Statistical Analysis
Descriptive statistics were generated to summarize the risk
characteristics of the study population. Chi square and
Student’s t tests were used to compare categorical and
continuous outcome variables in the study exposure
groups. Several women delivered multiple live births during
the study time period. To compensate for the lack of
independence among observations, the outcome of PTB
was modeled using generalized estimating equations
(GEE); the use of a log link function and a binomial distribution
assumption were used to derive relative risks [26].
An exchangeable correlation structure was assumed and
has been used in similar research [15]. Models were built
using a stepwise selection approach with entry criteria of a
p value of 0.2 and retention criteria of a p value of 0.15;
however, only variables that reduced the quasi-likelihood
information criterion (QIC) were retained in the final
models.