The Latent Classes and Predictors of Early Childhood Development in Low- and Middle-Income Countries. Poster presented at the biennial meeting of the Society for Research in Child Development Conference, Salt Lake City, USA, March 23 - 25, 2023

Abstract

<p>Investing early childhood development (ECD) are foundational for individual and societal development. Extant research has tended to use a variable-centered approach to examine predictors of ECD and few studies have adopted a person-centered approach. Variable-centered research has shown that family wealth, maternal education and parenting practices predict child outcomes overall. However, less is known if different patterns of child development exist and how these predictors may be associated with ECD patterns, which is critical to designing and delivering tailored ECD interventions. This study used latent class analysis (LCA) with data from 29 low- and middle-income countries (LMIC) to examine the latent patterns/classes of ECD from 2010 to 2020 and their associations with predictors at different ecological levels (individual, family, and country).<br><br>We used publicly available data from UNICEF’s Multiple Indicators Cluster Surveys (MICS) waves 4 to 6, the nationally representative household surveys conducted from 2010 to 2020 in LMIC countries. Although 68 countries participated in at least one of the three waves, we used data from 29 countries for which at least two waves of data were available for 3-5-year-olds to make the longitudinal comparisons meaningful. Primary caregivers reported children’s literacy-numeracy, physical, social emotion, and approaches to learning development with the Early Childhood Development Index (ECDI) (10 items, Wave 4: α =.57, Wave 5: α =.54, Wave 6: α =.51). They also reported their child’s preschool attendance, their partners’ and their own engagement in children’s learning, the numbers of toys and books at home, and other demographic information. The Human Development Index (HDI) of the countries at the year of data collection were also noted. The total sample size for analyses was 226,374 (nMICS4 = 70,082, nMICS5 = 91,652, nMICS6 = 64,640; Mage = 47.23(months), SDage = 6.87).<br><br>LCA was conducted separately for each wave of data using the 10 ECDI items as indicators. Models with two to five classes were evaluated. Three classes were chosen due to its best fit and substantive meaning: the Overall Academically Challenged group (14.6% in MICS 4, 12.6% in MICS 5, and 11.1% in MICS 6), the Academically Challenged but Approaches to Learning Competent group (63.5% in MICS 4, 60.9% in MICS 5, and 66.8% in MICS 6), and the Overall Competent group (22.0% in MICS 4, 26.5% in MICS 5, and 22.2% in MICS 6). The percentages of the each of the three classes were distributed consistently across the three waves, Mixed-effects random intercept logistic regression models with predictors at different ecological levels (i.e., child age, gender, maternal education, urbanicity residence, maternal engagement, paternal engagement, preschool attendance, number of books, number of toys, family wealth, and HDI) were further conducted with each wave of data. Preschool attendance was found to significantly predict the group membership with an acceptable effect size in MICS 4 (Ra = .15) and 5 (Ra = .13). Number of books and maternal education significantly predicted the ECD class membership in MICS 4 with an acceptable effect size (Rsa = .11). The implications for future research and policy-making are discussed.<br></p>