Effectiveness of the Moments that Matter parenting programme on caregiving practices, family violence, mental health and early child development in Kenya: a cluster-randomised controlled trial

Abstract

Introduction

Globally, parenting interventions have been shown to improve caregiving practices and child development. However, rigorous evidence from sub-Saharan Africa remains limited, particularly for programmes that leverage multiple community-based delivery platforms. We evaluated the effectiveness of the Moments that Matter (MTM) parenting programme delivered by community volunteers in partnership with faith leaders on child development and caregiving-related outcomes in Kenya.

Methods

We conducted a cluster-randomised controlled trial in two counties in western Kenya. Villages were randomly assigned to the MTM parenting programme or a standard-of-care waitlist control, stratified by county. Eligible participants were primary caregivers of children aged 0–18 months. The intervention consisted of monthly group sessions and home visits delivered over 18 months. Caregivers were surveyed at baseline and endline. The primary outcome was child development measured using the Caregiver Reported Early Development Instruments. Intervention effects were estimated using mixed-effects models with a difference-in-differences, intention-to-treat approach.

Results

Between February and March 2023, 595 caregiver-child dyads in 46 villages were enrolled and randomly assigned to the intervention (n=310) or control (n=285). At endline, there were no overall intervention effects on child development outcomes. Moderate-to-large effects were observed across multiple secondary outcomes, including caregiver stimulation, harsh discipline, caregiver mental health, social support, intimate partner violence victimisation, child dietary diversity, and male caregiver engagement, as reported by female primary caregivers. Exploratory subgroup analyses revealed treatment heterogeneity by county, with small significant improvements in child language and socioemotional development in one county but null effects in the other.

Conclusions

A multicomponent parenting programme delivered in partnership between community volunteers and faith leaders improved multiple caregiving and family well-being outcomes but did not lead to overall improvements in child development. Findings highlight the potential of integrating multiple community-based platforms for delivering parenting programmes, while underscoring the need for implementation research to enhance impacts on child development at scale.

Trial registration number

NCT05796934.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• The Moments that Matter parenting programme, delivered primarily by community volunteers with supportive engagement by faith leaders, improved multiple domains of nurturing care, including caregiver stimulation, caregiver psychosocial well-being, nutrition-related practices, reduced harsh discipline and intimate partner violence, and increased maternal reports of male caregiver involvement.

• Despite these gains, MTM had no overall impact on early child development outcomes.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The findings suggest that parenting programmes can be integrated through community- and faith-based platformsin low-resource settings.

• The results highlight the need to more explicitly strengthen opportunities for early learning and responsive caregiving and to incorporate more intensive behaviour change strategies in order to achieve meaningful improvements in early child development outcomes.

Introduction

Early childhood development (ECD) encompasses young children’s cognitive, language, motor and socioemotional skills, which collectively form the foundation for lifelong health, learning and well-being.¹ Globally, approximately 43% of children aged under 5 years are at risk of not reaching their developmental potential, with the highest burden in sub-Saharan Africa.² Multiple interrelated risk factors contribute to poor child development, including poverty, malnutrition, infectious diseases, limited access to maternal and child health services and suboptimal caregiving environments.^{3 4} These risk factors can also increase caregiver stress and adversely affect family caregiving environments and caregivers’ capacities to provide nurturing care for young children.^{5 6} Interventions that address these interconnected risks and strengthen protective factors are critical for supporting healthy child development in low-resource settings.⁷

Parenting interventions are among the most effective strategies for supporting young children’s development.⁸ These interventions often target caregiver stimulation, responsive caregiving, positive discipline and child behaviour management and have been shown to improve multiple domains of child development across diverse global contexts.⁸ Meta-analytic evidence indicates that parenting programmes implemented in low- and middle-income countries (LMICs) tend to yield larger effects on child development outcomes than those in high-income settings and those with explicit programme components to increase opportunities for early learning and responsive caregiving are also more effective on child development outcomes.⁹

Most parenting programmes in LMICs have been delivered by a single delivery agent, typically community health workers, non-governmental organisation staff or volunteer caregivers. Leveraging multiple delivery agents may enhance programme reach, acceptability and sustainability. In many contexts, faith leaders are influential community figures who are widely trusted and embedded within community life, and who can play an important role in shifting social norms related to parenting and ECD.^{10 11} Integrating faith leaders alongside community volunteers represents a promising approach to embedding nurturing care and parenting messages within existing community structures. However, evidence on the use of this delivery approach in parenting programmes in LMICs remains limited.

In sub-Saharan Africa, there has been growing implementation of parenting programmes to promote child development. Regional syntheses and evidence reviews document increasing investments in supporting caregivers and young children.^{12 13} However, they also highlight that relatively few programmes have been evaluated using rigorous designs and that evidence of impacts on child development remains inconsistent and influenced by programme design and implementation quality. In Kenya specifically, most evaluations of parenting programmes to date have been relatively small, relied on quasi-experimental designs, and reported improvements in parenting behaviours but null effects on child development.^14–16 One notable exception is a cluster-randomised controlled trial (RCT) in western Kenya that combined group sessions and home visits, was delivered by community health volunteers, and demonstrated improvements in both maternal parenting and child development outcomes.¹⁷ Together, this literature underscores the need for more rigorous evaluations of scalable, community-based parenting programmes to better understand their potential for improving both caregiver-level outcomes and child development.

This study evaluates the effectiveness of the Moments that Matter (MTM) parenting programme implemented in rural western Kenya. An earlier quasi-experimental evaluation of MTM in western Kenya found improvements in caregiver stimulation but no detectable impacts on child development.¹⁵ Building on findings from this earlier evaluation, the MTM model was revised and shortened from a 24-month to an 18-month duration while retaining its core multicomponent curriculum delivered through group sessions and home visits. This cluster-RCT rigorously evaluated the effects of the revised MTM programme on a comprehensive set of outcomes, including child development, nutrition, parenting, caregiver mental health and family violence, in line with the programme’s theory of change. We hypothesised that children in intervention communities would have greater improvements in child development and that caregivers would report more positive parenting, other nurturing care practices and psychosocial and mental health, compared with those in control communities.

Methods

Trial design and registration

We conducted a two-arm cluster-RCT with baseline and endline assessments to evaluate the effectiveness of the MTM parenting intervention on child development and caregiver-level outcomes in western Kenya (registered on 7 March 2023 with ClinicalTrials.gov: NCT05796934).

Setting and participants

The study was conducted in collaboration with Episcopal Relief & Development in partnership with the Anglican Church in Kenya Development Services (ADS). ADS-Nyanza oversaw programme implementation in Borabu subcounty in Nyamira County, while ADS-Western oversaw implementation in Luanda subcounty in Vihiga County. These two subcounties were selected in consultation with county government officials and because of their high levels of poverty and poor maternal and child health indicators. In addition, county government officials indicated that no large-scale parenting programmes were known to be operating in either subcounty at the time the study was planned.

Eligible participants were primary caregivers of children aged 0–18 months residing within the selected sublocations. Households were required to have lived within the study area and intend to remain for the duration of the programme. Caregivers younger than 18 years or households where the index child was older than 18 months were excluded.

The sample size calculation was based on a longitudinal comparison of outcomes from baseline to endline (difference-in-differences analysis) and was designed to detect a minimum detectable effect (MDE) of 0.25 SD in the primary outcome of child development, with 80% power and a two-sided α=0.05. The MDE was selected as a conservative threshold to ensure adequate sensitivity to detect modest but plausible effects, given that MTM is a multicomponent parenting programme addressing multiple caregiving domains beyond early learning and that some multicomponent interventions have found relatively small effects on child development.^{18 19} The longitudinal design accounted for within-child correlation over time (assumed baseline-endline correlation of approximately 0.55), which increases statistical efficiency relative to post-test-only designs. Clustering at the village level was incorporated using an average cluster size of 13 households and an assumed intracluster correlation coefficient of 0.03, and the calculation allowed for up to 15% loss to follow-up. Under these assumptions, a target sample size of 282 primary caregiver-child dyads per study arm provides adequate power to detect effect sizes of approximately 0.25 SD or larger.

Randomisation and masking

Within each of the two selected subcounties, we implemented a multistage random sampling procedure. First, eight sublocations per subcounty were randomly selected and then randomly assigned in a 1:1 ratio to the intervention or waitlist-control arm using sequential numbering in Excel. Second, villages within each selected sublocation were randomly selected for participation in the study. Prior to village selection, ADS staff consulted with local government officials to map villages that were not participating in existing parenting programmes. Remaining eligible villages were listed and randomly ordered in Excel. The first six villages in each intervention sublocation (24 total) and the first five or six villages in each control sublocation (22 total) were then selected for study inclusion. Within each selected village, 13 primary caregivers were randomly selected for enrolment to match the programme’s intended group size for implementation.

Control villages continued to access existing health and community services, including routine maternal and child health services such as antenatal and postnatal care, immunisation, growth monitoring, treatment of common childhood illnesses and community-based health promotion activities. The study employed a waitlist-control design, whereby villages assigned to the control arm were offered the parenting programme after completion of the evaluation. Participants were not restricted from accessing other parenting programmes that became available during the study period.

Randomisation was performed by an independent researcher at Emory University who had no role in study implementation. Enrolment of clusters and participants was carried out by B&M Consult in collaboration with ADS-Nyanza and ADS-Western. Due to the nature of the community-based intervention, participants and ECD Promoters were not masked to assignment. However, enumerators for the evaluation operated independently from programme implementation and were kept blind to treatment status to the extent possible.

MTM parenting programme

MTM is a community-led, multicomponent parenting programme for caregivers of children aged 0–3 years that aims to promote ECD.²⁰ The programme was developed by Episcopal Relief & Development and is currently implemented in multiple countries across sub-Saharan Africa through partnerships with national faith-based organisations and often those affiliated with the Anglican Church. In Kenya, MTM has been implemented by ADS-Nyanza since 2019, whereas ADS-Western implemented MTM for the first time through this study. For this trial, the intervention was implemented from May 2023 to November 2024.

MTM is grounded in the Nurturing Care Framework (NCF),⁷ which emphasises five interrelated domains—good health, adequate nutrition, responsive caregiving, opportunities for early learning and safety and security—as essential for promoting children’s cognitive, language, motor and socioemotional development. Across the curriculum, different sessions emphasise specific NCF domains, including play and opportunities for early learning, responsive caregiving, nutrition and feeding practices, and non-violent discipline. The curriculum also addresses broader factors that influence caregiving and child development, including caregiver mental health, family violence prevention and male caregiver engagement. Together, these components aim to strengthen the psychosocial environment and reduce sources of caregiver and family stress, thereby creating conditions that support the adoption and maintenance of nurturing caregiving practices.

MTM follows an 18-month hybrid delivery model, combining monthly primary Caregiver Support and Learning Group (CSLG) sessions with individualised monthly home visits, resulting in a total of 36 programme contacts per primary caregiver. Trained community volunteers (ie, “ECD Promoters”) facilitate group sessions involving approximately 13 primary caregivers per village and provide tailored counselling to families during home visits. Although the programme primarily targets primary caregivers (most often mothers), the programme encourages the participation of male caregivers (fathers) to promote equitable caregiving and strengthen male caregiver–child interactions.

ECD Promotors used structured manuals to facilitate each CSLG, which included guidance on what to say, structures of activities and discussion questions. Multiple social and behaviour change strategies, including reflection activities, role-plays, guided practices of new behaviours and goal-setting, were embedded throughout the curriculum to improve nurturing care behaviours. Caregivers were encouraged to set specific goals using ‘Caregiver Actions to Practice Passports’, which prompt caregivers to identify actions to practice with their child, who in the household will carry them out and when they will occur in daily routines; these actions are reviewed in subsequent sessions and during home visits to reinforce practice and address barriers to adoption. Finally, each CGSL session included a dedicated opportunity for peer learning and support in which caregivers share experiences, challenges and strategies for implementing new practices in their daily routines with one another.

In addition to the ECD Promoters, community faith leaders received their own training through MTM and specifically on how to promote nurturing care through their existing roles within local religious networks. In this context, faith leaders included both clergy and lay individuals who served and supported members of their faith communities. These leaders were predominantly affiliated with Christian denominations (including Catholic, Seventh-day Adventist and African Initiated Churches), but this role was inclusive of leaders from other religions active in the community, including Islam. Rather than serving as primary facilitators of the parenting sessions, faith leaders were trained to reinforce programme messages by incorporating nurturing care and father involvement into sermons, Bible studies and community gatherings and by providing counselling to families when needed. Faith leaders were encouraged to attend caregiver group sessions or home visits when possible, although their regular participation was not required.

Approximately 6 months after programme initiation, MTM introduced two additional components: (1) Savings and Loan groups facilitated by a separate cadre of trained community volunteers and (2) kitchen-garden training and cooking demonstrations. These components were intended to improve access to diverse foods for young children, reinforce recommended feeding practices, reduce sources of caregiver stress and enhance caregivers’ sense of control over household resources. Activities could be integrated into CSLG sessions, home visits or Savings and Loan meetings. Finally, at the sublocation level, ECD Committees comprising community leaders and government representatives strengthened referral linkages and local ownership for sustainability.

Data collection procedures

Quantitative data were collected from primary caregiver-child dyads at baseline (23 February–10 March 2023) and reassessed at endline (13–25 November 2024). Due to heavy rains, school holidays and seasonal migration that reduced follow-up rates at endline, a supplemental revisiting round was conducted from 26 February to 10 March 2025 to reach caregivers missed during the main endline period. Data from this revisiting exercise were combined with the main endline dataset to create the full endline sample.

12 enumerators and two supervisors from B&M Consult were trained on the study protocol, interview procedures, research ethics and the survey modules. Training included classroom sessions and then field piloting for a total of 8 days at baseline and 14 days at the endline. Surveys were administered via KoboToolbox on Android devices and took approximately 1.5 hours per household. Supervisors conducted daily quality checks and tracked data collection. Data collection took place either at caregivers’ homes or at central community locations, depending on geographic dispersion, weather conditions and logistical feasibility in each village.

Outcomes

The primary outcome of this was child development, assessed using the Caregiver Reported Early Development Instruments (CREDI) long form for children aged 0–35 months.²¹ The CREDI measures caregiver-reported developmental milestones across cognitive, language, motor and socioemotional domains. We used raw scaled scores for the overall CREDI and for the four domain-specific scores (cognitive, language, motor and socioemotional development), as well as the CREDI mental-health subscale capturing behavioural and emotional difficulties.

At the endline, the Global Scales for Early Development (GSED) long form was added as an exploratory outcome, as it had not yet been publicly released for use during the baseline. The GSED is a direct observational assessment of child development that provides a single developmental score (D-score) representing cognitive, language and motor skills.^{22 23} Child development assessors received 7 days of intensive GSED training, and assessments lasted approximately 45 min per child. At endline, children’s socioemotional development was additionally assessed through direct observation by trained child development assessors using an adapted version of the Wolke behavioural scales.²⁴

Secondary outcomes spanned multiple domains related to parenting and psychosocial well-being. Primary caregivers’ stimulation practices were measured using an adapted version of the Family Care Indicators,²⁵ which captured the frequency of early child learning activities conducted with the child (eg, storytelling, singing and play) and were modified to reference the past week.^{19 26} Caregivers also reported on the variety of play and learning materials available in the home (eg, homemade toys and store-bought toys) and the number of books in the household, using the ECD module of the UNICEF Multiple Indicator Cluster Survey (MICS) questionnaire.^{19 27 28} Primary caregivers’ disciplinary practices were assessed using the UNICEF MICS Child Discipline Module, which measures both positive and harsh disciplinary approaches, including physical punishment and psychological aggression. Binary indicators were created for any violent discipline, any positive discipline, any physical punishment and any psychological aggression. Caregivers additionally reported whether the child’s birth had been officially registered.

For households in which a male caregiver was present in the index child’s life, primary caregivers provided parallel reports on the male caregiver’s stimulation and discipline practices using the same items and scales described above. Primary caregivers also reported the number of household chores (eg, washing dishes, cleaning and childcare) performed by the male caregiver in the past 2 weeks. This approach of obtaining primary caregiver reports of male caregiver involvement due to feasibility and representativeness considerations has been used in other evaluations of parenting programmes in low-resource settings.^29–31

Caregiver psychosocial well-being was measured through the Parenting Stress Index-Short Form, Parental Distress subscale and the Multidimensional Scale of Perceived Social Support. The parenting stress measure was calculated as a total score across the items, with higher scores indicating greater parenting stress. The social support measure assessed caregivers’ overall perceived support from various sources (eg, partners, family and friends), with higher scores reflecting stronger support networks.

Additional secondary outcomes included child nutrition, health and intimate partner violence (IPV). Child dietary diversity was assessed as an indicator of nurturing care related to nutrition, a core domain of the NCF also addressed in MTM, and was measured as the number of WHO-defined food groups (out of eight) consumed by the child in the previous 24 hours. Health indicators captured caregiver reports of any child diarrhoea, cough or fever in the past 2 weeks, as well as appropriate care-seeking behaviours for such illnesses. IPV was asked to female primary caregivers and measured based on acts of physical, emotional and economic violence and a composite indicator for any overall IPV victimisation over the previous 3 months.

Finally, several exploratory outcomes were added and not preregistered. These included caregiver report of having a kitchen garden; caregiver depressive symptoms measured using the Centre for Epidemiologic Studies Depression Scale-10-item; perceived community connectedness developed by the study team (‘having things in common with other caregivers’ and ‘feeling cared for by other caregivers’); caregiver financial worries developed by the study team (relating to food, household expenses and child costs; which was added to the survey only at endline); whether caregivers received any referral for themselves or their child in the past year, and indicators of financial security, including caregiver-reported total monthly income, current savings and access to credit during the past month. Full descriptions of measures are summarised in online supplemental table 1.

Analysis

We first examined descriptive statistics for all sociodemographic characteristics in the sample at baseline and assessed balance across study arms. We then evaluated attrition at endline and used logistic regression models to assess whether loss to follow-up was associated with study arm, child gender, child age at baseline, primary caregiver role, county and household wealth quintile. Among participants in the intervention arm, we assessed MTM programme exposure, including attendance, message recall and sources of information. At endline, caregivers were asked whether they recalled learning about broad MTM topic areas and, for each topic endorsed, to identify the source of information and describe what they had learnt. Enumerators then coded these unprompted responses against a prespecified list of key messages to classify recall within each topic area, and response options were not read aloud.

For the impact evaluation, we employed an intent-to-treat (ITT) difference-in-differences approach using mixed-effects regression models. Models were estimated separately for each outcome and included random intercepts to account for repeated measures within individuals and clustering at the village level. Inverse probability weights were applied to address differential missingness. Fixed effects included time (baseline vs endline), study arm (intervention vs comparison), the time-by-arm interaction, child sex, child age, primary caregiver relationship (mother vs other), primary caregiver age, wealth quintile (assessed using the first principal component from 20 household assets), endline survey round (November 2024 vs February/March 2025) and county (Nyamira vs Vihiga). The GSED and observed socioemotional development outcome models additionally controlled for child development assessor and baseline child development using overall CREDI scores. Continuous outcomes were analysed using linear mixed-effects models, with interaction terms standardised by the pooled baseline SD to yield effect sizes presented in SD units (Cohen’s d). Binary outcomes were analysed using mixed-effects logistic regression, and estimates were reported as ORs. ITT analyses were conducted for the full sample and stratified by county.

We additionally estimated treatment-on-the-treated (TOT) effects for caregivers who attended at least 1, 4, 7, 11 and 17 MTM sessions or group visits, which are associated with any programme exposure and approximately 25%, 40%, 60% and 95% programme exposure, respectively. For the few exploratory outcomes measured only at endline, models excluded time and the interaction term, and instead intervention effects were estimated from the study-arm coefficient and represent between-group differences at endline.

Finally, we explored potential mediation of the intervention effect on child development (ie, overall CREDI and GSED scores) through several secondary outcomes (primary caregiver stimulation, learning materials, child dietary diversity and parenting stress) using path analysis. Mediators were first explored in separate models, then any significant mediators were included simultaneously in a multiple mediation model. Descriptive statistics, ITT analyses and TOT analyses were conducted using Stata V.18; mediation analyses were conducted in R V.4.5.2.

Patient and public involvement

Patients or the public were not involved in the design and conduct of this research.

Results

Between February and March 2023, 798 households across the 46 study villages were screened for eligibility, and 595 caregiver-child dyads were enrolled: 310 dyads from 24 intervention villages and 285 dyads from 22 control villages. Endline data were later obtained for 525 caregiver-child dyads (88% follow-up) (figure 1). Of these, 424 dyads were assessed during the main endline round in November 2024, and an additional 101 were assessed during a supplemental round in February–March 2025. Timing of follow-up (November 2024 vs February–March 2025) was not associated with study arm or baseline characteristics, except that households in Nyamira were more likely than those in Vihiga to be interviewed during the supplemental round (online supplemental table 2). Attrition rates were similar between intervention and control groups, and baseline characteristics were largely comparable between participants retained and those lost to follow-up.

Figure 1. Consolidated Standards of Reporting Trials (CONSORT) flow diagram.

Table 1 reports descriptive statistics for sociodemographic characteristics in the overall sample and by study arm. Overall, the mean age of the index child was 8.7 months (SD=5.5), most primary caregivers were mothers (91%), and 78% were in partnered relationships. Sociodemographic characteristics were generally balanced between study arms. The only notable difference between study arms was that the intervention group included more grandmothers as the primary caregiver (12% vs 3%) or fewer mothers (86% vs 96%), corresponding to a higher proportion of caregivers over 35 years.

Table 1. Sample demographic characteristics at baseline by study arm.

	Overall (n=595)	Study arm
		Intervention (n=310)	Control (n=285)	P value
Child age in months, mean (SD)	8.7 (5.5)	8.8 (5.3)	8.6 (5.6)	0.66
Child sex, n (%)
Male	271 (45.6%)	145 (46.8%)	126 (44.2%)	0.530
Female	324 (54.5%)	165 (53.2%)	159 (55.8%)
Primary caregiver relation to child, n (%)
Mother	539 (90.6%)	273 (85.8%)	266 (95.8%)	<0.01
Father	6 (1.0%)	4 (1.3%)	2 (0.7%)	0.47
Grandmother	45 (7.6%)	36 (11.6%)	9 (3.2%)	<0.01
Grandfather	2 (0.3%)	2 (0.7%)	0 (0.0%)	0.17
Aunt	2 (0.3%)	1 (0.3%)	1 (0.4%)	0.95
Other	1 (0.2%)	0 (0.0%)	1 (0.3%)	0.34
Number of individuals living in household, mean (SD)	5.0 (1.8)	4.9 (1.9)	5.2 (1.7)	0.09
Currently in partnered relationship, n (%)	461 (77.5%)	239 (77.1%)	222 (77.9%)	0.82
Primary caregiver age, n (%)
18–24 years	205 (34.5%)	107 (34.5%)	98 (34.4%)	<0.01
25–34 years	269 (45.2%)	125 (40.3%)	144 (50.5%)
35–44 years	85 (14.3%)	50 (16.1%)	35 (12.3%)
45+ years	36 (6.1%)	28 (9.0%)	8 (2.8%)
Primary caregiver education completion, n (%)
No education	5 (0.8%)	3 (1.0%)	2 (0.7%)	0.32
Some primary school (incomplete)	84 (14.2%)	51 (16.5%)	22 (11.6%)
Completed primary school	248 (41.7%)	129 (41.6%)	119 (41.8%)
Completed secondary school	258 (43.4%)	127 (41.0%)	131 (46.0%)
Household wealth quintile, n (%)
Lowest	119 (20.0%)	67 (21.6%)	52 (18.3%)	0.66
Second	120 (20.2%)	63 (20.3%)	57 (20.0%)
Middle	119 (20.0%)	65 (21.0%)	54 (19.0%)
Fourth	119 (20.0%)	57 (18.4%)	62 (21.8%)
Highest	118 (19.8%)	58 (18.7%)	60 (21.1%)

Programme attendance and message recall

At baseline, few caregivers in either study arm had participated in any parenting education or support groups in the previous 3 months (6.1% intervention and 2.8% control). By endline, attendance in a parenting group increased substantially among caregivers in the intervention arm (91.3%) but was relatively low in the control arm (37.6%). Among control-group caregivers who reported attending a parenting group, activities were delivered by a variety of local actors, including community health promoters, faith-based organisations and other non-governmental organisations. At endline, 93.1% of intervention caregivers versus 8.4% of control caregivers reported having heard of the MTM programme, suggesting that increased exposure to parenting groups in the control arm likely reflected other parenting programmes rather than contamination.

Among intervention participants, attendance in the group sessions varied: 20.3% of caregivers reported attending none or had missing attendance data, 29.0% attended 1–6 sessions, 33.9% attended 7–16 sessions and 16.8% attended ≥17 sessions. Home visit participation followed a similar pattern, with 23.2% receiving none or missing data, 31.9% receiving 1–6 visits, 28.7% receiving 7–16 visits and 16.1% receiving ≥17 visits. Intervention participants reported learning about various topics through MTM, including parenting (73.6%), nutrition (66.5%), spiritual counselling (54.8%), men’s engagement (50.3%) and savings and financial management (32.3%). Among participants who recalled learning about these topics, ECD promoters were the most frequently cited source for parenting (97.4%), nutrition (89.2%) and men’s engagement (90.5%), while faith leaders were most commonly cited for spiritual counselling (89.5%). Many participants (61.0%) reported sharing information from their group sessions with others, most commonly a friend (75.7%) or their male partners (29.6%). Attendance and message recall differed by county, with somewhat higher participation in CSLG sessions and home visits in Nyamira, along with greater recall of messages related to spiritual counselling, savings and finance and men’s engagement, compared with Vihiga (online supplemental table 3).

We also examined participants’ interactions with faith leaders before and after MTM. At baseline, few participants reported having ever received counselling related to ECD, parenting or nurturing care from a faith leader (10.3% in the intervention group and 13.0% in the control group). By endline, these proportions increased in both groups, but with a substantially larger increase observed in the intervention group (63.2%) compared with the control (28.5%). Among intervention participants who recalled receiving messages from faith leaders, the most commonly reported topic was spiritual counselling, followed by men’s engagement, parenting, nutrition and savings and finance.

ITT results

Table 2 reports ITT impact estimates. No significant differences were observed between the intervention and control arms for any child development outcome, including overall and subscale CREDI scores, observed socioemotional development and GSED.

Table 2. Outcomes by study arm and timepoint, with intervention effect sizes.

Outcome	Baseline		Endline		Effect size type	Estimate	CI lower	CI upper
	Control	Intervention	Control	Intervention
Early childhood development
CREDI overall	44.22 (3.93)	44.74 (3.84)	51.84 (1.50)	52.13 (1.49)	d	−0.033	−0.162	0.095
CREDI cognition	46.95 (2.30)	47.25 (2.20)	50.75 (1.00)	50.95 (0.98)	d	−0.026	−0.176	0.123
CREDI language	47.30 (1.65)	47.50 (1.70)	51.62 (1.22)	51.85 (1.22)	d	0.032	−0.090	0.154
CREDI motor	46.30 (2.55)	46.63 (2.55)	51.47 (1.22)	51.66 (1.24)	d	−0.025	−0.143	0.093
CREDI social-emotional	46.65 (2.57)	46.95 (2.42)	51.39 (1.12)	51.66 (1.09)	d	0.000	−0.129	0.130
GSED developmental score	N/A	N/A	67.24 (8.49)	68.36 (7.75)	d	0.084	−0.026	0.195
Observed socioemotional development	N/A	N/A	32.87 (9.16)	33.57 (9.09)	d	0.071	−0.105	0.247
Parenting
Primary caregiver stimulation	7.96 (4.26)	8.21 (4.14)	10.49 (4.35)	12.8 (5.22)	d	0.422	0.175	0.669
No. of learning materials	2.16 (1.69)	2.52 (1.73)	3.90 (1.04)	4.30 (0.94)	d	−0.013	−0.217	0.190
No. of books	0.54 (1.48)	0.70 (1.49)	1.44 (1.75)	2.00 (1.98)	d	0.208	−0.042	0.458
Child safety and security
Primary caregiver positive discipline	53.33%	55.48%	97.22%	94.64%	OR	0.369	0.094	1.459
Primary caregiver any violent discipline	57.54%	60.00%	88.10%	55.17%	OR	0.091	0.038	0.220
Primary caregiver physical punishment	49.12%	50.32%	79.37%	48.28%	OR	0.135	0.058	0.310
Primary caregiver psychological aggression	45.61%	45.81%	68.65%	39.08%	OR	0.230	0.115	0.461
Birth registration	30.74%	35.16%	57.37%	63.08%	OR	1.121	0.533	2.361
Caregiver psychosocial well-being
Social support	3.37 (0.67)	3.35 (0.62)	3.29 (0.81)	3.56 (0.68)	d	0.442	0.255	0.629
Community connectedness	3.04 (0.96)	3.01 (0.92)	2.92 (1.13)	3.34 (1.07)	d	0.392	0.139	0.645
Parenting stress	32.92 (6.41)	33.89 (7.16)	33.38 (7.52)	30.98 (7.28)	d	−0.496	−0.666	−0.325
Primary caregiver depressive symptoms	8.24 (5.87)	9.25 (6.69)	9.35 (7.23)	7.80 (6.27)	d	−0.452	−0.660	−0.245
Primary caregiver financial worries	N/A	N/A	12.88 (5.05)	11.7 (5.38)	d	−0.283	−0.427	−0.138
Any IPV	46.81%	47.62%	46.77%	31.16%	OR	0.399	0.182	0.875
Physical IPV	15.74%	18.65%	14.93%	9.30%	OR	0.332	0.118	0.937
Emotional IPV	40.43%	41.67%	38.81%	24.19%	OR	0.341	0.153	0.757
Economic IPV	32.34%	28.17%	29.35%	15.81%	OR	0.366	0.164	0.818
Nutrition, health and financial security
Kitchen garden	72.28%	67.74%	73.02%	84.67%	OR	3.360	1.602	7.051
Child dietary diversity	3.52 (1.61)	3.37 (1.54)	3.90 (1.39)	4.36 (1.38)	d	0.381	0.126	0.635
Child illness	72.28%	77.74%	57.20%	56.30%	OR	0.786	0.417	1.484
Sought appropriate care for illness	48.54%	62.24%	44.76%	53.85%	OR	0.782	0.336	1.820
Received referral	N/A	N/A	13.89%	29.89%	OR	3.098	1.581	6.071
Income in past month (Kenyan Shillings, log)	1.60 (4.76)	1.84 (4.82)	6.37 (3.23)	6.89 (2.38)	d	−0.032	−0.299	0.235
Current savings (Kenyan Shillings, log)	−0.35 (4.01)	−0.32 (3.95)	0.86 (4.8)	3.36 (5.17)	d	0.521	0.307	0.735
Male caregiver outcomes
Male caregiver stimulation	3.75 (3.82)	4.36 (4.27)	5.66 (5.07)	8.47 (6.28)	d	0.468	0.234	0.701
Male caregiver positive discipline	30.45%	35.96%	80.00%	80.56%	OR	0.762	0.308	1.885
Male caregiver any violent discipline	30.00%	31.58%	49.55%	26.85%	OR	0.257	0.114	0.580
Male caregiver physical punishment	23.18%	26.75%	36.82%	18.98%	OR	0.206	0.086	0.491
Male caregiver psychological aggression	20.45%	18.42%	33.18%	18.06%	OR	0.397	0.155	1.015
Male caregiver household chores	2.19 (2.28)	2.13 (2.35)	1.12 (1.76)	2.22 (2.48)	d	0.579	0.301	0.858

Note. Effect sizes are derived from adjusted mixed-effects difference-in-differences models controlling for time, study arm, and their interaction, with additional covariate adjustment for child sex and age, primary caregiver characteristics, household wealth, survey round and county. Models include random intercepts to account for repeated measures within individuals and village-level clustering and apply inverse probability weights for missingness. Models for GSED and observed socioemotional development outcomes additionally adjust for child development assessor and baseline overall CREDI score. Estimates are based on intent-to-treat. Bold values indicate p<0.05.

d refers to Cohen’s d, or standardised effect size expressed in SD units.

CREDI, Caregiver Reported Early Development Instruments; GSED, Global Scales for Early Development; IPV, intimate partner violence; N/A, not assessed.

Primary caregiver stimulation improved significantly more in the intervention arm compared with the control (Cohen’s d (SD) 0.422, 95% CI 0.175 to 0.669). All measures of harsh discipline by primary caregivers were substantially lower in the intervention arm, including any violent punishment (OR 0.091, 95% CI 0.038 to 0.220), physical punishment (OR 0.135, 95% CI 0.058 to 0.310) and psychological aggression (OR 0.230, 95% CI 0.115 to 0.461). Availability of learning materials and books, primary caregivers’ use of positive discipline and child birth registration did not differ significantly by study arm.

In addition to improvements in their own behaviours, primary caregivers in the intervention arm reported greater improvements in male caregiver involvement compared with those in the control. Primary caregivers reported increased male caregivers’ stimulation (d=0.468, 95% CI 0.234 to 0.701), reduced violent punishment of any type (OR 0.257, 95% CI 0.114 to 0.580), reduced physical punishment (OR 0.206, 95% CI 0.086 to 0.491) and greater participation in household chores (d=0.579, 95% CI 0.301 to 0.858).

Significant improvements were also observed across all measured psychosocial well-being outcomes among primary caregivers in the intervention arm. These included social support (d=0.442, 95% CI 0.255 to 0.629), community connectedness (d=0.392, 95% CI 0.139 to 0.645), parenting stress (d=−0.496, 95% CI −0.666 to −0.325), depressive symptoms (d=−0.452, 95% CI −0.660 to −0.245) and financial worries (d=−0.283, 95% CI −0.427 to −0.138). Partnered caregivers in the intervention arm also reported lower victimisation of any form of IPV (OR 0.399, 95% CI 0.182 to 0.875), physical IPV (OR 0.332, 95% CI 0.118 to 0.937), emotional IPV (OR 0.341, 95% CI 0.153 to 0.757) and economic IPV (OR 0.366, 95% CI 0.164 to 0.818).

Finally, there were also greater improvements in several nutrition, health and financial security outcomes. Intervention caregivers reported higher child dietary diversity (d=0.381, 95% CI 0.126 to 0.635), increased access to kitchen gardens (OR 3.360, 95% CI 1.602 to 7.051), greater savings (d=0.521, 95% CI 0.307 to 0.735) and higher likelihood of having received a referral (OR 3.098, 95% CI 1.581 to 6.071). No differences were observed in child illness, appropriate care-seeking behaviours or past-month income.

ITT results by county

County-specific analyses revealed considerable heterogeneity in intervention effects (table 3). In Nyamira, significant positive intervention effects were observed across multiple child development outcomes, including language and socioemotional CREDI subscales, observed ratings of socioemotional development and primary caregiver stimulation and the number of books in the home. In contrast, Vihiga showed no significant intervention effects on child development or early learning outcomes.

Table 3. Intervention effect sizes, by county.

Outcome	Effect size	Nyamira			Vihiga			County differences p value
		Estimate	CI lower	CI upper	Estimate	CI lower	CI upper
Early childhood development
CREDI overall	SD	0.079	−0.069	0.228	−0.128	−0.322	0.066	0.090
CREDI cognition	SD	0.136	−0.043	0.315	−0.169	−0.389	0.051	0.032*
CREDI language	SD	0.149	0.024	0.275	−0.064	−0.222	0.094	0.034*
CREDI motor	SD	0.033	−0.123	0.189	−0.072	−0.251	0.107	0.378
CREDI social-emotional	SD	0.174	0.021	0.327	−0.153	−0.335	0.029	0.006*
GSED developmental score	SD	0.163	−0.012	0.339	−0.011	−0.146	0.124	0.179
Observed socioemotional development	SD	0.171	−0.077	0.419	−0.046	−0.291	0.198	0.048*
Parenting
Primary caregiver stimulation	SD	0.661	0.377	0.946	0.195	−0.115	0.505	0.023*
No. of learning materials	SD	0.093	−0.205	0.392	−0.112	−0.391	0.168	0.307
No. of books	SD	0.436	0.065	0.806	−0.015	−0.240	0.209	0.036*
Child safety and security
Primary caregiver positive discipline	OR	0.671	0.120	3.754	0.219	0.035	1.376	0.352
Primary caregiver any violent discipline	OR	0.076	0.022	0.267	0.105	0.033	0.333	0.663
Primary caregiver physical punishment	OR	0.112	0.036	0.353	0.157	0.046	0.537	0.628
Primary caregiver psychological aggression	OR	0.198	0.087	0.448	0.249	0.088	0.705	0.484
Birth registration	OR	1.725	0.634	4.692	0.832	0.259	2.674	0.311
Caregiver psychosocial well-being
Social support	SD	0.390	0.134	0.646	0.497	0.224	0.771	0.604
Community connectedness	SD	0.373	0.214	0.532	0.423	0.113	0.732	0.854
Parenting stress	SD	−0.413	−0.656	−0.171	−0.573	−0.785	−0.362	0.297
Primary caregiver depressive symptoms	SD	−0.426	−0.729	−0.122	−0.475	−0.736	−0.213	0.786
Primary caregiver financial worries	SD	−0.267	−0.490	−0.044	−0.337	−0.528	−0.146	0.921
Any IPV	OR	0.470	0.169	1.304	0.371	0.125	1.102	0.991
Physical IPV	OR	0.212	0.036	1.258	0.651	0.210	2.022	0.260
Emotional IPV	OR	0.474	0.178	1.259	0.249	0.068	0.917	0.644
Economic IPV	OR	0.478	0.173	1.324	0.314	0.088	1.123	0.736
Nutrition, health and financial security
Kitchen garden	OR	2.527	0.798	8.006	4.294	1.560	11.818	0.495
Child dietary diversity	SD	0.636	0.348	0.925	0.105	−0.306	0.515	0.038*
Child illness	OR	1.245	0.548	2.829	0.504	0.193	1.315	0.138
Sought appropriate care for illness	OR	1.018	0.252	4.108	0.663	0.223	1.970	0.552
Received referral	OR	5.546	3.046	10.098	1.627	0.405	6.539	0.015*
Income in past month (KSH, log)	SD	−0.267	−0.571	0.037	0.279	−0.072	0.629	0.023*
Current savings (KSH, log)	SD	0.657	0.362	0.953	0.379	0.068	0.690	0.221
Male caregiver outcomes
Male caregiver stimulation	SD	0.762	0.454	1.071	0.158	−0.126	0.443	0.004*
Male caregiver positive discipline	OR	0.902	0.268	3.037	0.621	0.159	2.430	0.628
Male caregiver any violent discipline	OR	0.168	0.055	0.510	0.403	0.122	1.333	0.130
Male caregiver physical punishment	OR	0.134	0.047	0.383	0.340	0.085	1.353	0.141
Male caregiver psychological aggression	OR	0.371	0.103	1.330	0.379	0.093	1.542	0.664
Male caregiver household chores	SD	0.884	0.460	1.309	0.203	−0.113	0.518	0.015*

Note. Effect sizes are derived from adjusted mixed-effects difference-in-differences models controlling for time, study arm, county and their interaction, with additional covariate adjustment for child sex and age, primary caregiver characteristics, household wealth, and survey round. Models include random intercepts to account for repeated measures within individuals and village-level clustering and apply inverse probability weights for missingness. Models for GSED and observed socioemotional development outcomes additionally adjust for child development assessor and baseline overall CREDI score. Estimates are based on intent-to-treat. Bold values indicate p<0.05.

^*Estimates are significantly different (at p<0.05) between counties.

CREDI, Caregiver Reported Early Development Instruments; GSED, Global Scales for Early Development; IPV, intimate partner violence.

Several male caregiver outcomes were also improved in Nyamira, including stimulation, use of violent discipline, physical punishment and participation in household chores, whereas no significant effects were observed in Vihiga. Intervention benefits on child safety and security outcomes and primary caregiver psychosocial well-being were consistent across both counties, though Vihiga experienced a significant reduction in emotional IPV, which was not observed in Nyamira. Nyamira additionally showed benefits for children’s dietary diversity and caregivers’ receipt of referrals, which were not seen in Vihiga, whereas caregivers in Vihiga reported significantly increased access to kitchen gardens, which was not observed in Nyamira.

TOT results

Figure 2 presents TOT effects of varying levels of MTM participation on overall CREDI scores and primary caregiver stimulation. Overall CREDI scores did not differ significantly across levels of MTM participation. Similar patterns were observed across CREDI domain-specific scores. In contrast, intervention effects on primary caregiver stimulation showed more of a dose-response trend, with greater improvements at higher levels of MTM participation.

Figure 2. Intervention effects on CREDI overall and primary caregiver stimulation for caregivers with differing Moments that Matter participation. Note. Dashed lines represent the average intent-to-treat effect estimated for each outcome. CREDI, Caregiver Reported Early Development Instruments.

Mediation results

Mediation analyses based on single-mediator models showed that primary caregiver stimulation and learning materials significantly mediated the intervention effect on both overall CREDI and GSED scores (online supplemental table 4). Child dietary diversity also emerged as a significant mediator of intervention effects on GSED scores. In the multiple-mediator models that include all candidate mediators simultaneously (online supplemental figure 1), primary caregiver stimulation remained a significant mediator of both overall CREDI and GSED scores, while learning materials remained a significant mediator of GSED scores.

Discussion

In this cluster-RCT, we evaluated the effectiveness of the MTM parenting programme delivered on caregiving practices, caregiver well-being and child development outcomes in rural Western Kenya. MTM improved multiple outcomes for primary caregivers, while impacts on child development were not observed in the overall sample and only in one county. These findings contribute to a growing body of evidence from LMICs showing that multicomponent parenting interventions can improve multiple caregiving domains, including psychosocial stimulation, parenting knowledge and attitudes, dietary practices, and caregiver mental health.^32–35

One of the most central caregiver-level outcomes targeted by MTM was parenting, and the programme improved primary caregivers’ reported stimulation practices and reduced harsh discipline. These findings are consistent with a robust body of evidence demonstrating that parenting interventions can achieve medium-to-large effects on caregiving knowledge, attitudes and practices across diverse settings in LMICs.^{36 37}

Despite improvements in caregiving practices, these gains did not translate into measurable effects in child development. One potential explanation is that, although MTM increased caregiver stimulation practices, these changes may not have been sufficiently frequent, sustained or integrated into caregivers’ everyday interactions with children to meaningfully impact child development. In addition, the programme’s content and delivery may not have sufficiently emphasised interactive, practice-based behaviour change strategies (eg, hands-on coaching, guided practice and feedback) that support caregivers in translating knowledge into consistent, responsive interactions with children to spur improvements in child development.^{9 30 38} Although children were invited to attend sessions, attendance was not required, which may have limited opportunities for caregivers to directly practise newly learnt play and interaction strategies with their children during sessions. Future adaptations of MTM may benefit from more intentional inclusion of children and structured caregiver-child practice during group sessions. This is also consistent with findings from our early qualitative process evaluation of the MTM programme, which revealed that responsive caregiving was among the least understood and least adopted concepts by both participants and delivery agents during implementation.³⁹ Prior studies from LMICs have similarly highlighted the importance of strengthened facilitator training, ongoing coaching and supervision—especially in programmes relying on community-based volunteers—to support high-quality delivery of early learning and responsive caregiving activities.^40–42

Although MTM did not produce consistent overall effects on child development across the full sample, results from the mediation analyses provide important insight into potential mechanisms of influence. Improvements in primary caregiver stimulation emerged as a consistent mediator of intervention effects on both overall CREDI and GSED, even after accounting for multiple other mediators (ie, learning materials, parenting stress and child dietary diversity) and in the absence of a significant overall direct effect. This finding further reinforces the importance of promoting opportunities to learn through play and stimulation, alongside strengthening responsive caregiving practices and aligns with other programme evaluations that have conducted mediation analyses and identified stimulation as a particularly influential pathway through which parenting interventions benefit child development.^{17 29 32 43}

Notably, MTM also improved caregiver mental health, perceived social support and reduced women’s IPV victimisation, despite these topics receiving only limited attention within the curriculum. These effects are noteworthy considering that mental health promotion and violence-prevention programmes explicitly designed to target these as primary outcomes have not consistently achieved success.^{18 44} These results strengthen the case that parenting programmes may be effective platforms for addressing other underlying risks in family dynamics that can positively affect multiple caregiver outcomes.^{32 45}

Primary caregivers in MTM reported improvements in not only their own outcomes but also their perceptions of their male partners’ parenting behaviours compared with the control group. Quantitative evidence on whether maternal-focused parenting programmes can benefit fathers’ parenting outcomes has been limited and mixed. For example, the Msingi Bora programme in Kenya found no effects of inviting fathers on fathers’ self-reported outcomes,¹⁷ whereas other maternal-focused parenting programmes in Pakistan⁴³ and Bangladesh²⁹ found increases in maternal reports of fathers’ stimulation. A prior qualitative evaluation of MTM in Kenya highlighted positive changes in fathers’ parenting practices and more equitable gender attitudes through both men’s direct programme participation and indirectly through mothers sharing programme lessons at home.⁴⁶ This study provides quantitative evidence that MTM improved female primary caregivers’ reports of fathers’ stimulation, engagement in household chores and reduced harsh discipline. However, these findings should be interpreted cautiously because fathers’ outcomes were reported by female primary caregivers rather than directly by fathers, considering that fathers were not the primary participants in the programme. Future studies should incorporate fathers’ self-reports and observational measures to better understand whether and how parenting programmes, including those that do not explicitly target fathers, may influence fathers’ parenting practices.⁴⁷

While the primary delivery agents of MTM were community-based volunteers, faith leaders were also trained to play a supportive role by reinforcing programme messages through their existing religious and community interactions and assisting families needing additional support. Given that faith leaders are trusted and widely embedded community figures in Western Kenya and leveraged as facilitators of other types of community-based programmes,⁴⁸ the training they received through MTM may have facilitated the dissemination of programme messages among participating families and potentially the broader community, particularly with respect to gender norms and promoting psychosocial well-being.^{49 50} However, this study was not designed to evaluate the independent contribution of faith leader engagement but rather the effects of the MTM programme as a whole. Future implementation research could examine how different delivery agents, including but not limited to faith leaders, can be engaged within community-based parenting programmes and how such multi-actor delivery approaches can be best designed and evaluated.

Finally, we uncovered notable variation in programme impacts by county. There were significant positive effects on ECD in Nyamira but none in Vihiga and significantly larger improvements in parenting in Nyamira, whereas impacts on IPV reduction and access to kitchen gardens were stronger in Vihiga. Importantly, the alignment between stronger impacts on caregiver stimulation and observed gains in child development in Nyamira reinforces evidence that improvements in early learning opportunities often accompany and help explain programme effects on child development.^{9 32} These county-level differences may also reflect implementation and contextual variations. The implementing partner in Nyamira had prior experience delivering MTM, which may have contributed to more consistent and high-quality delivery of key programme components and fostered greater caregiver engagement, while the implementing partner in Vihiga was new to MTM. At the same time, households in Vihiga appeared to be more disadvantaged with lower wealth and education levels. However, it is worth noting that county-stratified analyses were not powered a priori and are should therefore be regarded as exploratory and hypothesis-generating.

This trial represents one of the first rigorous evaluations of a parenting intervention delivered through a partnership between community-based volunteers and faith-based structures to support caregivers of young children. Strengths include its cluster-RCT design and the use of a wide range of theory-informed outcome measures. Key limitations of this study include reliance on primary female caregiver reports for nearly all outcome measures, including the CREDI, caregiving practices, child diet, psychosocial outcomes, fathers’ parenting behaviours and programme exposure, which may be subject to social desirability or reporting bias. Except for the GSED, all intervention effects observed were caregiver-reported and should therefore be interpreted with caution. Although the GSED was used as a direct assessment of child development, it was administered only at endline, and despite controlling for baseline CREDI scores, residual baseline differences may remain. Future research would benefit from incorporating multiple informants, including fathers directly, and observational measures when feasible to triangulate findings and strengthen validity.^{51 52} Several measures in this study (eg, learning materials) were adapted from existing tools and were not formally validated in this study context. In addition, the study was powered to detect a minimum detectable effect of 0.25 SD in the primary outcome of child development. However, observed effects in the overall sample were smaller than this threshold, suggesting limited power to detect small effects. Finally, sub-county analyses were not powered a priori and should therefore be interpreted as exploratory.

Conclusion

MTM improved multiple domains of nurturing care, including parenting, nutrition, safety and security and other caregiver- and family-level outcomes, including caregiver mental health, violence prevention and fathers’ engagement. It also highlights the feasibility of linking community and faith-based platforms for delivering nurturing care interventions. However, these caregiver-level benefits did not translate into measurable improvements in child development in the overall sample. Together, these results suggest that while MTM shows promise as a multicomponent parenting programme for strengthening caregiving environments in low-resource settings, additional refinements may be needed to achieve more consistent impacts on child development. Future iterations of MTM should consider placing greater emphasis on behaviour change strategies that strengthen opportunities for early learning and responsive caregiving practices within children’s daily routines to promote child development, while maintaining the relational and psychosocial elements of MTM that enabled the broad range of caregiver benefits. Overall, this trial contributes evidence on the potential of multicomponent parenting interventions to holistically support caregivers of young children in low-resource settings.

Data availability statement

Data are available upon reasonable request.