Dear Editor,

Dehaene and colleagues’ [1] criticise the description and reporting of the methodology behind development of the QUiPP app’s prediction models. The algorithms have been extensively peer-reviewed and published [2, 3, 4] and the data is available on request for those wishing to confirm our findings or, indeed, to scrutinise the methodology. Deheane et al.’s in-depth feedback will assist us in honing our presentation of how QUiPP achieves its predictive accuracy. Any group creating a preterm predictive tool will encounter a range of choices regarding diverse statistical approaches and compromises to ensure clinical utility and safety. However, the authors present no evidence to support the suggestion that the device may be inaccurate. Validation of the QUiPP prediction models was robust, and subsequently proven on many different populations based on data collected through large prospective multi centre studies. QUiPP’s rigorous external validation in a real-world setting supersedes their concerns. The onus is on investigators to show other populations are different from the many thousands of participants already investigated.


Yours faithfully,


Jenny Carter, Naomi Carlisle, Anna David, Jane Sandall, Paul Seed, Andrew Shennan, Rachel Tribe and Helena Watson.