Determinants of institutional delivery in Afghanistan: evidence from the 2022/2023 Multiple Indicator Cluster Survey (MICS)

Afghanistan has one of the highest maternal mortality ratios in the world, and the place where a woman gives birth is a critical determinant of maternal and newborn survival. Although the determinants of institutional delivery have been widely studied, nationally representative evidence from the most recent post-2021 data covering both urban and rural women remains scarce. This study examined the determinants of institutional delivery in Afghanistan.

We conducted a secondary analysis of the Afghanistan Multiple Indicator Cluster Survey (MICS) 2022/2023. The descriptive analysis included 12,578 women aged 15-49 years who had a live birth in the two years preceding the survey. The outcome was institutional (health-facility) versus home delivery. Descriptive and bivariate analyses used survey-weighted methods accounting for the multistage cluster design, stratification, and sampling weights. Multivariable associations were estimated using three-level mixed-effects logistic regression, with women nested within households, which were nested within primary sampling units (PSUs). A common complete-case sample of 12,462 women was used for four models: Model 0 (null), Model I (individual-level factors), Model II (household- and community-level factors), and Model III (full model). PSU-level contextual measures of community educational attainment and community socioeconomic status were derived from within-PSU aggregates and modeled alongside urban/rural residence. Adjusted odds ratios (AORs), 95% confidence intervals (CIs), random-effects variances, intraclass correlations (ICCs), median odds ratios (MORs), proportional change in PSU variance (PCV), and model-fit statistics were reported. A two-level PSU-only model was fitted as a sensitivity analysis.

Overall, 66.3% of women delivered in a health facility (95% CI 64.5-68.1). In the fully adjusted three-level model, higher education was associated with higher odds of institutional delivery compared with pre-primary/ECE or no education (AOR 7.92; 95% CI 2.40-26.11), while women with five or more children had lower odds than those with 0-1 child (AOR 0.19; 95% CI 0.12-0.31). Antenatal care was strongly associated with institutional delivery (AOR 7.12; 95% CI 5.32-9.53), as were urban residence (AOR 2.46; 95% CI 1.61-3.78) and residence in communities with high socioeconomic status (AOR 6.90; 95% CI 4.23-11.24 versus low socioeconomic status). The PSU-level variance declined from 7.44 in the null model to 2.32 in the full model (PCV: 68.8%), although residual PSU clustering persisted (ICC: 19.7%; MOR: 4.25). The full three-level model had the lowest AIC and BIC. In the two-level sensitivity analysis, the directions of the principal associations were generally consistent, although several effect estimates were attenuated.

Institutional delivery in Afghanistan remains inequitably patterned by factors operating at individual, household, and community levels. Women's education, household wealth, community socioeconomic context, urban-rural residence, parity, and antenatal care were independently associated with institutional delivery. Equity-oriented strategies should address educational, financial, and geographic barriers while recognizing that substantial contextual heterogeneity remains after adjustment. Given the cross-sectional design, findings are interpreted as associations rather than causal effects.