Determinants of frequency of meal intake among children (9-23 months) in Uganda

Abstract

Frequency of meal intake is a major principle of optimal complementary feeding. Therefore, modelling this indicator that takes on real-valued integers (count); requires attention to the probability distribution of the outcome measure. Also, the use of full information from this measure is often required. Thus, in this study, we modeled frequency of meal intake among children (9-23months) in Uganda as a count while accounting for its distribution as Poisson in examining its factors. Secondary data were sourced from the 2016 Uganda Demographic and Health Survey (UDHS). Descriptive characteristics of the respondents were generated. This was followed by fitting a simple Poisson regression model to assess the association between each covariate and meal frequency intake, and in turn, identify variables for the inferential analysis. At the inferential level, there was a comparison between the Poisson and binary logistic regression models to identify the most suitable model. The study established that the Poisson regression model of frequency of meal intake offered a better precision in the identification of associated factors. Finally, a Poisson model based on case-wise deletion was fitted to examine the determinants of frequency of meal intake. It was found that birth order of the fourth position and above (IRR= 0.915; 95% CI = 0.842 – 0.997) was associated with reduced frequency of meal intake; while being aged from the richest household wealth quintile (IRR= 1.170; 95% CI = 1.004 – 1.398), being from the Bukedi, Ankole and kigezi regions (IRR= 1.429; 95% CI = 1.170 – 1.743), 24% (IRR= 1.239; 95% CI = 1.023 – 1.502) respectively were associated with increased frequency of meal intake. The findings proved that attention to distributional analysis and the use of Poisson regression rather than binary logistic regression improves the precision for identification of factors associated with frequency of meal intake. We recommend the use of the Poisson regression model for modelling the risk factors associated with frequency of meal intake. Also, we recommend a future study on the potential determinants of frequency of meal intake stratified by region.