METHODS FOR DYNAMIC PREDICTION: MULTI-STATE MODELS AND LANDMARKING
7-10 September 2015 – Residential course
Ponte di Legno, Brescia, Italy
FACULTY: HEIN PUTTER & HANS VAN HOUWELINGEN
Leiden University Medical Center (LUMC)
COORDINATORS: Maria Grazia Valsecchi & Laura Antolini
Center of Biostatistics for Clinical Epidemiology,
University of Milano-Bicocca
Prediction models play an important role in medicine to guide treatment decisions and to inform patients on their prognosis. The vast majority of prediction models developed in the medical literature have been designed to predict (disease-free) survival from diagnosis or start of treatment. But in clinical practice the patient regularly returns to the physician and it is important to be able to provide updated predicted probabilities of survival, taking into account clinical events that may have occurred, or clinical measurements that may have been made, between start of treatment and the time of prediction. Such prediction models, to be used after start of treatment and taking into account time-dependent information, are called dynamic prediction models. In this course we focus on the development and validation of dynamic prediction models in clinical survival analysis. It will be discussed how dynamic prediction probabilities can be obtained using traditional models and new approaches will be presented that have been developed in the last few years. The course will consist of a mix of lectures and computer practicals.
Applied statisticians, in particular biostatisticians who actively involved clinical research. Graduate PhD students interested in survival analysis.
Intermediate knowledge of survival analysis, basic knowledge of the R package.
This course is part of the summer course activities of Univerisity of Milano Bicocca in the Expo2015 framework