housing relatively poor. Some households had no latrines, while traditional unimproved latrines (eg, pit latrines without slab), which are prone to collapse during the rainy season, are common. Relatively high sanitation coverage in the communities indicates the short-term success of CLTS, however, other TAs in Chiradzulu District have reverted to OD after achieving ODF status. Thus, the district provided a suitable environment for this study to assess alternative approaches for promoting sustainable sanitation interventions.

Handwashing with soap is low in low- and middle-income countries, including Malawi [40-44]. Similar findings were observed in our baseline results. Access to easy-to-use and effective handwashing facilities was limited among the study population as most households did not have a dedicated handwashing facility. Observations at critical times indicated a challenge with soap availability for hand washing. As such, nearly half of the study participants did not use soap during handwashing.

The baseline results indicate some imbalance between trial arms on self-reported water treatment practices. Treatment of household water was more frequently reported among the control group compared to the treatment groups. A possible explanation for this imbalance could be that immediately before baseline data collection, other NGOs promoted the use of safe water through the drilling of boreholes and installation of chlorine dispensers in all the water points in the control communities. This included the installation of chlorine dispensers in both treatment groups, although a similar trend was not observed in responses. In terms of sanitation, we observed that the control arm had the highest coverage of latrines without a slab, while more latrines with a slab were observed in the CLTS group. We could not find an explanation for this difference and expect it to be by chance.

This study has several limitations. Structured observations were used to collect the data. This approach has limitations as the presence of the observer can influence the behavior of the person under observation [45]. To minimize bias, study participants were not informed about the observed hygiene behaviors, though they were generally informed about the purpose of the study. Nonetheless, structured observations remain the gold standard for measuring hygiene behaviors [46,47]. Treatment arms and the control were purposively selected based on the presence of care groups and the CLTS program for the treatment arms rather than random assignment. The control arm was selected based on the absence of care groups and CLTS. Systematic random sampling was used to select study households in each arm.

We highlight the study design and baseline results from an ongoing BCA trial implemented in rural households in Chiradzulu District, Malawi. The study, implemented within a district-wide WASH program, assesses the effectiveness of CLTS alone against CLTS combined with the CG model for improving sanitation coverage and sustained uptake. The baseline observations indicate a balanced distribution of potential demographic confounders in the trial arms with a slight variation on some WASH proxy measures. We expect to publish the trial findings in peer-reviewed journals in 2025 and share findings at local and national events in Malawi, as well as at international conferences.