Spatial Patterns and Healthcare Access in Early-Onset Breast Cancer Diagnosis in Lagos, Nigeria: A Bayesian Multilevel Analysis

Abstract

Early-onset breast cancer (diagnosis before age 40) is an emerging public health concern in Nigeria,
yet its spatial distribution and determinants remain poorly understood, particularly in urban settings
with unequal access to diagnostic services. This study analysed retrospective breast cancer registry
data from the Lagos University Teaching Hospital (LUTH), with early-onset diagnosis defined as a
binary outcome. A Bayesian structured additive logistic regression model was used to assess socio-
demographic, clinical, and spatial effects. Nonlinear effects of age and distance to LUTH were
modelled using smooth functions, while spatial heterogeneity across Local Government Areas
(LGAs) was captured using a Gaussian Markov Random Field. A three-level hierarchical structure
accounted for clustering, and estimation was performed using Integrated Nested Laplace
Approximation (INLA). The unadjusted model showed geographic variation, but this was
substantially attenuated after adjustment, with no strong evidence of elevated risk across LGAs.
Distance to LUTH showed a decreasing nonlinear relationship with early-onset diagnosis, indicating
higher probabilities among individuals living closer to the facility. Age also exhibited a declining
nonlinear association. Overall, spatial variation is modest after adjustment and largely reflects
diagnostic access rather than true geographic clustering.