Statistical modelling of the association between dietary diversity, dietary patterns and non-communicable diseases in Namibia
Loading...
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Namibia
Abstract
Globalization coupled with urbanization has placed a significant pressure on the food systems of
many developing countries. This has led to lifestyle changes that have become one of the most
important influences on dietary patterns. The nutritional transition has affected the dietary pattern
and nutrient intake greatly and has led to a rise in the purchases and consumption of processed and
convenience foods. Analysis in nutritional epidemiology typically examined diseases in relation
to a single or a few nutrients or foods. However, people do not eat isolated nutrients. Instead, they
eat meals consisting of a variety of foods with complex combinations of nutrients. The high degree
of inter-correlation among nutrients as well as among foods makes it difficult to attribute effects
to single dietary components. Dietary patterns can influence health and the risk of developing
chronic conditions. Therefore, to gain full understanding of the relationship between diet and the
development of non-communicable diseases (NCD), it is desirable to use several methodological
approaches.
The main objective of this study was to explore the linkages between dietary patterns, dietary
diversity and prevalence of non-communicable diseases. Specifically, the study aimed at: (i)
applying count models on dietary diversity in Namibia, (ii) using bivariate count modelling
approach in analyzing convenience and non-convenience consumption food preference in
Windhoek, (iii) applying copula joint modelling of food insecurity indicators with application to
food insecurity prevalence (FIP), household dietary diversity score (HDDS) and months of
inadequate household food provisioning (MIHFP), (iv) fitting multiple indicators-multiple-cause
modelling to examine the relationship between foods consumed and non-communicable diseases.
The analysis used two representative survey data, namely the AFSUN-HCP Household Food
Security Baseline Survey (2016) and Namibian Household and Income Expenditure (NHIES) of
2015/2016.
The study focused on dietary diversity by using different count models. The household dietary
diversity score presented a mean score of 6.5, suggesting a moderate diverse diet, with less
consumption of food made from beans/lentils; eggs; fruits/vegetables and more consumption of
starch food. Determinants for household dietary diversity included educational level, sex of head
of household and main source of income (p-value <0.005). The study further used bivariate
iii
modelling approaches to analyze the food consumption patterns. The results found that, whereas
the consumption of food monthly was more on the non-convenience foods, the purchases of
convenience was frequent on a weekly basis and in multiple food sources. Moreover, the study
employed copula joint modelling of food security indicators. The findings show that AIC of the
untruncated (conditional/marginal) Poisson regression model was lower and thus proved to fit the
data better. The Frank Copula and Bivariate Normal Copula best fitted the data of establishing the
relationship between HFIP and HDDS, and between HFIP and MIHFP respectively. Lastly, we
analyzed multiple indicators-multiple causes examining the relationship between foods consumed
and non-communicable disease. Principal Component Analysis (PCA) and Structural Equation
Models (SEM) were used as data reduction methods to derive dietary patterns. Fruits, foods such
as condiments/tea/coffee and potatoes, yams, cassava, or any foods made from roots and tubers
accounted for majority of the variation.
The study concluded that the usage of appropriate methods for specific data types is very critical.
Generalized Poisson Regression models through the usage copula approaches are best to analyze
jointly two outcomes in order to test for significant relationships between high-level hierarchical
effects (e.g., random effects). Specifically, the bivariate normal and the Frank Copula were found
to fit the data best. The unique nature of the bivariate normal model is that it does not allow for a
different dependence structure between the outcomes while the frank copula does not have tail
dependence and it can model both positive and negative dependencies as the normal copula. SEM
and PCA’s were used as data reduction methods. Lastly, the study concludes that food and nutrition
insecurity is a major threat to the development of the country and the study recommends for
strengthened advocacy for consumption of healthy and diverse diets in the country in order to slow
down and arrest proliferation of non-communicable diseases
Description
A Dissertation submitted in fulfillment of the requirements for the Degree of Doctor in Science in Applied statistics
Keywords
Namibia, Statistical modelling, Dietary diversity, Dietary patterns, University of Namibia