Oxford RSS meeting with Ruth Keogh
Wednesday, 17 May 2017, 4pm to 5pm
Richard Doll Building Lecture Theatre, Richard Doll Building, Old Road Campus, University of Oxford
All those with an interest in statistics are invited to join the Oxford local chapter of the Royal Statistical Society for an afternoon talk by Ruth Keogh
Royal Statistical Society - oxford local chapter
The Oxford local chapter of the Royal Statistical Society, hosted by the University of Oxford, brings together anyone with an interest in statistics in Oxfordshire and the surrounding areas. We organise an annual programme of free events and lectures on statistical topics.
Academics and non-academics are all welcome to join this local RSS chapter. For advance notice of events, join our mailing list by emailing firstname.lastname@example.org. We are always interested in new ideas for events and lectures, so do get in touch!
Speaker: Ruth Keogh
Ruth Keogh is an Associate Professor in the Department of Medical Statistics at the London School of Hygiene & Tropical Medicine (LSHTM). She is currently funded by a MRC Methodology Fellowship and is focusing on methods for dynamic prediction of survival in complex observational data using landmarking, and extensions of landmarking to address other questions in survival analysis. She is especially interested in applications in cystic fibrosis.
Ruth’s other research interests include case-control study design and analysis, methods for handling missing data and measurement error, and causal inference methodology.
Dynamic prediction of survival using landmarking in large healthcare databases, with an application in cystic fibrosis
In 'dynamic' prediction of survival we make updated predictions of individuals' survival over time as new information becomes available about their health status via longitudinal measurements. Landmarking is an attractive and flexible method for dynamic prediction. I will give an introduction to landmarking and a practical overview of how to use this approach, including some recent developments.
Large observational patient databases, which provide longitudinal data on clinical measurements, present opportunities to develop 'personalised' dynamic predictions of survival. I will present an example application of landmarking for dynamic prediction of survival in people with cystic fibrosis, using data from the US Cystic Fibrosis Foundation Patient Registry. This will include discussion of some of the challenges faced in making dynamic predictions using routinely collected data and how they can be addressed in the landmarking framework. I will also show some comparisons between landmarking and the alternative approach of joint modelling, and hopefully convince you that landmarking has a number of advantages.