Covariate Effect as an offset in Spatio-temporal Disease Mapping Modelling

Abstract

This paper focuses on extending spatio-temporal model in disease mapping. The objective is to incorporate covariate variable as an offset to analyse rea-life data across state and health care areas. The method of generalized linear mixed model (GLMM) that followed a Poisson distribution and built within a Bayesian approach was used. Model fitting and inference were carried out using integrated nested Laplace approximations (INLA), while deviance information criteria (DIC) were used to choose the best performing model. The performance of the proposed model compared to existing model was assessed using female breast cancer mortality data. Result revealed that the proposed model has the lowest DIC value of 1514.053 as against DIC value of 1514.407 for the existing model. The proposed model overcomes the usual model as measured by DIC for a difference of 0.354 (35%) which showed a significant improvement in the proposed model, The proposed model has yielded a classical methodology for assessing the risk variations in the state areas and their influence in health care areas on disease outcome. The study recommended that covariate variable such as age should be included as an offset in a spatio-temporal disease mapping modelling.