Postnatal care practices and linear growth among children aged 6–23 months in Central Kalimantan, Indonesia: A structural equation modeling approach

Background Childhood stunting affects an estimated 148.1 million children worldwide. In Indonesia, prevalence was 21.5% in 2023, and Central Kalimantan, whose riverine geography limits repeated access to health services, safe water, and sanitation, was higher (23.5%). Prior work has largely pooled children aged 0–59 months and favoured risk prediction over explanation. Objective To estimate the direct and indirect associations of sociodemographic, water-sanitation-hygiene (WASH), prenatal, natal, and postnatal domains with linear growth (continuous height-for-age z-score, HAZ) among children aged 6–23 months in Central Kalimantan; to identify the domains most proximally and strongly associated with growth; and to derive a district policy roadmap. Methods Using a cross-sectional subset of the 2023 Indonesian Health Survey (n = 431), we estimated a partial least squares structural equation model (SmartPLS 4.1.1.2) operationalising the WHO Childhood Stunting Conceptual Framework, with five formative domains and continuous HAZ as the outcome. Inference used bootstrapping (10,000 subsamples; bias-corrected and accelerated [BCa] 95% confidence intervals), and prediction used PLSpredict. A sensitivity analysis removed child weight, measured concurrently with HAZ, from the postnatal composite. Survey weights, clustering, and stratification were not incorporated. Results In the main model, the postnatal domain had the largest direct association with HAZ (β = 0.470, p < 0.001; f 2 = 0.284), followed by prenatal conditions (β = 0.122); natal conditions were not directly associated. Sociodemographic and WASH domains, specified as indirect, were associated with growth through prenatal (β = 0.068) and postnatal (β = 0.083) pathways. The model explained 25.1% of HAZ variance; out-of-sample prediction was positive but marginal (Q 2 predict = 0.009). Without child weight, the postnatal coefficient fell from 0.470 to 0.149 (95% BCa interval 0.045–0.236; f 2 = 0.022), comparable to the prenatal coefficient (0.164); explained HAZ variance fell from 25.1% to 5.9%, and the WASH pathway was no longer supported. Conclusion Postnatal feeding practices remained positively associated with linear growth without concurrent child weight but were not the dominant correlate; most of the original association reflected that indicator. Antenatal and postnatal care emerge as complementary district priorities. Given the cross-sectional, unweighted design, findings denote associations within the analysed sample, not causal or population-level effects.