The Applied Statistics to Health & Medicine (YSM Prize Winners) event on 2 May was jointly hosted by the Glasgow local group and the Young Statisticians section of the RSS. The event consisted of Young Statisticians Meeting (YSM) prize winners Emily Granger (University of Manchester) and Kathryn Leeming (University of Bristol) presenting their PhD statistical research in health and medical contexts. The event was chaired by Johnathan Love and had approximately 30 in person attendees at the University of Glasgow and 16 attendees that joined via livestream.
Emily Granger (YSM Prize Winner, 2016) presented her research on propensity score diagnostics and the challenges we face when ascertaining if a propensity-based estimate is unbiased. In a health context and when dealing with observational patient data, these scores (defined as a patient’s probability of receiving treatment conditional on their baseline characteristics) are becoming an increasingly popular method used to account for confounders. However, Emily highlighted that there is some debate on the reliability of these scores, particularly the balancing of these scores across covariate distributions. Emily outlined why some of the balance diagnostics could possibly mislead users of propensity scores: results from a simulation study indicated commonly used diagnostics, eg standardised differences, can be unreliable. Emily also discussed her future plans on developing new propensity score diagnostics, such as cumulative prevalence.
In the second part of the session, Kathryn Leeming (YSM Prize Winner 2017), presented her research on network time series, which are data collected over time at nodes of a graph/network and occur in a wide variety of areas including environmental, financial and indeed medical. Kathryn presented results of analyses, which involved the Network Autoregressive (NAR) framework, using England-based Mumps disease incidence data. Kathryn also highlighted features of the NAR model and demonstrated how to recover an un-weighted graph from a multivariate time series using the NAR model. She also highlighted the advantages of using network-based methods for analysing such medical data and compared the performance of these methods to those which do not involve utilising network structures.
Overall, this was an engaging session on modern statistical methods applicable to the fields of health and medicine.