Caregiver parenting practices, dietary diversity knowledge, and association with early childhood development outcomes among children aged 18-29 months in Zanzibar, Tanzania: a cross-sectional survey

Background: Many children in low- and middle-income countries fail to reach their cognitive potential, with experiences before age 3 critical in shaping long-term development. Zanzibar’s Jamii ni Afya program is the first national, digitally enabled community health volunteer (CHV) program promoting early childhood development (ECD) following the Nurturing Care Framework within an integrated maternal and child healthcare package. Using program baseline data, we explored home environment, caregivers’ parenting, health and nutrition knowledge and practices, and ECD outcomes in Zanzibar. Methods: We conducted a national household survey among 499 children aged 18-29 months using two-stage cluster sampling in February 2019. The primary outcome was child development score measured using the Caregiver Reported Early Developmental Index (CREDI), with higher scores representing higher levels of child development. We analyzed CREDI scores, along with MICS questions on parenting knowledge, practices, and characteristics of the home environment. We developed multivariate regression models to assess associations between caregiver-child interactions, knowledge of dietary diversity, and ECD. Results: Ten percent of children had overall CREDI z-scores 2 standard deviations [SD] or more below the global reference population mean, with 28% of children at risk of developmental delay with z-scores 1 SD or more below the mean. Cognitive and language domains were of highest concern (10.2 and 12.7% with z-score < − 2 SD). In 3-day recall, 75% of children engaged in ≥4 early stimulating activities with all caregivers averaging 3 total hours of play. CREDI scores were positively associated with greater frequency of caregivers’ engagement (β = 0.036, p = 0.002, 95%CI = [0.014, 0.058]), and dietary diversity knowledge (β = 0.564, p < 0.001, 95%CI = [0.281, 0.846]). Conclusions: Our findings demonstrate a positive association between both the frequency of caregiver child interactions and knowledge of adequate dietary diversity, and ECD outcomes. This aligns with global evidence that promoting early stimulation, play and learning opportunities, and dietary diversity can improve developmental outcomes. Further study is needed to establish causal relationships and assess the impact of ECD programming in Zanzibar. In 2006, Zanzibar expanded access to early childhood education with a mandate for preschool education for all, though by 2015 still less than half of Zanzibar’s children had access to preschool [28]. Despite these initiatives to increase access to early learning, programs focusing on development during the critical window of the first 3 years of life remained unaddressed. In response, the Zanzibar Ministry of Health (MOH) and Ministry of Local Government (President’s Office Regional Administration and Local Government and Special Departments [PORALGSD]) launched the National Community Health Strategy 2019-2025, which formalized a cadre of over 2000 community health volunteers (CHVs) as part of the national health system with the creation of the Jamii ni Afya (‘Communities are Health’) program. The program aims to improve early childhood developmental outcomes by targeting promotion of, access to, and utilization of health care services, improved nutrition, and nurturing caregiver interactions for children in utero to age 5 and is built around WHO/UNICEF’s Nurturing Care Framework [13] and supported by a digital system co-developed with D-tree International. We conducted a cross-sectional, nationally-representative household survey in February 2019 in all 11 districts of Zanzibar, Tanzania. We used two-stage cluster sampling to randomly select 50 clusters using probability proportionate to size. Clusters and respective populations were defined by the enumeration areas provided by the Office of the Chief Government Statistician. We then used systematic random sampling within each cluster to identify and recruit 10 child-caregiver pairs from each cluster, for a total sample size of 500 participant pairs. Eligible participant pairs included children aged 18 to 29 months, with birth dates verified from their child health cards, and the child’s primary caregiver. Both caregiver and child were required to have their primary residence in Zanzibar. No other exclusion criteria were applied. Therefore, the sample is representative of the whole of Zanzibar, with all eligible children being equally likely to be selected for participation. Notably, this cross-sectional study takes place within a larger study to compare Caregiver Reported Early Developmental Index long-form (CREDI) scores at baseline to future survey implementations planned in 2023 and beyond. The power and sample size considerations for this larger study are provided in the supplemental materials (See Supplement 1). Household interviews with primary caregivers of eligible, enrolled children were conducted by trained data collectors in February 2019. The study questionnaire was approximately 40 min in length and administered in Kiswahili. The tool was previously translated and tested in Kiswahili in Tanzania by the tool’s developers. Data were collected on tablets and smartphones using ODK Collect. Data collectors and field supervisors were independent to the Ministry of Health and collaborating organizations, and participated in a 5-day training on ethical considerations, survey methodology, and administration of all data collection tools prior to data collection. The primary outcome measure was child development score as measured by the CREDI tool, which is based on caregiver report of easily observable and understandable child milestones and behaviors by age group. The tool has been validated in 17 low, middle- and high-income countries, including Tanzania [29, 30]. We report z-scores and raw scaled scores by language, cognitive, motor, and social-emotional domains, as z-scores are best used for comparison to other populations, and scaled scores are most appropriate for linear regression. For both measures, higher scores represent greater achievement in child development outcomes. Although the CREDI was not developed as a tool for diagnosing delay in individual children, we established a cut-off to define “developmental concern” for the purpose of comparison to the reference population and ease of communicating for policy and advocacy. We follow the conventions of other developmental assessment tools [31, 32] to consider to a z-score 1- < 2 standard deviations below the mean a “developmental concern” and “significant developmental concern” to be equal to a z-score 2 or more standard deviations below the mean. Because this is a standard normal distribution, we would expect 13.5% of children be in the “developmental concern” group. Likewise, we would expect 2.5% of children to be in the “significant developmental concern”. In accordance with CREDI guidelines, CREDI z-scores were used for descriptive analysis, while the continuous score was used for all hypothesis testing [33]. Our primary exposure variables of interest were caregiver reports of interactions with the child in the form of early stimulating activities and caregiver knowledge of dietary diversity. Questions were drawn from the UNICEF Multiple Indicator Cluster Survey (MICS) questionnaire. Caregiver report of interactions with the child was analyzed as a continuous variable defined as the summative number of early stimulating activities a child engaged in with any caregiver in the past 3 days. Types of early stimulating activities included: reading or looking at picture books, telling stories, singing songs, taking outside the home compound, playing a game, or naming/counting/drawing together. Knowledge of dietary diversity was the number of food groups the caregiver reported as appropriate for the child to eat, and knowledge of feeding frequency was the number times per day the caregiver reported the child should be fed. For the diet questions, we modified the MICS questions to reflect knowledge rather than practice, as the later was not feasible for implementation. The variables were categorized as those who named four or more food groups compared to those who named less, and those who named feeding thrice daily or more compared to those who suggested less. Knowledge of feeding practices were only assessed in a subgroup of children who were still breastfed and 18-23 months of age (n = 122). We also gathered data on individual and household-level covariates related to the home environment, caregiver engagement and play, disciplinary practices, care-seeking behaviors, and health and nutritional knowledge and practices. All questions and indicators were defined and assessed using standard UNICEF indicators from MICS and a standardized monitoring and evaluation tool for UNICEF’s Care for Child Development checklist. We measured sociodemographic characteristics relevant to understanding the relationship between our independent exposures of interest and child development outcomes. Wealth was measured using the Tanzania EquityTool (https://www.equitytool.org/) which is a validated tool that analyzes household wealth using a simplified version of the DHS asset-based questionnaire. Using the EquityTool standard analysis package, each household was assigned a score and then categorized according to their relative wealth compared to quintile levels established by the Tanzania DHS 2015 population. We completed descriptive analyses of the overall and domain-specific CREDI z-scores and compared to the CREDI reference population1, using the CREDI scoring package developed in R V3.6.0 (R Core Team, Vienna, Austria). To explore associations between the individual-level covariates and overall and domain-specific CREDI continuous scores (herein: child development outcomes), we performed bivariate analyses for all categorical variables using Wald’s t-test and ANOVA. We fit two multivariate linear regressions to quantify the relationship between number of early stimulating activities and child development outcomes, and knowledge of adequate dietary diversity and child development outcomes. In both models, we adjusted for known confounding variables and those found to have a significant association in the bivariate analysis (at α = 0.05 significance level) including: geographic region, age of the caregiver, age and sex of the child, if caregiver is married or living with their partner or not, maternal and paternal education levels, parity, wealth, if the child was left alone for more than an hour in the past week, and (for caregiver engagement only) if the caregiver believed that domestic abuse was justifiable in any situation. We accounted for clustering by utilizing the svyset function on Stata. Our primary sampling unit was the enumeration area, and each individual was weighted by the probability of selection within their cluster. All tables and regression analyses account for the survey sampling plan our standard errors were adjusted accordingly. Given the difficulty to interpret meaningful changes in the raw scaled CREDI score, we standardized the results of our multivariate model analysis. To do so, we divided the coefficient of the CREDI outcome variable estimated by the model by the standard deviation within the study population for the specific CREDI domain, to express the effect size as change in standard deviation among the study population. With the exception of the CREDI scoring, all statistical analyses were performed using Stata Version 14 (StataCorp, College Station, TX). Ethical approval to conduct this study was obtained from the institutional review boards at the Ministry of Health/Zanzibar Health Research Institute (Ref. No: ZAHREC/01/DEC/2018), and Boston Children’s Hospital (Ref. No.: P00029981). Every child’s parent or primary caregiver provided written informed consent on behalf of the child-caregiver pair prior to enrollment in the study, and all research was performed in accordance with approved study procedures and ethical guidelines.