Empirical Evidence That There Is No Such Thing As A Validated Prediction Model
Published in arXiv preprint, 2024
In the paper we show that there is a lot of heterogeneity between the performance measure AUC in external validations of the same clinical prediction model. This heterogeneity cannot be estimates well if there are only a couple of external validations, in those cases the predicition interval around the AUC will be to slim. A way to fix this problem is to use a prior for the heterogeneity.
Recommended citation: van Leeuwen, F. D., Steyerberg, E. W., van Klaveren, D., Wessler, B., Kent, D. M., & van Zwet, E. W. (2024). "Empirical Evidence That There Is No Such Thing As A Validated Prediction Model." arXiv preprint, arXiv:2406.08628.
Download Paper | Download Slides