Macro-micro dynamics of child undernutrition in Pakistan: a structural learning-based Bayesian network analysis of...

Child undernutrition remains a major public health concern and creates health-related inequalities in developing countries, with long-term consequences for human development deficits. Most of these early studies were conducted using conventional econometric techniques; however, understanding the complex, multi-level determinants of undernutrition may require more advanced analytical methods.

This study, PC algorithm is used for Structural Learning-Based Bayesian Networks [SL-BN] (a graphical model based on directed acyclic graph with joint and conditional associations)-which is conceptually different than the weight-based networks, to examine the macro- and micro-level determinants of child undernutrition in Pakistan. Macro-level analysis combines historical nutrition policy evaluation (critical discourse) with WDI and UNDP data (2000-2022), spanning the MDG and SDG eras. Micro-level analysis draws on the Pakistan Demographic and Health Survey (PDHS 2017-18) to identify socioeconomic drivers of child malnutrition and nutritional inequalities.

In micro analysis, the SL-BN network structure identifies household-level determinants-such as poor wealth status, rural residence, poor environmental conditions, low maternal education, and diarrhea-as key drivers of child undernutrition. Inequality is stark, with malnutrition disproportionately concentrated among the poorest. However, the SL-BN findings and critical assessment of nutritional policies and programs found that low investment in human development, foreign dependency on health, poverty, inflation, and a history of inconsistent nutritional policies to combat hunger and undernutrition are the fundamental issues at the macro level in Pakistan.

The study concluded that the underlying micro level determinants of child undernutrition in Pakistan cannot overcome without long-term targeted policies at macro level.