From the Brass Indirect Method to IGME: A Critical Review of Under-Five Mortality Estimation Approaches
S.O. Adeyemo *
Federal Polytechnic, Nekede, Owerri, Nigeria.
F.N. Nwobi
Department of Statistics, Imo State University, Owerri, Nigeria.
E.U. Ohaegbulem
Department of Statistics, Imo State University, Owerri, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
Background: Indirect estimation of under-five mortality from Summary Birth Histories (SBH), has been a central tool in low- and middle-income countries since the 1970s. Classical Brass methods map proportions of children dead to mortality probabilities but rely on model life tables, quasi-stability assumptions, and population homogeneity. Modern Bayesian B-spline models adopted by the United Nations Inter-Agency Group for child Mortality Estimation (IGME), produce smooth national trends with uncertainty, but are not designed for routine subnational monitoring using SBH alone. This paper contributes a critical review of indirect U5MR estimation evolution, highlighting limitations in heterogeneous populations, and suggests avenues for adaptation in Sub-Sahara Africa (SSA).
Objective: This paper reviews indirect estimation evolution and identifies five structural limitations in heterogeneous settings: dependence on model life tables, inadequate survey design treatment, limited covariate integration, weak subnational disaggregation, and lack of transparent benchmarking.
Methods: A systematic literature review of indirect estimation methods from Brass (1975) to UN IGME (2024), focusing on assumptions, data requirements, and performance in heterogeneous LMICs. The review draws on studies using Demographic and Health Surveys (DHS) data from Nigeria (2018 & 2023/24), Ghana (2022), Malawi (2024), and Senegal (2023).
Results: Classical methods (Brass, Trussell, Sullivan) rely on rigid multipliers and quasi-stability, failing to account for heterogeneity (e.g., rural-urban gaps, wealth gradients), leading to biased subnational estimates (MAE >15 per 1,000 vs. FBH in Nigeria). Variants like Preston-Coale (1989) and sibling histories (Timæus & Dorrington, 2014) improve recency but still overlook covariates. UN IGME's Bayesian hierarchical model (Alkema & New, 2014) integrates multiple sources for national trends, with B-spline smoothing and bias correction, but lacks covariate adjustment for subnational SBH. Recent DHS applications in Nigeria (calibrated 105 per 1,000), Ghana (43 per 1,000), Malawi (38 per 1,000), and Senegal (37 per 1,000) demonstrate the value of covariate-driven frameworks that preserve subnational ratios (e.g., 2.86 in Nigeria, 1.641 in Ghana, 1.192 in Malawi).
Conclusions: From Brass method to IGME, indirect U5MR estimation has evolved from model-life-table dependence to Bayesian integration, but gaps persist in handling heterogeneity. For SSA, future methods should incorporate survey weighting, covariates, and calibration while preserving subnational ratios. This review underscores the need for adaptable approaches to support SDG 3.2.1 monitoring in diverse populations.
Keywords: Indirect estimation, summary birth histories, under-five mortality, survey design, quasi-stability, heterogeneity, Nigeria, SDG 3.2.1