Clinical prediction rules for preterm birth in patients presenting with preterm labor.

Abstract

To develop prediction rules to identify which women with preterm labor are at greatest risk for delivery within 10 days and before 37 weeks of gestation using demographic and clinical risk factors alone. We analyzed data collected for a prospective cohort study of singleton pregnancies at 22-33 6/7 weeks of gestation with preterm labor. Potential risk factors were included in multivariable logistic models for each outcome. Using backwards regression, we identified combinations of risk factors that generated the most parsimonious yet predictive models. Adjusted odds ratios of covariates in the final models were used to estimate weights for each risk factor and were summed to generate a predictive score. The score associated with the highest negative predictive value was defined as a positive test result for each outcome. Bootstrapping techniques internally validated the scoring systems. We include data from 583 women. The risk of delivery within 10 days was 15.4% (n=90) and before 37 weeks of gestation it was 35.0% (n=204). The final model for delivery within 10 days included initial cervical dilatation, no prenatal care, and tobacco use (area under curve=0.75), and for delivery before 37 weeks of gestation it included initial cervical dilatation, obstetric history, and tobacco use (area under the curve=0.73). A positive test result was associated with 84% sensitivity, 51% specificity, 24% positive predictive value, and 95% negative predictive value in predicting delivery within 10 days and 79% sensitivity, 50% specificity, 46% positive predictive value, and 82% negative predictive value in predicting delivery before 37 weeks of gestation. Based on their strong negative predictive values, these prediction rules could identify patients who do not require intensive monitoring when they present with preterm labor. II.