About this paper symposium

---

Panel information
Panel 27. Solicited Content: Global South

Paper #1
The Moderating Role of Caregiver Mental Health in Parenting Stimulation Interventions in LMICs
Author information	Role
Qi Jiang, School of Public Health, University of California, Berkeley, United States	Presenting author
Yiwei Qian, Research Institute of Economics and Management; Southwestern University of Finance and Economics, Chengdu, China, China	Non-presenting author
Lia Fernald, School of Public Health, University of California, Berkeley, United States	Non-presenting author
Scott Rozelle, Stanford Center on China's Economy & Institutions, United States	Non-presenting author
Abstract
Introduction: As caregivers shape a child’s early experiences, researchers and practitioners have increasingly turned their attention toward the role of caregiver mental health problems in early childhood development (ECD). Because the majority of the ECD interventions were designed to improve child development outcomes through enhancing parental stimulating practices, caregivers with mental health problems may benefit less from the ECD interventions through reduced responsive stimulation and lack of compliance. A few studies in the US found significant moderating effects of maternal depression in ECD interventions (Gardner et al., 2010; Oxford et al., 2023), yet no studies to date have been conducted in LMICs. The current study aims to examine the moderating effects of caregiver mental health problems in parenting stimulation interventions on intervention compliance, child development outcomes, and responsive stimulation. Methods: The data used in this study come from a cluster-randomized controlled trial conducted in 100 rural villages in northwestern China. Within each village, all children in our desired age range (6-24 months) and their primary caregivers were enrolled in the study, consisting of 2040 caregiver-child dyads from both the treatment and control arm. The intervention was year-long weekly one-on-one parenting stimulation sessions delivered in the parenting centers constructed in each treatment village. Baseline data were collected before the intervention and follow-up surveys were conducted after the intervention, both of which measured caregiver mental health (Depression Anxiety and Stress Scale-21), child development (the third edition of the Bayley Scales of Infant Development), stimulating parenting practices (Family Care Indicators), demographic information. Caregiver-child dyad’s attendance was collected during the intervention course. Ordinary least squares (OLS) regressions with interaction terms between treatment indicators and caregiver mental health problems were employed. Sensitivity analyses were conducted to examine the degree to which an unobserved covariate could bias our estimation. Results: We found caregiver depressive symptoms played a significant moderating role in treatment effects on child cognitive development and caregiver-child interactions. Specifically, the parenting intervention significantly improved the cognitive development of children whose caregivers did not have depressive symptoms at baseline by 0.167 standard deviations while the effect was not significant for children whose caregivers had depressive symptoms at baseline. Correspondingly, the treatment effects on time investment and material investment were 0.213 and 0.079 standard deviations among caregivers without depressive symptoms at baseline, respectively, while caregivers with depressive symptoms did not have significant effects on either of these outcomes. Similar moderating effects were found in caregiver stress symptoms. In contrast, caregiver anxiety symptoms did not show significant moderating effects. Conclusions: These results suggest that integrating mental health support into parenting programs can enhance the impact of interventions, even when mental health is not the primary focus. This study provides key policy insights for improving ECD outcomes through targeted mental health support in LMICs.

Paper #2
"The difficulty of life”: shocks as moderators of ECD intervention implementation
Author information	Role
Eleanor Tsai, School of Public Health, University of California, Berkeley, United States	Presenting author
Emanuela Galasso, World Bank, United States	Non-presenting author
Lia Fernald, School of Public Health, University of California, Berkeley, United States	Non-presenting author
Abstract
Introduction Exogenous shocks may pose challenges to the delivery and take-up of ECD interventions. For example, shocks such as economic, weather, and disease events can create logistical barriers to intervention participation, or cause income losses that increase the opportunity costs of intervention participation. Understanding how shocks moderate intervention implementation is an important step towards adapting interventions to maintain fidelity and quality even in unstable or unpredictable contexts. This study analyzes the impacts of a variety of shocks on the delivery and take-up of MAHAY Mikolo, an intervention integrating ECD sessions and health and nutrition sessions for children ages 6-30 months old in Madagascar. Methods MAHAY Mikolo randomized 75 villages into one of three treatment arms: [C], the status-quo health and nutrition sessions; [T], which is [C] + ECD sessions; or [T+], which is [T] + a play materials package introduced six months after baseline. ECD sessions ran fortnightly, in two cycles of 12 sessions each (“Cycle 1”: February 2021-August 2021; “Cycle 2”: September 2021-March 2022). Community health workers (CHWs) and village informants reported on their experience of economic, weather, disease, security, agricultural, and household shocks at ten survey waves throughout the intervention (Galasso et al. 2022). The primary outcomes in this study are (1) the number of ECD/health and nutrition sessions delivered (delivery); (2) the number of ECD/health and nutrition sessions attended (take-up); and (3) the likelihood of attrition from/transition between ECD session cycles (take-up). Results After baseline, almost all villages were subject to shocks in the last 30 days: only about 6 percent of villages experienced no shocks, whereas 87 percent of villages experienced 1-4 shocks, and the remaining 7 percent of villages experienced 5-7 shocks. Villages that experienced more weather shocks tended to hold fewer ECD sessions (an average of 8.3 sessions per month in villages that experienced no weather shocks, versus 5.8 sessions in villages that experienced 3 weather shocks) (Figure 1). Meanwhile, villages that experienced agricultural shocks tended to see fewer children in attendance at ECD sessions (7.2 children at a given session in villages that experienced no agricultural shocks, versus 5.9 children in villages that experienced 3 agricultural shocks) (Figure 1). By contrast, at the child level, no clear relationships emerge between the number of shocks and the number of ECD sessions attended or attrition by cycle (Figure 2). These findings demonstrate that the moderating effects of shocks may be heterogeneous not only by the type of shock, but also by the level at which delivery and take-up are measured (e.g. delivery agent, community, or individual participant). Planned work will explore the timing and co-occurrence of shocks to determine whether patterns of shocks diverge between villages with the highest and lowest rates of delivery and take-up. OLS, logistic, and autoregressive models will be used to test hypotheses that the number of shocks experienced is inversely related to delivery and take-up. Interpretation will draw on qualitative interviews with CHWs regarding challenges carrying out ECD/health and nutrition activities, perceived caregiver challenges, and facilitating factors.

Paper #3
How do WASH and Nutrition interventions work to improve child development? Evidence from Bangladesh
Author information	Role
Helen Osborne Pitchik, Ph.D., UC Berkeley School of Public Health, United States	Presenting author
Fahmida Tofail, icddr,b, Bangladesh	Non-presenting author
Mahbubur Rahman, icddr,b, Bangladesh	Non-presenting author
Peter J. Winch, Johns Hopkins Bloomberg School of Public Health, United States	Non-presenting author
Stephen P. Luby, Stanford University, Stephen P. Luby	Non-presenting author
Kara Rudolph, Columbia Mailman School of Public Health, United States	Non-presenting author
Lia C. H. Fernald, UC Berkeley School of Public Health, United States	Non-presenting author
Abstract
Though most children in LMICs experience sub-optimal water, sanitation, and hygiene (WASH), and interventions designed to improve these conditions can improve children’s developmental through reduced enteric pathogen infection, there is little existing research from intervention trials on WASH interventions and child development outcomes (Oriá et al. 2016; Pinkerton et al. 2016). Only a handful of interventions have evaluated the impact of interventions that target improvements in WASH on both child diarrhea or enteric infections and early child development outcomes. A cluster randomized controlled trial of individual and combined WASH and nutrition interventions in rural Bangladesh found improvements in both child diarrhea and child development outcomes at 7 years of age (manuscript in progress). However two other WASH interventions in rural Zimbabwe and Kenya did not find impacts on either diarrhea, enteric infection, or child development (Stewart et al. 2018; Gladstone et al. 2019). The mechanism of the effective WASH intervention in rural Bangladesh is unknown. Knowledge about how effective WASH interventions work can inform future interventions. We use data from the WASH-B intervention trial to examine mediators of intervention effects on child development outcomes at 7 years of age (Luby et al. 2018). The outcomes will be outcomes for which the initial intervention had sustained impact at 7 years of age, which are WPPSI-IV Full-scale IQ, prosocial behaviors as measured by the SDQ-prosocial, the home environment (measured by the middle childhood HOME), and maternal mental health (measured by the CESD-2060). The exposures examined at each time point are the intervention arms that demonstrated impact the outcome of interest (e.g., for the prosocial behaviors outcome they would be W, H, WSH, N and WSHN) as well as the control arm. The mediators included were chosen based on a review of the literature on modifiable risk factors for the outcomes, data availability, and an examination of the early life exposures measured at the 2-year intervention that are impacted by the intervention. A total of 3,629 children have data from baseline as well as both the 2- and 7-year follow-up. At 2 years of age the potential mediators impacted by the intervention for any intervention arm included indicators of child health (child 7-day diarrhea prevalence, child growth), caregiver depressive symptoms, and quality and quantity of stimulation in the home environment. A preliminary analysis demonstrates that potential mediators at 2 years are associated with outcomes at 7 years of age: stimulation at 2 years is associated with 7-year outcomes for social-emotional development, and diarrhea, home stimulation, and caregiver depressive symptoms at 2 years are associated with 7-year full-scale IQ outcome. Results from the mediation analyses will be presented and discussed. This work builds on evidence on key intervention components that contribute to sustained intervention effects in childhood.

Paper #4
Pathways of Change in Early Childhood Development: Evidence from a parenting intervention in rural Kenya
Author information	Role
Douglas Newball-Ramirez, University of Southern California, United States	Presenting author
Helen O. Pitchik, UC Berkeley School of Public Health, United States	Non-presenting author
Italo Lopez Garcia, University of Southern California, United States	Non-presenting author
Pamela Jervis, University of Chile, Chile	Non-presenting author
Abstract
Background: An estimated 43% of children under age 5 in low- and middle-income countries (LMICs) do not reach their full developmental potential due to poverty, poor nutrition, and inadequate psychosocial stimulation. Early childhood development (ECD) programs that promote responsive play, communication, and nutrition can improve children's development. However, the mechanisms through which these programs work are not well understood. While some programs show early success, their benefits often diminish over time, suggesting challenges in maintaining new parenting behaviors. It is therefore critical to understand how ECD programs can sustainably change behaviors and improve child outcomes. Methods: This study builds on an evaluation of two group-based delivery models of a responsive parenting curriculum in rural western Kenya. The curriculum was delivered in 16 bi-weekly sessions across 60 villages, involving 1,200 households with children aged 6-24 months. At the 10-month endline, the interventions significantly improved children's cognitive (0.42 SD), receptive language (0.31 SD), and socioemotional development (0.2 SD). They also improved maternal parenting behaviors (0.8 SD), knowledge of child development (0.2 SD), and nutrition practices (0.2 SD), though there were no effects on maternal mental health, perceived social support, or child-rearing beliefs. To explore the causal pathways, we used baseline and endline data to estimate dynamic production functions for child development. First, we constructed a measurement system linking child and maternal latent factors (e.g., cognitive, socioemotional traits, stimulation behaviors, knowledge, and nutrition) to individual measures, addressing measurement error. Next, we estimated production functions for children's cognitive and socioemotional development at endline, with maternal stimulation, knowledge, and nutrition inputs as key variables, controlling for baseline child cognition and maternal traits. To correct for potential bias due to the endogeneity of maternal inputs, we applied an instrumental variable (IV) approach. Results: Initial results show that, without correcting for the endogeneity of maternal inputs, stimulation at endline, and to a lesser extent knowledge, were strongly associated with children's cognitive and socioemotional development. The elasticity of stimulation was particularly high for cognition (0.8), indicating a significant indirect effect. The direct effect of the intervention on child outcomes was smaller but positive, suggesting the presence of unmeasured mediators. After correcting for endogeneity using IVs, the causal effect of stimulation on cognition nearly doubled (elasticity of 1.7), while the effect on socioemotional development remained consistent. However, our IV results are too imprecise to identify the causal effect of other maternal inputs on outcomes and intervention’s direct effects. While our initial results suggest that maternal stimulation is the primary mediator of change in child outcomes, ongoing work is focused on testing new IVs with better statistical power to help address causality. Conclusions: Our initial findings suggest that maternal stimulation explain nearly all of the intervention's effects on child development. Maternal knowledge played a minor role. No other maternal inputs were found to mediate the intervention’s impacts on parenting behaviors or child development. Future research should focus on uncovering the mediators of parenting behavior changes, exploring mediators such as maternal social networks, cognitive load, and more sensitive measures of maternal wellbeing.

⇦ Back to session