Ranking Predictors of Child Dietary Diversity in Nepal Using a Decision Tree

Abstract

Introduction: Minimum dietary diversity among children (MDD-C) remains low in Nepal, contributing to persistent malnutrition. This study aimed to rank and characterize the association of dietary, socioeconomic, and health-related predictors of MDD-C among children aged 6–59 months in Madhyapur Thimi Municipality of Nepal.

Method: A secondary analysis of survey data from 375 children was conducted. Chi-squared tests were used to identify baseline associations of the predictors with MDD-C. A Chi-squared Automatic Interaction Detection (CHAID) decision tree was then fitted to identify key split variables and interaction structures. A CHAID-based sensitivity analysis was used to estimate food groups’ contributions to predicted dietary adequacy by toggling individual food groups from absent to present. Bootstrap resampling was used to quantify the internal variability of these estimates.

Result: Overall, 73.6% of children were predicted to achieve MDD-C. The decision tree placed other fruits and vegetables (Group G) at the root level, followed by splits on eggs (Group E), legumes and nuts (Group B), and vitamin A–rich foods (Group F). Sensitivity analysis suggested that enabling consumption of Group G or Group E was associated with the largest expected improvements in predicted dietary adequacy, and bootstrap resampling indicated that this ranking was relatively stable across resamples.

Conclusion: This model-based analysis indicates that Group G and Group E foods are strong positive predictors of adequate dietary diversity and offer the largest expected gains among diet-deficient children. Hence, access to Group G and Group E foods could be prioritized as dietary interventions to improve child nutrition.