Geospatial patterns and multilevel determinants of cesarean section rates in Iran: a Bayesian spatial analysis using INLA

Background Cesarean section (CS) rates in Iran remain among the highest globally, far exceeding the WHO’s recommended 10–15%. While individual determinants are well-studied, provincial spatial heterogeneity and contextual influences are underexplored. This study quantifies multilevel predictors and examines spatial patterns of CS using Bayesian spatial modeling. Methods Cross-sectional analysis of 24,982 deliveries from the 2010 Iran Demographic and Health Survey (IrMIDHS). Outcome: mode of delivery (cesarean vs. vaginal). Predictors: maternal age, education, economic status, contraceptive use, smoking. Multilevel logistic regression with provincial random intercepts was fitted first, followed by a Bayesian spatial logistic model using INLA with BYM2 structure. Model fit compared via DIC/WAIC; spatial autocorrelation via Moran’s I. Survey weights were not applied due to the focus on spatial patterns rather than population prevalence. Results National CS rate: 52.1%. Advanced age (OR = 1.036/year, 95% CrI: 1.028–1.045), university education (OR = 1.584, 95% CrI: 1.422–1.762), and high income (OR = 1.142, 95% CrI: 1.045–1.247) increased odds; contraceptive use was protective (OR = 0.458, 95% CrI: 0.418–0.501). Moderate clustering (ICC = 0.18). Spatial model showed superior fit (ΔDIC = 670; ΔWAIC = 670) and significant autocorrelation (Moran’s I = 0.42, p < 0.001). Northern/central provinces exhibited excess adjusted risk; southeastern regions showed reduced risk. Conclusions CS in Iran is associated with individual socioeconomic factors and substantial spatial heterogeneity. Bayesian spatial modeling with INLA/BYM2 identifies high-risk clusters, supporting geographically targeted interventions to reduce unnecessary CS and promote equitable obstetric care nationwide.