Abstract:
One of the most important measures of a country’s development is the number of healthy
infants. Infant health is mostly related to birth weight, which is greatly influenced by
the diet of pregnant women, and nowadays most infants are underweight or overweight.
This study aims to investigate the pregnancy-related factors that are associated with
the birth weight of infants in the Western Province of Sri Lanka. Data for the study
were obtained from a census that having data of 8313 conducted by the Department
of Census and Statistics, Sri Lanka in 2016 and the birth weights of the infants were
categorized into three sectors as Low Birth Weight (< 2500g), Normal Birth Weight
(2500g - 4000g), and Extreme Birth Weight (>= 4000g). Statistical tests like Kruskal-
Wallis and Kendall’s rank correlation test were applied. The Ordinal Logistic model was
fitted to find out the factors that are associated with infant birth weight, and machine
learning models were also applied. According to the statistical tests and the fitted ordinal
logistic model, pregnancy duration, time spent in the hospital, gender of the infant, birth
type (single or multiple), getting iron pills, and delivery type were identified as the key
factors that are significantly associated with the infant’s birth weight. The best model
out of fitted machine learning models was found to be the Extreme Gradient Boost
model (overall accuracy = 65%), where pregnancy duration in months (below 6, 7, 8, 9,
and 10), getting iron pills and folic acid, and the number of injections of tetanus given
throughout pregnancy are reported as the most important factors with the newborn
birth weight.
