Performance of QUiPP app v.2 tool for prediction of preterm birth in asymptomatic high-risk women attending preterm specialist clinic: external validation study

L Creswell; D L Rolnik; B Burke; S Daly; N O'Gorman

doi:10.1002/uog.27619

Performance of QUiPP app v.2 tool for prediction of preterm birth in asymptomatic high-risk women attending preterm specialist clinic: external validation study

Ultrasound Obstet Gynecol. 2024 Feb 21. doi: 10.1002/uog.27619. Online ahead of print.

Authors

L Creswell¹, D L Rolnik², B Burke¹, S Daly¹, N O'Gorman¹

Affiliations

¹ Department of Obstetrics and Gynaecology, The Coombe Hospital, Dublin, Ireland.
² Department of Obstetrics and Gynaecology, Monash University, Melbourne, Australia.

PMID: 38379428
DOI: 10.1002/uog.27619

Abstract

Objectives: External validation of the QUiPP App v.2 algorithms in an independent cohort of high-risk asymptomatic women attending an Irish preterm birth surveillance clinic (PSC).

Methods: Retrospective, single center, observational study assessing discrimination and calibration of the QUiPP App v.2 at the six pre-determined clinical time points (birth prior to 30, 34, 37 weeks of pregnancy, and birth within one, two and four weeks of testing). Discrimination was assessed by estimating the area under the receiver-operating characteristics (ROC) curve (AUC) and sensitivity at fixed false-positive rates of 5%, 10% and 20%. Model calibration was assessed to evaluate the concordance between expected and observed outcomes. P-values <0.05 were considered statistically significant. No adjustments for treatment effects were made.

Results: 762 women and 1660 preterm birth surveillance clinic (PSC) visits utilizing the QUiPP between 2019 and 2022 were analyzed. The study population included 142 patients who later experienced a PTB (18.6%). QUiPP's performance to predict short-term outcomes such as birth within one week (AUC 0.866, 95% CI 0.755-0.955), two weeks (AUC 0.721, 95% CI 0.569-0.854) and four weeks (AUC 0.775, 95% CI 0.699-0.842), and delivery before 30 weeks (AUC 0.747, 95% CI 0.613-0.865) was superior to its ability to predict longer-term outcomes (birth <37 weeks; AUC 0.631, 95% CI 0.596-0.668). Calibration was generally good for low-risk results as the predicted risk in these patients tended to match the observed incidence. However, in women deemed to be at greater risk of delivery, the predicted probability superseded the observed incidence of PTB.

Conclusion: QUiPP accurately discriminates women who are at short-term risk of PTB. A treatment paradox may influence calibration in high-risk women. Further research is needed to ascertain if QUiPP treatment thresholds can be safely adjusted in women receiving prophylactic treatment to prevent PTB, and whether this improves outcomes. This article is protected by copyright. All rights reserved.

Keywords: prediction; preterm; risk assessment.