The development of an iphone application to predict preterm birth in high risk women. (7th June 2011)

• Record Type: Journal Article
• Authors: Smout, E M; Seed, P T; Shennan, A H
• Is Part Of: Archives of disease in childhood. Volume 96(2011)Supplement 1
• Page Start: Fa124
• Page End: Fa124
• Publication Date: 2011-06-07
• Subjects: Infants -- Diseases -- Periodicals; Newborn infants -- Diseases -- Periodicals; Fetus -- Diseases -- Periodicals; 618.920105
• DOI: 10.1136/archdischild.2011.300157.5
• Languages: English
• ISSNs: 1359-2998

Abstract

Abstract : Introduction: Fetal fibronectin (fFN) and cervical length are the best predictors of preterm birth (PTB). No studies have attempted to predict PTB in asymptomatic high risk women by combining demographics, past obstetric history and biochemical markers. We have developed an algorithm for predicting PTB before 30, 34 and 37 weeks' gestation. Methods: We analysed 219 women attending the Preterm Surveillance Clinic at our institution between 1 September 2007 and 31 August 2009. These women were at high risk of PTB following previous PTB (16–37/40) or cervical surgery. Multiple linear logistic regression analysis was performed using Stata. Results: fFN, shortest cervical length and gestation of fFN test were the only variables that influenced PTB risk when combined; previous PTB, ethnicity, smoking and BMI had no effect on ability to predict PTB. The area under the ROC curve for the prediction models was 0.96, 0.84 and 0.77 for delivery before 30, 34 and 37 weeks' gestation respectively. We have produced a formula for calculating percentage risk of delivering before 30, 34 and 37 weeks', and within 2 and 4 weeks of the calculation. This has been developed into a freely available iPhone application. Conclusion: Predicting PTB in high risk women is principally based upon cervical length, fFN result and gestation of fFN test; women with a positive fFN at an earlier gestation, with a shorter cervix are at greatest risk. Additional demographic information and past obstetric history are superseded by these variables and need not be applied to a high risk prediction model.