<![CDATA[Background: Stunting remains a critical public health emergency in Indonesia, with national prevalence of 21.6% in 2022 still exceeding the WHO 20% emergency threshold; West Sulawesi province recorded the third-highest provincial prevalence nationally (35.0%), with Polewali Mandar Regency identified as a principal contributing district. Severe stunting carries substantially greater risk of child mortality and cognitive impairment than moderate stunting, yet stunting severity remains underanalyzed in district-level policy, as most studies rely on binary stunted/not-stunted classification. Objective: This study examines socioeconomic determinants of stunting severity among children under two in rural Polewali Mandar. Methods: A cross-sectional study was conducted among 165 children under two years of age in rural households of Mapilli sub-district, Polewali Mandar Regency, enrolled through consecutive sampling from an estimated 580 eligible children (28.4% coverage). Stunting severity, based on height-for-age z-scores, was categorized into three ordinal levels (not stunted, moderate stunting, severe stunting). Six predictors household economic status, maternal education, latrine ownership, drinking water source, posyandu visit frequency, and health facility access were assessed via structured interviews. Ordinal logistic regression under the Proportional Odds Model (POM) was applied, with the parallel regression assumption confirmed via the Brant test (omnibus X² = 14.07, df = 11, p = 0.23). Results: Stunting prevalence reached 46.67% (32.12% moderate, 14.55% severe), more than twice the WHO emergency threshold. Four significant predictors were identified: household economic status (moderate vs low, OR = 2.194, 95% CI 1.063–4.605, p = 0.035; high vs low, OR = 5.072, 95% CI 1.628–17.679, p = 0.007), ineligible drinking water source (OR = 0.180, 95% CI 0.038–0.744, p = 0.021), irregular posyandu attendance (OR = 0.352, 95% CI 0.140–0.877, p = 0.024), and difficult health facility access (OR = 0.048, 95% CI 0.009–0.208, p < 0.001). The model explained a moderate share of variance (Nagelkerke R² = 0.321).Conclusions: Household economic status, drinking water source, posyandu attendance, and health facility access were significantly associated with stunting severity in this rural setting. Given the cross-sectional design, these findings reflect associations rather than causal relationships. Severity-graded analysis provides more operationally relevant evidence than binary approaches for targeting rural interventions, pending confirmation from longitudinal research.]]>
