Factors Associated With Stunting Among Under-five Year Old Children in Kenya: a Multilevel Logistic Regression Approach

Abstract

Background: Stunting remains a global public health problem affecting millions of children. Stunting refers to growth retardation because of food deprivation or recurrent infections that tend to put a child at higher risk of illness or death. Stunting not only affects the health of individual in childhood but also has long term effects including; elevated risk of nutrition-related chronic diseases, low intellectual quotient, loss of productivity, and consequently poverty. Whereas significant efforts have been made towards reducing stunting in children under five, the rate of reduction remains slow in developing countries. Prevention of stunting in under five-year old children requires a multi-sectoral approach with a clear understanding of how certain factors play in different contexts. Most researchers in the past have studied stunting using the ordinary logistic regression even though most of the data used is complex survey data that may exhibit clustering in observations requiring the use of multilevel logistic regression to account for clustering. Study Objective: To determine factors associated with stunting among under five children in Kenya using the most suitable regression model Methodology: The study was a cross-sectional study design to examine factors associated with stunting among children under- five years in Kenya. The study used historical survey data from Kenya National Bureau of Statistics. The Kenya Demographic and Health Survey 2015 historical data was used. The study will use a total sample size of 18,582 which was collected from across the country using multi-stage sampling technique. Data was collected using household questionnaires and the nutrition status of children was recorded by measuring the height and weight of children 0-59 months. The ordinary logistic and multi-level logistic regression models which takes care of inter-correlation were fitted to select the best model for..........................................................................................................