Predicting pregnancy outcomes using longitudinal biomarkers: Analysis of urinary human chorionic gonadotrophin levels in normal and failing pregnancie

My enthusiasm for perinatal medicine has been longstanding. In fact, my MSc project was based around building a prediction model to predict the number of days that preterm babies received respiratory intervention. The project helped me to familiarise myself with a standard perinatal dataset including clinical variables such as APGAR scores. I have endeavoured to remain linked to this area of medicine, even whilst working in diabetes, by taking on the role of trial statistician for a study evaluating the effectiveness of an educational intervention, to reduce the risk of Type 2 diabetes in women who have previously developed gestational diabetes. Outside of my job my desire to make a difference to the care of women, particularly in pregnancy, encouraged me to take part in a focus group with researchers considering the barriers to giving birth in midwife led birth centres. Prediction modelling is a prevalent method of analysis in diabetes, whether this involves a direct outcome of development of diabetes or predicting the reduction in glucose and related outcomes, due to an intervention. Many analyses I have conducted have involved several follow up measurements for each patient, resulting in analysis of longitudinal data using multilevel modelling. I have also been fortunate enough to conduct a competing risks analysis modelling clinical inertia; the time taken for a healthcare professional to prescribe a drug after diagnosis of high glucose levels. I am excited to potentially explore the joint modelling of these types of data in a field I am passionate about.