Dr Glen Martin, University of Manchester
Risk prediction models are tools that compute the risk of an adverse outcome given a set of patient characteristics. Arising from the desire to move health systems away from managing or curing disease towards preventative medicine, these tools have become popular and several are now embedded in clinical practice. They are typically based on statistical or machine learning models, derived by analyzing historical patient data (e.g., routinely collected electronic health records). Our group is engaged in a program of methodological research to improve the ways in which these models are developed and validated. In this talk, I will overview some of this methodological work, including topics such as: (i) updating of existing risk prediction models to suit local settings, (ii) missing data/ informative presence, (iii) incorporating longitudinal data in risk prediction models, (iv) penalisation and shrinkage for prediction models, and (v) development of models for multiple outcomes (multivariate prediction models).
This will be a virtual seminar. If you would like to attend, please email email@example.com for the joining information.