Analytical Dataset: Machine Learning-Based Prediction of Acute Morbidity Among Children Aged 6-23 Months in Punjab Pakistan

Dataset: Punjab MICS 2017-18 Analytical DatasetStudy: Machine Learning-Based Prediction of Acute Morbidity Among Children Under Five in Punjab PakistanAuthors: Bakhtawar Majeed, Furqan Awan et al. VARIABLES:- disease_status: Outcome variable (Yes=Diseased, No=Healthy)- wealth_quintile_index: Household wealth (Poorest to Richest)- mothers_education: Maternal education level (None to Higher)- vitamin_mineral_supp: Supplement access (Yes/No)- HAZ_category: Height-for-age z-score category- WHZ_category: Weight-for-height z-score category- Achieve_MDD: Minimum dietary diversity (Adequate/Inadequate)- Breastfeeding: Breastfeeding status (Yes/No)- bottle_feeding: Bottle feeding (Yes/No)- health_insurance: Health insurance coverage (Yes/No) DATA SOURCE: Punjab Multiple Indicator Cluster Survey 2017-18Original data available at: http://mics.unicef.org/surveysAnalytical sample: 10,128 children aged 6-23 months This study demonstrates that Random Forest machine learning combined with counterfactual policy simulation provides a powerful analytical framework for translating population survey data into actionable child health policy guidance. Among the policy-relevant determinants of acute childhood morbidity examined in Punjab, vitamin and mineral supplement access emerged as the highest-impact modifiable factor, with universal supplement access predicted to increase population-level healthy rates by 12.79 percentage points. Maternal education and dietary diversity represent secondary priorities with meaningful but more modest predicted effects. The findings support urgent investment in targeted micronutrient supplementation programmes, girls’ education, and dietary diversity interventions as high-yield strategies for reducing the acute morbidity burden among children in Punjab, especially in divisions with highest burden of illness like D.G. Khan recording the highest prevalence of acute morbidity (66%), Sargodha (56.8%), and Rawalpindi (52.3%) as identified in geographical mapping. This analytical framework is replicable using MICS data from other provinces and countries, offering a scalable tool for evidence-based child health policymaking in similar resource-constrained settings.