Subnational regional dispersion in zero-dose vaccination: a multidimensional comparison across 339 country–year surveys

Zero-dose children are disproportionately concentrated in disadvantaged populations, but the dimension along which inequality is greatest may differ across settings. Comparisons across household wealth, maternal education, place of residence, and subnational region are complicated by differences in the number and population size of their constituent groups. The study compared these four stratifiers using a common descriptive measure and assessed uncertainty in identifying the leading observed inequality dimension. A total of 339 eligible country–year surveys from 86 countries conducted during 1993–2023 were analysed using data from the WHO Health Inequality Data Repository. For each stratifier, inequality was measured using the population-weighted between-group standard deviation of subgroup-specific zero-dose prevalence. The stratifier with the largest value was classified as the leading observed inequality dimension. Classification uncertainty was assessed using 5,000 parametric simulations based on reported subgroup standard errors. A maximum simulated leading-dimension probability below 0.80 was considered indicative of an uncertain classification. Sensitivity analyses examined alternative inequality measures, probability and dominance-margin thresholds, country weighting, regional group count, survey period, and additional inequality dimensions. Subnational region was the leading observed inequality dimension in 209 of the 339 surveys (61.7%), followed by household wealth in 67 (19.8%), maternal education in 61 (18.0%), and place of residence in 2 (0.6%). However, only 145 surveys (42.8%) had a maximum simulated leading-dimension probability of at least 0.80. Across six alternative inequality measures, the proportion of region-led surveys ranged from 50.1% to 70.5%. The regional share remained similar when countries were weighted equally (62.5%; 95% CI, 54.9–70.1) and when only the most recent survey from each country was retained (68.6%; 95% CI, 59.3–77.9). Among 66 countries with repeated surveys, 37 (56.1%) changed their leading dimension at least once. Regional dispersion was positively associated with the number of reported subnational regions (Spearman’s ρ=0.355; p < 0.001). Adding child sex and, where available, maternal age did not change the original classifications. Subnational regional dispersion was most frequently the largest observed dimension of zero-dose inequality, but this pattern was neither universal nor consistently certain or stable over time. Subnational region should be interpreted as a composite spatial stratifier rather than an independent causal mechanism. The findings support repeated multidimensional assessment to identify where disadvantage is concentrated, while emphasising that the numerically largest dimension does not necessarily represent the most appropriate intervention target.