Statistical Analysis
All continuous data were assessed for normality and summary statistics were reported as mean (standard deviation), median (interquartile range) or number(percent) as appropriate. Demographics of the women before and after the guideline was implemented were tabulated compared using either a t-test, a Kruskal Wallis or chi2 test as appropriate. The rate of stillbirth per 1000 births before and after the guideline for south Asian born women and all other women were determined. The association between the change in clinical practice and rates of stillbirth was then determined using log linked generalised linear regression. Adjustment with maternal age, BMI, parity, and presence of pre-existing medical conditions (pre-existing hypertension, type 2 diabetes or thyroid disease) or obstetric conditions (gestational hypertension, pre-eclampsia and gestational diabetes) was also performed. The rate of stillbirth per ongoing week for south Asian born and all other women before and after the change in practice was determined and plotted. Differences in secondary outcomes stratified by maternal exposure group before and after the change in practice were also tabulated and compared using standard approaches. When evaluating the impact of interventions to reduce stillbirth, understanding potential harms or changes to intervention rates need to be balanced. The rates of induction of labour are widely acknowledged to be increasing therefore pooling and comparing rates before and after the change is not appropriate. Therefore we performed a two-group, interrupted time-series analysis using Newey-West standard errors and a lag of 1 to determine the change in rates of induction of labour over time following the implementation of the guideline, taking into account the trends in induction over time prior to the guideline between south Asian and all other women18. We did not correct for seasonality. All statistics were performed using StataCorp 12.