2.5 Statistical analysis
The analysis was performed using the statistical software program SAS
version 9.4(20). Descriptive statistics included means and standard
deviations for parametric data, medians and interquartile ranges for
non-parametric data and percentages for categorical data. Multivariable
logistic regression was used to explore the relationship between volume
of IV fluids in labour (high volume versus low volume) and estimated
maternal blood loss ≥500 mL. A p value of <0.05 was considered
statistically significant. Explanatory variables were determined prior
to the analysis based on a review of the current literature. These were:
maternal age, BMI, country of origin, parity, model of care, IV
antibiotics for infection/suspected infection, type of birth, degree of
perineal injury, duration of active labour, and birth weight. Continuous
explanatory variables (e.g. maternal age, BMI, and birthweight) were
tested for linear association by sorting into clinically relevant groups
(e.g., maternal age <25, 25-29, 30-34, and ≥35 years) and
plotting the beta-coefficients against the midpoints for each group.
Variables without linearity were analysed by these groups. Multiple
imputation was attended for missing data using 20 iterations. Logistic
regression was performed on each of the 20 datasets and summary
regression parameters were reported.