Comparing performance of generalized linear mixed model and generalized estimating equations in modelling under-five mortality : a case study of Iganga-Mayuge Districts Uganda

Abstract

Introduction: Reduction of childhood mortality is a global health priority and an indicator of child survival. Over 5 million in 2021 children under five years died every day from preventable and curable diseases which is undesirable. Between 1990 and 2020, the global mortality rate for children under five years declined by 61 per cent, from 93 deaths per 1,000 live births to 37 deaths per 1,000 live births. Over the last three decades, under-five mortality has steadily been declining in Uganda. Between 1990 and 2020, the rate declined by more than 70%, from 183 to 43.3 deaths per 1,000 live births. This study aimed at determining the association between the predictor variables and under-five mortality rates in the Iganga and Mayuge districts. Methods: Data from Iganga-Mayuge Health and Demographic Surveillance Site (IMHDSS) of 2010 to 2015 collected on women of reproductive age (13 to 49 years) were used. The outcome was death before the fifth birthday and the independent variables were gender, place of residence, birth order, mother’s age, education, wealth index, marital status, place of delivery, attend antenatal care, occupation and year. Village was used as a random effect. For descriptive analysis, proportions were used. Means and standard deviation were used for continuous variables. The explanatory variables were checked for multicollinearity to ensure validity for further analysis. AIC and QIC were used for model selection. Data management and analysis was done using excel and STATA 18. Results: Between 2010 and 2015, 2011 had the highest number of under-five deaths. Mothers from both the rural and Peri_urban areas had an average of one child. The odds of dying before the age of five reduced among the children whose mothers were cohabiting and those whose highest level of education was below primary. Being male, residing in the rural areas and children whose mothers were teenagers increased the odds of dying before the age of five. Conclusion: The choice between these methods should be guided by the underlying data structure, assumptions, research objectives, and the practical significance of the differences observed. Interventions like healthcare access could improve maternal and child healthcare services, enhance education, thus improving future research, guide policy development aimed at improving overall child well-being and reduce under-five mortality in Iganga and Mayuge districts.