Logistic Regression Additive Model: Application to Tanzania Demographic and Health Survey Data

Abstract

Background: The well-being of a child reflects household, community and national involvement on family health. Currently, the global under-five child mortality rate is falling faster compared to any time in the past two decades. However, the progress remained insufficient to match the Millennium Development Goal 4 targets especially in the Sub-Saharan African region. Objective: This study aims to visualize and identify factors associated with under-five child mortality in Tanzania, which is essential for formulating appropriate health program and policies. Methods: The survey data used for this paper was taken from 2011-2012 Tanzania HIV/AIDS and Malaria Indictor Survey. The study utilizes statistical model that accommodate a response, which is dichotomous and account for non-linear relationship between binary response and independent variable. Generalized additive models was adopted for the analysis. The sample was selected using stratified, two-stage cluster sampling that gave a sample size of 10494 mothers. The model was fitted using proc gam in statistical analysis software version 9.3. Results: The results showed that human immunodeficiency virus status of the mother and breastfeeding were associated with under-five child mortality. Furthermore, the results also indicated that under-five child mortality had a quadratic pattern relationship with the number of children ever born, the number of children alive, the number of children five or under in a household and child birth order number. Conclusion: Based on the study, our findings confirmed that under-five mortality is a serious problem in the Tanzania. Therefore, there is a need to intensify child health interventions to reduce the under-five mortality rate even further with the development of policies and programs to reduce under-five child mortality.