Geoadditive Latent Variable Models of Childhood Diseases: Case Study for Nigeria
Khaled Khatab, Ngianga-Bakwin Kandala
Institute of Occupational and Social Medicine, Medical Faculty, RWTH Aachen University, Aachen, Aachen, Germany; Clinical Sciences Research Institute, Warwick, Medical School, University of Warwick, Warwick, United Kingdom

Background: Childhood morbidity and malnutrition are among the most serious health issues facing developing countries. Previous analyses are often based on Demographic and Health Surveys (DHS) as a well established data source with reliable information on childhood diseases and undernutrition, and they rely on statistical inference with various forms of regression models.

Aim: This study investigates the impact of various bio-demographic and socio-economic variables on childhood disease with flexible geaodditive probit models.

Methods: We apply a recently developed geoadditive latent variable models where the three observable disease (diarrheoa, cough, fever) variables are taken as indicators for the latent individual variable “health status” or “health status” of a child. This modelling approach allows to study the common influence of risk factors on individual frailties of children, thereby automatically accounting for association between diseases as indicators for health status. The latent variable model has been applied in this work to analyze the impact of risk factors and the spatial effects on the unobservable varaible “health status” of a child less than 5 years age using the 2003 DHS data for Nigeria.

Results: The analysis suggests some strong underlying spatial patterns of the three ailments with a clear southeastern divide of childhood morbidities and this might result of shared and overlapping the various risk factors.

Conclustion: The spatial effect suggests the need to give more attention to some areas that have high rates of diseases, such as the south eastern regions as well as some regions in the north part which are associated with a high rate of diseases. On other hand, latent variable models offer a new methodology opportunity for considering special types of diseases as indicators for latent morbidity and to flexibly model covariate and spatial effects on this latent variable.

Biography: He is Assistant Professor in Medical Statistics and Statistical Methods in Epidemiology in RWTH Aachen Germany. With a background in Econometrics, Applied Statistics and Epidemiology and public health, Khaled's area of research interest is statistical modelling, especially Bayesian approaches. He is also Chartered Statistician-RSS-LONDON.