An improvement in the method used to assess discriminatory ability when predicting the chances of a live birth after one or more complete cycles of in vitro fertilisation

Our paper presents two clinical prediction models that estimate the chance of having a baby over multiple complete cycles of in vitro fertilisation (IVF)-that is, cumulative live birth.1 The pretreatment model predicts the chance of cumulative live birth before treatment starts, and the post-treatment model predicts the chance of cumulative live birth just after the first embryo transfer. Through a collaboration with researchers from the University of Utrecht, who have externally validated these models, we have decided to revise the method used to assess the discriminatory ability of our models in the original study. In time to event models, such as ours, discrimination indicates the proportion of all pairs of women who can be ordered such that the woman with the lower predicted chance of live birth is the one who either did not have a live birth or had more complete cycles of IVF to have a live birth. Discrimination is assessed using the C index, where 1 is perfect discrimination and 0.5 is no better than a coin toss.2 We request this amendment so that other researchers externally validating these models can use this revised approach to calculate the C index. They can then compare the discriminatory ability of their own cohort with the model applied to the original development cohort (as revised and presented here). © 2018 BMJ Publishing Group. All rights reserved.