BSU Virtual Seminar: 'PROGRESS in sample size calculations for clinical prediction model research'
Duration: 57 mins 20 secs
Share this media item:
Embed this media item:
Embed this media item:
About this item
| Description: |
Speaker: Professor Richard Riley, Centre for Prognosis Research, School of Medicine, Keele University
Title: ‘PROGRESS in sample size calculations for clinical prediction model research' |
|---|
| Created: | 2020-12-09 09:36 |
|---|---|
| Collection: | BSU Virtual Seminars 2020 |
| Publisher: | University of Cambridge |
| Copyright: | A.S. Quenault |
| Language: | eng (English) |
| Distribution: |
World
(downloadable)
|
| Keywords: | biostatistics; clinical trials; |
| Explicit content: | No |
| Aspect Ratio: | 4:3 |
| Screencast: | Yes |
| Bumper: | UCS Default |
| Trailer: | UCS Default |
| Abstract: | There is a growing demand to personalise treatment and healthcare for individuals based on their prognosis and/or predicted response to treatment. For this reason, prognosis and prediction research has never been more important. Sadly, empirical evidence has shown that prognosis and prediction studies are often poorly designed, badly analysed, and selectively reported. The Prognosis Research Strategy (PROGRESS) framework was established to help address such shortcomings. In this talk, I will describe the PROGRESS framework, and highlight latest methodology guidance for calculating the sample size required for developing and validating clinical prediction models.
In terms of sample size for model development, current “rules of thumb” are based on having at least 10 events per predictor variable, but I will describe a more scientific approach based on minimising expected overfitting and ensuring precise parameter estimation. In terms of sample size for model validation, I will introduce a new approach that targets precise estimation of key model performance measures. Real examples are used to illustrate the concepts. The talk is intended for a wide audience. |
|---|---|