Quantifying Inequality, Volatility, and Stability in Early Postnatal Care Coverage by Multilevel Beta Regression

Abstract

The early postnatal care in the first 48 hours of childbirth is crucial in preventing maternal morbidity and death. However, worldwide evidence remains dominated by mean-based and binary analyses that do not account for instability in service coverage. This research aims to measure the extent and consistency of early postnatal care coverage across nations and socioeconomic status. The three objectives include estimating changes in coverage over time, identifying socioeconomic and geographic disparities, and understanding the factors driving fluctuations in coverage. A multilevel beta regression was used with harmonised multi-country data from 73 countries covering 2010-2023, using 230 country-year-stratum observations to model bounded coverage percentages with country-random and survey-source random effects. The volatility was measured using a complementary instability model that relies on logarithmically transformed squared deviance residuals. Findings indicate a significant increase in temporal coverage (β = 0.058, p <.001), with coverage consistently lower in low-income contexts and in the poorest and rural populations (p <.001). The volatility of the poorest and richest groups was much greater than that of the nation groups, suggesting masked subgroup volatility. It is novel in that the coverage and instability of the models are jointly modelled in a distribution-consistent manner, thereby overcoming the significant statistical shortcomings of previous literature. The following recommendations should be considered: Stability metrics should be incorporated into SDG 3, resilience-based interventions should be prioritised for disadvantaged populations, and routine postnatal follow-up systems should be strengthened to ensure quality and consistency in maternal care.