2.4 Study sample size
Study sample size was calculated using the odds ratio for multiple logistic regression for a binary outcome using the method described by Hsieh et al.(19) We assumed an alpha value of 0.05, a beta value of 0.2, P1 (rate of PPH in the unexposed group) of 20%, an odds ratio of 1.65 (a relative risk of approximately 1.5), and B (proportion of the population exposed to IV fluid ≥ 2500 mL) = 50%. This gave a required sample size of 686. We then calculated the R2 using data from a previous study(11) to estimate that 32% of the variance in the use of IV fluids was due to the covariates and adjusted the sample size using a variance inflation factor of 1.47 [=1/(1-R2)] to calculate the final required sample size of 1009 participants.