Study protocol for a multi-arm, parallel group randomised controlled trial evaluating the isolated and combined effects of behavioural promotion and hardware provision on handwashing with soap in peri-urban Lusaka, Zambia

Abstract

Introduction

Diarrhoea and lower respiratory infections remain leading causes of under-five mortality in sub-Saharan Africa. Handwashing with soap (HWWS) is a cost-effective intervention, yet adherence remains low. Effective interventions must address multiple behavioural influences, as knowledge alone is insufficient without supportive environments. Multi-component handwashing interventions target diverse behavioural determinants, yet few studies disentangle the isolated and combined effects of different components. This multi-arm randomised-controlled trial in Lusaka, Zambia, aims to assess the isolated and combined effects of hand hygiene behaviour promotion and handwashing hardware and supply provision on HWWS at handwashing opportunities, strengthening the evidence base for designing sustainable handwashing interventions.

Methods and analysis

This superiority, multi-arm, parallel group randomised-controlled trial will be conducted in five peri-urban communities in Lusaka, Zambia, with 1800 households randomly assigned to four arms (450 households per arm; 1:1:1:1): (A) handwashing hardware and supply provision only, (B) hand hygiene behavioural promotion only, (AB) hand hygiene behavioural promotion+hardware and supply provision, and (C) control (no intervention). Households receiving the hardware and supply intervention (A, AB) will receive a handwashing facility and supplies to make liquid soap solution with a maintenance check after 3 months. Households receiving the hand hygiene behaviour promotion intervention (B, AB) will receive five bi-weekly visits with a sixth follow-up visit 5 weeks later. HWWS behaviour of one household member will be assessed through structured observations at baseline and endline (6 months after intervention delivery begins).

Ethics and dissemination

Ethical approval was obtained from London School of Hygiene & Tropical Medicine, Research Ethics Committee, London, UK (Ref: 31387), the University of Zambia Biomedical Research Ethics Committee (Ref: UNZABREC 5834–2024) and the Zambia National Health Research Authority (NHRA −1823/23/12/2024). Results of the main trial will be submitted for publication in a peer-reviewed journal.

Trial registration number

NCT06865495.

Strengths and limitations of this study.

• The large sample and comprehensive outcome measures will provide robust evidence on the isolated and combined effects of intervention components on behaviour and health.

• The interventions are culturally sensitive, evidence-informed, and family-centred, designed to avoid placing responsibility for handwashing solely on caregivers (usually women in these contexts).

• Structured observation, while subject to the Hawthorne effect, remains the gold standard for measuring routine behaviour in domestic contexts.

• Limited blinding of participants is an acknowledged limitation, inherent to behavioural trials of this nature.

• Potential for contamination exists between households, though this risk is minimised by the large sampling interval and will be assessed at endline.

Introduction

The health benefits of handwashing with soap (HWWS) are well established. Interventions promoting HWWS have been shown to reduce the risk of acute respiratory infection by 17% and diarrhoea by 30% in low-income and middle-income settings.^{1 2} However, HWWS practice remains low, particularly in low-resource settings.^{3 4}

While there is robust evidence linking changes in key hygiene behaviours to health outcomes, the mechanics of supporting and promoting lasting changes in individual behaviour remain uncertain. Households with a dedicated handwashing facility are approximately twice as likely to practice HWWS after faecal contact compared with those without.^{4 5} However, these estimates are largely based on cross-sectional data, limiting insights into causal mechanisms. Global monitoring suggests that 73% of households in sub-Saharan Africa lack access to a handwashing facility with soap and water at home.⁶ Traditional handwashing promotion interventions often focus on cognitive determinants of HWWS, with limited attention to a broad range of behavioural and infrastructural factors. Several systematic reviews of handwashing determinants in community settings found that knowledge of disease transmission alone is insufficient to drive HWWS, particularly in the face of unconducive behavioural settings or competing priorities.^7–9 Reviews highlight the critical role of environmental enablers—such as the presence of a dedicated handwashing facility—in overcoming psychological trade-offs that hinder HWWS. Successful handwashing interventions therefore must integrate multiple elements to address these diverse influences.^{10 11}

A recent scoping review of hand hygiene guidelines highlighted a lack of clear, evidence-based guidance on how to develop, implement and sustain hand hygiene behaviour interventions.¹² Multi-component interventions targeting cognitive, behavioural and infrastructural determinants of handwashing are needed. However, typical evaluation approaches comparing a multi-component intervention as a single package against a control group limit the ability to isolate the effects of individual components and understand their specific contributions to HWWS behaviour change.^{13 14} A recent systematic review reported substantial heterogeneity in the behaviour change techniques used across hand hygiene interventions and noted that the complexity of intervention packages often hinders the ability to disentangle the isolated and combined effects of specific components.¹⁵ The authors emphasised the need for research that is intentionally designed to evaluate these effects.

This multi-arm randomised-controlled trial aims to address this evidence gap by evaluating the isolated and combined effects of handwashing hardware and supply provision and hand hygiene behavioural promotion on HWWS at handwashing opportunities (ie, when hands could have been washed or not) in peri-urban communities in Lusaka, Zambia. By examining the effect and interactions of these intervention components, the study aims to enhance the evidence base for designing more effective and sustainable handwashing programmes.

Objectives

1. To evaluate the isolated and combined effects of intervention components on HWWS behaviour at handwashing opportunities (box 1).

2. To evaluate the isolated and combined effects of intervention components on HWWS at specific handwashing opportunities (eg, before eating; box 1).

3. To evaluate the isolated and combined effects of intervention components on hand-rinsing at handwashing opportunities (box 1).

4. To assess if water insecurity is an effect modifier on the relationship between the intervention components and hand hygiene behaviour.

5. To explore the effect of the intervention components on handwashing knowledge and self-reported handwashing.

6. To explore the durability of handwashing facilities over a 6-month period.

Box 1. Handwashing opportunities.

Before cooking.

Before eating.

Before feeding a child (spoon feeding, manual feeding or serving).

Before breastfeeding.

After toilet use.

After handling faeces (child, adult or animal) or cleaning a child’s bottom.

After contact with animals.

After coughing or sneezing.

Methods and analysis

Study design

The study is a superiority, multi-arm, parallel group randomised-controlled trial assessing both the isolated and combined effects of two interventions—hand hygiene behavioural promotion and handwashing hardware and supply provision—on observed HWWS. Given the potential for interaction between intervention components, a multi-arm trial using an ‘inside-the-table’ analysis will be used, rather than a factorial trial (using an ‘at-the-margins’ analysis).^{16 17} The research question is summarised in table 1.

Table 1. Research question (PICOT^* format)⁴¹.

Research question	What are the isolated and combined effects of handwashing hardware and supply provision and behavioural promotion on HWWS behaviour at handwashing opportunities?
Population	Households in peri-urban communities of Lusaka, Zambia, with at least one child under the age of 5 years, no fixed handwashing facility, and at least one adult (18+ years) who can consent to the study; participants must be permanent residents of the selected communities (ie, expect to remain in the community for the next 3–6 months)
Intervention	Households randomised to one of the four arms (1:1:1:1): (A) handwashing hardware and supply provision only, (B) hand hygiene behavioural promotion only, (AB) hand hygiene behavioural promotion+handwashing hardware and supply provision, (C) no intervention; at the end of the study period, the most effective combination of interventions will be delivered in the control group
Comparison	Our analysis will concern five pairwise comparisons: Behavioural promotion against control (B vs C) Hardware and supply provision against control (A vs C) Behavioural promotion+hardware and supply provision against control (AB vs C) Behavioural promotion+hardware and supply provision against hygiene promotion alone (AB vs B) Behavioural promotion+hardware and supply provision against hardware and supply provision alone (AB vs A)
Outcome	The primary outcome is the probability of HWWS within 3 min of any handwashing opportunity (box 1) in the correct sequence (eg, before or after the relevant opportunity)
Time	Outcomes measured at baseline and endline (6 months after intervention delivery begins)

^*Population, Intervention, Comparison, Outcome, Time

HWWS, handwashing with soap.

A one-stage approach will be used to randomise 1800 households to one of four trial arms (450 households per arm; 1:1:1:1): (A) handwashing hardware and supply provision only, (B) hand hygiene behavioural promotion only, (AB) hand hygiene behavioural promotion+handwashing hardware and supply provision, and (C) control (no intervention; figure 1). A block randomisation approach will be used during allocation of households to study arms, ensuring proportionate representation of participants across study areas. This protocol was drafted using the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) reporting guidelines.¹⁸ The SPIRIT checklist can be found in online supplemental material table 2.

Figure 1. (A) Description of the intervention components within each arm and (B) table of the 2×2 multi-arm study design.

Study setting

This community-based study is based in Lusaka, the capital and largest city in Zambia. Nearly 62% of Lusaka’s population live in peri-urban areas,¹⁹ characterised by dense, informal settlements, with unplanned housing and persistent challenges in accessing WAter, Sanitation and Hygiene (WASH) services. These communities experience seasonal flooding, leading to cyclical cholera outbreaks and other public health risks. Several WASH-related indicators—including limited soap use for handwashing, reliance on unimproved sanitation facilities, and uncovered household water storage—were identified as risk factors for waterborne and diarrhoeal illness in these communities.²⁰ An estimated 71% of the urban population in Zambia lack access to a household handwashing facility with soap and water.²¹

Lusaka city is in Lusaka district, which is divided into 33 wards (small administrative units). Wards where we recently conducted feasibility research for this study were excluded (n=2). We then selected wards that fit the following criteria: (1) low-income, peri-urban communities; (2) high population density; (3) limited access to WASH services; and (4) high burden of hygiene-related disease. From these wards (n=15), five were selected in collaboration with the study team, drawing on prior engagement with the District Health Office’s Environmental Health Technician representing the five sites. Although ward selection was not random, the findings will likely be applicable to other low-income, peri-urban communities with similar characteristics in the region. The five selected wards are Chainda, Mpulungu, Chawama, Mtendere and Kanyama (figure 2).

Figure 2. Map of Zambia with Lusaka District highlighted in red (left) and map of Lusaka District with the five selected study wards highlighted (right). Administrative boundaries of Zambia were sourced from the Humanitarian Data Exchange, Zambia – Subnational Edge-matched Administrative Boundaries, licenced under CC BY-IGO (available from: https://data.humdata.org/dataset/cod-em-zmb; left graph) and Stanford Digital Repository – Zambian 2006 to 2010 Constituency and Ward Boundaries, not subject to copyright (available from: http://purl.stanford.edu/yc436vm9005; right graph).⁴².

Eligibility criteria

Households eligible for the trial must meet all the following criteria at enrolment:

• At least one adult (aged 18 years or older) who can consent to the study on behalf of all members of the household.

• At least one child under the age of 5 years.

Exclusion criteria:

• Non-permanent resident of the selected community and/or plans to leave the community within the next 3–6 months.

• Household already own a fixed handwashing facility (eg, sink, handwashing station).

The head of household or a nominated representative will provide written, informed consent for household participation in the study. Data collection activities will be completed with the adult (over 18 years or older) most responsible for caring for the child under the age of 5 years at the time of data collection. If a child under 5 years is not present, the adult (18 years or older) most responsible for childcare will be selected for observation. If such an individual is not present, another randomly selected household member (18+ years) will be observed instead. Separate consent will be obtained from the household member participating in data collection activities.

Interventions

This household-level, multi-component intervention includes the delivery of hand hygiene behavioural promotion and handwashing hardware and supplies to households in pre-selected peri-urban communities in Lusaka, Zambia. The intervention is outlined according to the Template for Intervention Description and Replication for WASH (TIDieR-WASH) guidelines,²² for which the checklist can be found in online supplemental material table 3.

A Theory of Change (ToC) was developed to illustrate how the intervention components are expected to increase HWWS, and how this may be influenced by other factors (figure 3). We hypothesise the behavioural messaging component (Arms AB+B) will improve the capability, opportunity and motivation of household members to practice HWWS. For households receiving hardware and supplies (Arms AB+A), we hypothesise the presence of a handwashing facility with soap and water in the household will increase the opportunity for HWWS at key opportunities. Assumptions that need to hold true for change to occur are outlined in the ToC.

Figure 3. A Theory of Change for the behavioural messaging (Arms AB+B) and hardware and supplies provision (Arms AB+A) components of the intervention. Elements specific to each intervention component are grouped under sub-headings to highlight their unique pathways. Assumptions that need to hold true for change to occur are detailed at the bottom of the figure. HWF, handwashing facility; HWWS, handwashing with soap.

This multi-component intervention will be delivered by a team of community volunteers known as neighbourhood health committee (NHC) members, all of whom are residents of the selected communities. These volunteers typically oversee health promotion efforts, vaccination campaigns and disease surveillance within the communities they serve. They will be recruited on a short-term basis by the local research partner (Center for Infectious Disease Research in Zambia, CIDRZ) and overseen by the research team there. Intervention delivery teams will be given custom t-shirts with the intervention logo to create a distinct visual identity for the team and avoid direct associations with staff employed for data collection. The interventions will be delivered to all members of the household present at the time of delivery.

Handwashing hardware provision (Arms AB+A)

Households will be provided with a locally available handwashing facility called the Kalingalinga bucket (figure 4A), which was the favoured handwashing facility out of several designs in a previous feasibility study.²³ Facilities are manufactured locally, typically by metal workers who operate from a community called Kalingalinga, after which the bucket and stand are named. Handwashing facilities will be tested and quality checked prior to distribution. As cost and affordability of handwashing materials are commonly reported barriers to HWWS,⁷ and as this study will be conducted in low-income communities, handwashing facilities will be provided to households free of charge. The cost of the facility is a maximum of US$10. Of note, US$17 per household is the mean costs included in economic models of the cost of achieving universal hand hygiene in domestic settings,²⁴ thus representing a viable intervention option for scale.

Figure 4. (A) Kalingalinga bucket and (B) plastic detergent container for soapy water. Both photographs are original and were taken by the author.

Households will also be given instructions on how to make liquid soap (soapy water) from locally available soap products. Households will be provided with a sachet of powdered soap and a plastic detergent container which can be attached to the handwashing facility, ensuring soap can be co-located with water (figure 4B). Soapy water was selected as an alternative to bar soap because it can be co-located with the handwashing facility and easily made in households with existing resources. A small pilot conducted prior to the trial confirmed the acceptability of soapy water. Participants reported that soapy water made from powdered (ie, laundry) soap was easier to make and use compared with soapy water made with bar or liquid soap. Intervention delivery workers (NHC members) will set-up the handwashing facility and demonstrate how it works, including where to place soap near the stand. Discussions will also be held on maintenance and positioning of the handwashing facility when it is provided. Participants will be advised to report if their handwashing facility is damaged or stolen between follow-up visits. Any problems will be immediately addressed, or the handwashing station replaced. Intervention delivery workers will return to households after 3 months to ensure there are no issues with the handwashing station (eg, broken stand or tap, or stolen), to bring more powdered soap and to remind households how to make soapy water. Both visits should take no longer than 30 min.

Hand hygiene behavioural promotion (Arms AB+B)

Household-based hand hygiene behavioural promotion will be based on an innovative intervention informed by multiple behavioural theories and models.^25–27 Households will receive a behavioural intervention called Ulupwa (the family intervention), which uses interactive storytelling and visual aids to promote HWWS. Designed as a participatory experience, Ulupwa actively involves families through storytelling, games and discussion. Intervention delivery workers will act as storytellers, narrating stories about three characters with different handwashing habits. Interactive elements are embedded throughout to encourage reflection on personal handwashing behaviours, explore common barriers to sustained practice, and support families to develop their own practical solutions. A key prop for the intervention is a set of visual ‘playing’ cards that illustrate the characters, handwashing barriers and other related concepts. The intervention supports the household to realise all family members should take responsibility for maintaining safe handwashing practices and is designed to involve the whole family, rather than placing the responsibility solely on caregivers (usually women in these contexts). The intervention progresses through five 30 min visits conducted on a bi-weekly basis, with a final sixth visit 5-weeks after the fifth visit (3-months after the first visit; table 2). This schedule was designed to mirror the structure typical of handwashing promotion programmes. More details are outlined in online supplemental material text 1.

Table 2. Intervention visit content.

Visit topic	Content
Visit 1: identifying behaviours	This visit aims to build a connection and trust with the family and create an environment where resources can lead to transformative conversations; during the visit, storytellers will understand handwashing behaviours of each family member and discuss their perceived handwashing barriers
Visit 2: understanding disease risks	This visit tells a story about Nevers, a boy who does not understand the importance of handwashing behaviour; the visit aims to build awareness around the disease risks associated with inadequate HWWS and provide information on when to wash hands
Visit 3: encouraging practical solutions	This visit tells the story of Sami, Nevers’ sister, who knows handwashing is important but often forgets or does not have time to wash hands; this visit aims to introduce nudges or other environmental cues to prompt HWWS; during the visit, households not receiving hardware will receive guidance on how to set-up and position a handwashing facility in their home and will be shown how to make soapy water; all households will be given handwashing reminder cards to position around the household
Visit 4: promoting practice	This visit tells the story of Always, Sami and Nevers’ cousin (male), who always washes hands with soap and water and wants to ensure all family members take responsibility for maintaining safe handwashing practices; this visit aims to motivate all household members to take responsibility for handwashing, focusing on protecting household members and the community from disease; wider benefits such as improved school attendance will also be discussed
Visit 5: reflection and commitment	This visit reinforces key messages across the intervention period and aims to encourage participants to re-assess their behaviours and pledge to adopt improved hygiene practices
Visit 6: final visit	This visit asks participants to reflect on the intervention as a whole and covers any information participants would like to revisit; this includes making more soapy water if relevant

HWWS, handwashing with soap.

Households in the comparison group will receive no intervention during the study period. However, at the end of the study period, the most effective combination of interventions will be delivered to these households.

Criteria for discontinuation

Households that miss more than three consecutive visits will not receive further intervention visits. However, the trial will follow the intention-to-treat principle, so no households will be excluded from the analysis regardless of discontinuation.

Monitoring adherence

Intervention delivery workers will complete electronic data capture forms at each household visit. Data captured includes, but is not limited to, the number of household members present (all intervention arms), content discussed (all intervention arms) and condition/use of the Kalingalinga buckets (Arms AB+A).

Concomitant interventions

There are no prohibited concomitant interventions during the trial. However, households will be asked if they have received any other health-related interventions before (at baseline) and during the study period (at endline).

Patient and public involvement statement

Households from similar peri-urban communities in Lusaka participated in the feasibility and pilot testing of intervention components. Eight households received the first five hygiene promotion visits, and were invited to evaluate content, comprehensibility and willingness to engage in the full programme. The programme was revised based on participant feedback. Households identified the Kalingalinga bucket as the favoured handwashing facility out of several designs in a previous ranking exercise²³ and Trial of Improved Practices. A small pilot on the acceptability and use of soapy water with 10 households was also conducted. Sixteen households participated in pilot baseline outcome assessments, including the 3-hour structured observation.

Before study commencement, the research team will share the study protocol and ethical board approvals with the provincial and district health offices and request permission to conduct the study in the pre-defined communities. On written approval, the research team will hold introductory meetings with the sister-in-charge (health facility manager) at community health facilities and sensitisation meetings with community representatives (NHC members) who are the gatekeepers responsible for health promotion. Sensitisation meetings will be conducted in the local language and include information on the study objective and aims, targeted study participants, and the procedures outlined in the protocol. Input will be requested on the best way to recruit participants. Following these meetings, attendees will disseminate information through established communication channels to ensure awareness of research activities.

Outcomes

Primary outcome

1. HWWS at all handwashing opportunities.

The primary outcome is the probability of HWWS within 3 min of any handwashing opportunity (box 1) in the correct sequence (eg, before or after the relevant opportunity).

Secondary outcomes

2. HWWS at each specific handwashing opportunity.

3. Hand-rinsing measured as a three-level outcome: no action, hand-rinsing or HWWS at all handwashing opportunities.

The probability of either HWWS, hand rinsing with water only or taking no action within 3 min of any handwashing opportunity (box 1) in the correct sequence (eg, before or after the relevant opportunity).

Outcomes 1–3 will be measured through direct observation of individual handwashing behaviour over a 3-hour observation period at baseline and endline (6 months after intervention delivery begins).

4. Handwashing facility presence.

The probability that a household has a fully functional handwashing facility with water and soap present. This outcome will be measured at baseline and endline (6 months after intervention delivery begins).

5. Handwashing knowledge and reported practice.

Knowledge: the number of opportunities for HWWS (box 1) correctly specified in response to the question ‘When do you think it is important to wash hands with soap?’ Reported practice: the number of opportunities for HWWS (box 1) reported in response to the question ‘When did you wash your hands with soap yesterday?’

Both questions will be assessed through a survey of open-ended questions with pre-coded responses and multiple rounds of probing. This outcome will be measured at baseline and endline (6 months after intervention delivery begins) with one household member.

Tertiary outcomes

6. Diarrhoeal illness.

The risk of self-reported diarrhoea at least once in the past 7 days (experiencing three or more loose stools in 24 hours).

7. Respiratory illness.

The risk of self-reported respiratory illness at least once in the past 7 days (experiencing coughing or difficulty breathing).

Outcomes 6 and 7 will be measured through a survey with primary respondents at baseline and endline (6 months after intervention delivery begins) and will be collected for all household members. Results will be stratified by age group.

Participant timeline

Data will be collected from each household over a period of 6 months. Households will be enrolled, randomised and have a baseline visit at week 0. Due to the length of visit required for baseline measurements, and to ensure observations start on time, households will be enrolled and consented a few days ahead of baseline data collection. Intervention delivery will begin 2 weeks later. The research coordinator will provide the intervention allocation list to intervention delivery teams on Friday the week before intervention delivery begins. Households that do not receive a baseline visit before this list is provided will still receive the intervention. Endline data will be collected 6 months after intervention delivery begins. The 6-month follow-up period was chosen to capture behaviour beyond the intensive intervention phase. A shorter follow-up period would likely have captured only the short-term behaviour changes commonly observed immediately post-intervention. We acknowledge, however, that our follow-up window remains too short to reflect long-term, sustained behaviour change. Measurement time points are detailed in table 3 and figure 5.

Table 3. SPIRIT figure showing schedule of enrolment, intervention and assessments.

	Study period
	Enrolment	Intervention visits						Endline
		V1	V2	V3	V4	V5	V6/V2*
Time point (week)	0	2	4	6	8	10	15	27
Enrolment
Eligibility screen	X
Informed consent	X
Randomisation	X
Interventions
Hardware and supplies (AB+A)		X					X
Behavioural promotion (AB+B)		X	X	X	X	X	X
Assessments
Observation: handwashing behaviour (outcomes 1–3)	X							X
Survey: observed HWF presence (outcome 4)	X							X
Survey: handwashing knowledge+reported practice (outcome 5)	X							X
Survey: diarrhoea+respiratory illness (outcomes 6+7)	X							X
Intervention adherence indicators		X	X	X	X	X	X

^*Visit six for households receiving behavioural promotion (Arms AB+B)/visit two households receiving the hardware intervention (Arm A).

HWF, handwashing facility.

Figure 5. Study design flow diagram.

Household enrolment and baseline data collection started on 14 March 2025 and was completed on 30 May 2025. Endline is due to begin on 22 September 2025 and will be completed by 5 December 2025.

Sample size

For the sample size calculation, we modelled the primary outcome as a set of repeated binary outcomes (multiple handwashing opportunities over the observation window for one person) assuming intra-individual correlation. We assumed that the prevalence of HWWS at a handwashing opportunity would be 25% in the control group.⁴ We also assumed an average of six observed handwashing opportunities per household per 3-hour observation period based on prior research in Malawi,²⁸ and unpublished data in Zambia and Uganda; and an intra-cluster correlation coefficient of 0.3 based on unpublished data from Uganda. To account for five multiple comparisons (A vs C, B vs C, AB vs C, AB vs B, AB vs A), sample size calculations were adjusted by the Bonferroni method (family-wise α/5) as a conservative assumption.²⁹ We calculated that a sample of 450 clusters (households) per arm would allow us to detect an absolute minimum difference in HWWS of 6.6% (proportion of opportunities in which hands were washed with soap) between any two groups 6 months after intervention delivery begins, with 80% power (α=0·01).

Recruitment

A sampling interval for household enrolment was calculated using ward-level population estimates from the central statistics office in Zambia, assuming a household size of five (based on urban estimates from the Zambia Demographic and Health Survey).³⁰ We assumed that any inaccuracies would be consistent across wards. The total sample was distributed proportionally across wards, and the interval (k) was calculated as k=N/n, where N is the estimated number of households in the ward and n is the target sample size for that ward. This gave a preliminary sampling interval of 49 households (table 4). Assuming an average of four households per plot,³¹ data collectors will visit every 12th plot.

Table 4. Sampling interval calculation.

Ward	Population (p)	Estimated number of households per ward (n=p/5)	Proportion (p=N/total N)	Sample size per ward (n=p×1800)	Sampling interval (=N/ n)
Chainda	38 393	7679	0.09	157	49
Mpulungu	57 596	11 519	0.13	235	49
Chawama	70 181	14 036	0.16	286	49
Mtendere	106 128	21 226	0.24	433	49
Kanyama	169 253	33 851	0.38	690	49

Sampling and enrolment will be concurrent and conducted by members of the study data collection team. Data collectors will be distributed evenly across the zones within wards. Teams will proceed to the first plot or free-standing house along their sampling interval. In a plot, the team will select one household at random using a random number generator available on data collection tablets. If a household is empty at the time of enrolment, two additional attempts will be made to contact that household—once on the same day and once the next day. If households are unreachable or not eligible, the next household in the plot or along the street will be approached.

Allocation

Sequence generation

Randomisation will be stratified by ward and use a permuted block design to reduce predictability and ensure approximate balance among the four trial arms over the 11-week enrolment period. Prior to enrolment, the trial statistician (SB), with no prior knowledge of the participants, households or communities, will use randomly permuted blocks (sizes 4, 8 and 12) to generate a randomisation sequence for each ward with its length corresponding to the expected enrolment for the respective ward.

Implementation and concealment mechanism

The random sequences will be embedded into a password-protected online dashboard (Rshiny app) integrated with the electronic data capture system Open Data Kit (ODK), which will be used to assign households to treatments automatically and sequentially as they are enrolled, according to fixed characteristics of ward and precise (millisecond) timestamp of when enrolment data were submitted to the ODK server. The original randomisation sequence, including details of blocking, will be kept securely and the resulting allocations will be visible only to study coordinators and made available only to those responsible for providing or preparing the interventions and not to those enrolling households or collecting outcome data at any point.

Blinding

Due to the nature of the intervention, it will not be possible to blind study participants to intervention allocation. This is an inherent challenge in behaviour change and environmental health research and is an acknowledged limitation. Data collection teams—who are separate from the team who will deliver the intervention—will be unaware of study arm allocation. However, it is possible they will notice the presence of hardware while conducting structured observations at endline. Investigators aware of study objectives and hypotheses will not participate in data collection or intervention delivery. Investigators conducting data analysis will be blinded to intervention allocation.

Data collection methods

Structured observations

Outcomes will be measured using 3-hour structured observations at baseline and endline (6 months after intervention delivery begins). One eligible adult from each household will be observed at each timepoint. If a child under 5 years is present, the adult (18 years or older) most responsible for caring for the child at the time of the visit will be selected for observation. If a child under 5 years is not present, the adult (18 years or older) most responsible for childcare will be selected for observation. If such an individual is not present, another randomly selected household member (18+ years) will be observed instead.

A list of key handwashing behaviours that we will observe with definitions is provided in table 5. Other behaviours we will observe include handwashing opportunities (box 1) and hygiene management activities (refilling handwashing facility, moving soap near to handwashing facility, moving handwashing facility, preparing soap and disposing of grey water from handwashing facility).

Table 5. Handwashing behaviour definitions.

Outcome category	Definition
Hands washed with soap and water (HWWS)	Both hands washed under running water (eg, from a tap or jug/cup) with soap
Hands rinsed (water only)	Both hands washed under running water (eg, from tap or jug/cup) with no soap Dipping hands in a basin with water to clean hands (with or without soap) Both hands have contact with water (with or without soap), but other activities are being performed other than handwashing (eg, doing laundry or washing dishes)
No action	Only one hand washed No action

HWWS, handwashing with soap.

Observation data will be captured using electronic data capture forms on ODK. Forms will be configured to automatically record a timestamp when a data collector documents a behaviour. To determine whether hands were washed at a handwashing opportunity, we will allow a 3 min timeframe between the opportunity and the handwashing behaviour. This is to account for the time taken to fetch soap from the household when handwashing should occur after a handwashing opportunity.¹⁴ Past 3 min, or if the household member engages in another activity between the handwashing opportunity and handwashing, the event will be classified as ‘no action’. If the household member does not wash their hands at a handwashing opportunity but performs an activity involving the use of soap (eg, laundry or washing dishes), within 3 min before/after the key occasion, this will be recorded as ‘contact with soap and water but not HWWS’ but categorised as hand rinsing.

Structured observations will be conducted for 3 hours, between 07:00 and 16:00. Our prior work in both this context and other settings has identified two critical windows where we anticipate higher numbers of hand hygiene opportunities—07 to 10:00 (specifically around morning mealtimes and domestic chores) and 13:00 to 16:00 (specifically around midday and during evening meal preparation and the return of individuals to the home after going to the market or employment). We anticipate that due to data collection logistics (one observer completing two observations per day), approximately half the sample will be observed at each timepoint (early morning vs afternoon observation).

Data collection staff will position themselves in a place where most of the household—specifically food preparation areas and toilet infrastructure—are in view. Care will be taken to ensure the observer’s position and presence does not disturb the daily organisation and activities of the household. Indoor observations will occur only in shared or common spaces within the home. Based on previous research in the area, the behaviours of interest (food preparation, childcare, etc.) occur in the areas immediately outside the house.

While there are known limitations to structured observations, it remains the gold standard for the measurement of routine behaviours in the domestic context.^{32 33} Other measures of behaviour using self-reported practice have very low reliability and poor association with actual behaviours.^{34 35}

Household surveys

A household survey will be completed after the observation at baseline and endline. At baseline, the survey will collect information on household demographics, including assets, income and household size, as well as access to WASH services and experiences of water insecurity. The survey will also collect demographic information on each household member as well as self-reported cases of diarrhoea or respiratory illness in the past week. At endline, the survey will reassess household WASH access and update selected demographic information, including diarrhoea and respiratory illness among household members. Demographic data for the specific household member observed at endline will also be re-collected. An additional module at endline will assess exposure to the intervention as a measure of contamination.

The household survey completed at baseline and endline will include a health behaviour module. Participants will be asked to self-report when they washed hands yesterday, then will be asked when they believe HWWS is important. To reduce potential bias associated with focusing solely on handwashing, the survey will also incorporate additional modules on child immunisation, child feeding and child engagement/play. These modules are adapted from UNICEF’s Multiple Indicator Cluster Sruveys (MICS-6) questionnaire for children under 5 years. All surveys will be completed after observations to avoid biasing behaviour. Surveys will be completed using electronic data capture forms on ODK.

Data quality

Data collection teams will undergo a week-long training before study implementation, focusing on data collection methods. The training will include practical sessions where team members role-play household visits and practice translating and delivering questions in different languages (English, Nyanja and Bemba). Additionally, they will conduct practice observations using videos of domestic settings. To reinforce learning, team members will complete a test observation using a video, with feedback provided on submitted data.

Process data collection

Process data will be collected by intervention delivery workers during their intervention visits using electronic data capture forms on ODK. Data captured includes, but is not limited to, the number of household members present (all arms), reported content discussed (all arms) and condition/use of the Kalingalinga buckets (Arms AB+A). Households will be asked if they have received any other health-related interventions before (at baseline) and during the study period (at endline).

Retention

Participants will receive study booklets where the dates of their upcoming visits will be recorded. Separate booklets will be maintained for intervention delivery workers and data collection teams (who conduct baseline and endline assessments) to prevent unblinding. Before endline data collection, intervention delivery workers will remove intervention booklets to ensure data collectors remain blinded to household allocation.

Data management

All households will be given a QR code corresponding to a unique alphanumeric code which will be used to link household data through the various stages of the study. QR codes will be kept with the household and attached to signed consent forms. Electronic data entry forms will be built using ODK. All data will be anonymised prior to sharing with the London School of Hygiene & Tropical Medicine (LSHTM). Data will be transferred between LSHTM and CIDRZ using online secure file sharing systems. De-identified data will be sent weekly for monitoring purposes.

Participants will have the right to gain access to, change or oppose the collection of any data during the interviews or observations. Participants have the right to withdraw altogether from the study, without providing any justification, by contacting any of the investigators listed on their participation information sheet by mail or phone. Data and all appropriate documentation will be stored for a minimum of 5 years after the completion of the study, including the follow-up period.

Statistical analysis

The analysis and reporting of results will be in line with the Consolidated Standards of Reporting Trials (CONSORT) statement for randomised controlled trials³⁶ and its extension for multi-arm trials.³⁷ Statistical analysis will be conducted using R (V.4.5.0). Full details of all models and analytic procedures, including covariates to be included in the adjusted models, additional subgroup analyses and sensitivity analyses, are set out in a statistical analysis plan (SAP; V.1 dated 30 January 2026), finalised before investigators had access to endline data (a modification to the SAP (V.2 dated 6 May 2026) was published and locked prior to unblinding, revising the age stratum for tertiary outcome analyses).³⁸

Our analysis will concern five pairwise comparisons:

1. Behavioural promotion against control (B vs C).

2. Hardware and supplies provision against control (A vs C).

3. Behavioural promotion+hardware and supplies provision against control (AB vs C).

4. Behavioural promotion+hardware and supplies provision against behavioural promotion alone (AB vs B).

5. Behavioural promotion+hardware and supplies provision against hardware and supplies provision alone (AB vs A).

The remaining comparison (A vs B) may be analysed as an exploratory analysis if there is evidence of effectiveness of either intervention.

A two-stage approach to analysis of each outcome across the multiple trial arms will be used, as described in Korn et al,³⁹ involving first comparing individual intervention arms A, B and AB to arm C (pairwise comparisons 1–3), followed by comparing AB (the combination of A and B) to the best of A, B and C (comparison 3, 4 or 5) if AB is found to be superior to C in the first step (the null hypothesis for comparison 3 is rejected). This approach allows us to retain a larger statistical power than testing the five comparisons simultaneously as a different nominal significance threshold is used for each step, while maintaining the same overall family-wise type I error rate.

Descriptive statistics of demographic and outcome measures at baseline will be reported by trial arm to determine any imbalances. The primary outcome and secondary outcomes 2a–h (HWWS at specific handwashing junctures) will be analysed using Poisson regression, estimating the rate ratio of HWWS events at handwashing opportunities between study arms, with the log of the total number of handwashing opportunities included as an offset. Results will be presented as incidence rate ratios with appropriate CIs. The three-level ordinal hand-rinsing outcome (secondary outcome 3) will be analysed using mixed-effects ordered logistic regression with a random effect at the household level. Other secondary and tertiary outcomes will be compared between arms using appropriate regression models, as specified in the SAP, taking account of clustering where necessary. Distributional assumptions will be verified as appropriate.

Sensitivity to clustering at other levels (such as the data collector conducting observations) may be explored. Unadjusted and adjusted results will be presented for all analyses. All analyses will adjust for ward (stratification variable) and presence of an on-site water source at endline (defined as access to any water source on the premises or plot). This structural characteristic is unlikely to be affected by the handwashing intervention over the study period; the endline measure is used in preference to baseline to retain households missing baseline covariate data in the primary analysis and to reflect water access status at the time of outcome measurement. Secondary analyses of handwashing outcomes will additionally adjust for any HWWS at a handwashing opportunity at baseline, the observed participant’s age and sex, time of day of the observation and household water insecurity (Household Water Insecurity Experiences (HWISE-4) scale score ≥4). Households missing a baseline observation will not be included in secondary analyses.

Bias due to missing data will be investigated by comparing the baseline characteristics and randomised allocation of households lost to follow-up with those retained at endline, to assess whether attrition appears informative. Where patterns suggest differential attrition, sensitivity analyses will additionally adjust for characteristics associated with missingness. We will also explore the extent of missingness due to zero handwashing opportunities during the observation period. Complete case analysis will be used if missingness is less than 10%; if missingness exceeds 10%, sensitivity analyses based on multiple imputation or inverse probability weighting will be conducted.

Secondary outcomes 2a–h (HWWS at each specific handwashing opportunity) will be compared in the same way as the primary outcome, restricting HWWS events and the offset to those relevant to each specific juncture. Since these outcomes are components of the primary outcome and analyses are not adjusted for multiple testing beyond the between-group comparisons, interpretations will be highly cautious. Analysis will not proceed for a specific juncture unless there are at least 120 households in total and at least 25 per arm with at least one opportunity of that type.

To assess possible heterogeneity of the intervention effects, subgroup analyses will be undertaken as secondary analyses, including the planned subgroup of water insecurity experienced by the household measured using the Household Water Insecurity Experiences Scale.⁴⁰

As the primary aim of the trial is to evaluate the effectiveness of offering the intervention/s, a modified intention-to-treat principle will be used in the primary analysis of all outcomes, including all randomised households with observed endline outcome data regardless of baseline data availability, intervention receipt or protocol adherence. Endline outcome data will still be collected for households that did not receive the intervention as per protocol, and these households will be included in the primary analysis according to their original group, even if they received no intervention, partial intervention or the wrong intervention. Per-protocol analyses may be conducted as exploratory analyses, but these will be interpreted with caution. As the primary analysis will be based on differences between arms at endline, households who were missing at endline or who withdrew from the trial between baseline and endline will not be included in the primary analysis.

Data monitoring and auditing

As data are only collected at two time points and there is no interim analysis, a data monitoring committee is not needed. The study may be subject to audit by LSHTM under their remit as sponsor and other regulatory bodies to ensure adherence to Good Clinical Practice (GCP).

Ethics and dissemination

Research ethics approval

Ethical approval was obtained from LSHTM‘s Research Ethics Committee, London, UK (Ref: 31387), the University of Zambia Biomedical Research Ethics Committee (Ref: UNZABREC 5834–2024) and the Zambia National Health Research Authority (NHRA−1823/23/12/2024).

Protocol amendments

Important protocol modifications will be submitted to relevant ethics and regulatory committees mentioned above and updated in the trial registries (clinicaltrials.gov). Participants already enrolled in the study will be re-consented on updated consent forms.

Consent

There are two levels to consent: (1) consent to participate in the study intervention or control and (2) consent for data collection (online supplemental text 4). Procedures are outlined for each of these below.

All study information will be read in a language participants understand (ensuring a witness is present if the participant is illiterate). Information sheets and consent forms will be translated into the local language (Bemba and Nyanja) by the local research team. Questions on any aspect of the study will be invited from the participant. The study team will check for understanding of the information provided using a basic five-question knowledge assessment questionnaire. If not understanding, the forms should be reviewed again with the participant.

If the participant understands the forms and is willing to participate, the team will ask the participant to sign the consent form. If the participant is illiterate, the participant should leave a thumbprint or X on their signature line, and the witness will sign in the appropriate part of the consent form.

Consent to participate in the study

Informed consent for household study participation will be obtained at enrolment by the data collection team. The team will explain the purposes of the study in broad terms to households that meet eligibility requirements and invite them to learn more about the study. If they agree, the team will read the Study Participant Information Sheet (PIS) to available members of the household, or ask the participant to read themselves, and give the household time to ask questions. The Study PIS will explain randomisation procedures, anticipated interaction/time with the households, participant rights and data collection procedures. The Study PIS will explain that participating households have the right to end involvement in the study at any point without penalty.

The team will obtain written informed consent from the head of household or a nominated representative for household participation in the study, on behalf of all household members. If the head of household is not present, we will obtain verbal consent from the head household for a nominated representative to provide written consent for household study participation. Additional signatures will be collected from all adults in the households present at the time of enrolment.

Consent for data collection

A separate informed consent process will be completed for members of participating households that participate in specific data collection activities. Consent will be obtained at both baseline and endline, with the household member being observed. The data collection PIS will provide a brief overview of the specific data collection activities, explain the rights of the participants, and that they are free to terminate data collection at any moment. Participants will also be informed that they can choose to have their data destroyed or withdrawn. Written informed consent will be collected for the person who is participating in data collection.

Confidentiality

Identifiable data of the participants will be stored securely and their confidentiality protected. All electronic data will be stored on a secure, password-protected server. Only named individuals on the protocol will have access to the data. All mobile devices used for data collection will be encrypted and password protected. Any non-electronic documents will be stored in a locked filing cabinet, with access limited to field supervisors and study team members.

Data access

De-identified participant data, data dictionary, and analytical code used to process the data will be posted on a public repository for third-party use. To completely anonymise the data prior to uploading it onto an open data site, any information that holds the potential to identify the respondents will be redacted. Explicit consent to put de-identified data on a public repository will be sought from trial participants. Only named investigators will have access to the full trial dataset.

Ancillary and post-trial care

At the end of the study period, the most effective combination of interventions will be delivered to households in the control group. LSHTM holds Clinical Trial/Non-Negligent Harm Insurance and Medical Malpractice Insurance policies which apply to this trial. Due to the nature of the study, we anticipate the risk of harm to be very low to participants. Participants are informed to contact the local principal investigator or the research ethics committee/research governance and integrity office if they experience harm or injury because of taking part in the study.

Dissemination policy

All results from the study will be submitted for publication in peer-reviewed journals. All code and fully anonymised data will be posted on a public repository for third-party use. The final list and order of authors will reflect each researcher’s contribution and will comply with the Vancouver recommendations. Results may also be communicated through additional formats, such as a plain language summary.