Analisis Regresi Robust Determinan Sosial Stunting antar Provinsi di Indonesia

Abstract

Stunting remains a social-policy and equity issue because its distribution reflects not only nutritional intake but also family support, access to basic services, and regional inequality. This study examines the association of social interventions and basic services with provincial stunting prevalence in Indonesia using aggregate data from the 2024 Indonesian Nutrition Status Survey (SSGI). The ecological cross-sectional analysis covered 36 provinces with complete published data. Provincial stunting prevalence was modeled using family-assistance coverage by Family Support Teams (TPK), exclusive breastfeeding, access to adequate drinking water, and access to adequate/safe sanitation. Multiple linear regression with HC3 heteroskedasticity-robust standard errors was used, followed by Huber M-estimation as a sensitivity analysis. The model explained 40.4% of provincial variation (R²=0.404). Sanitation coverage showed the most consistent association: each one-percentage-point increase was associated with a 0.427 percentage-point lower stunting prevalence (95% CI -0.817 to -0.037; p=0.033). A 10-point increase therefore corresponded to approximately 4.27 points lower prevalence, other predictors held constant. TPK assistance, exclusive breastfeeding, and drinking-water access had negative but statistically non-significant adjusted coefficients. Huber regression confirmed the sanitation association (b=-0.427; p=0.003). These findings support a multisectoral approach that treats sanitation as a structural social determinant while maintaining nutrition-specific and family-based interventions. Because the unit of analysis is the province, the results indicate ecological association rather than individual-level causation.