Prediction of delivery among women with early preterm labor by means of clinical characteristics alone

Objective: This study was undertaken to assess whether individual clinical factors or combinations thereof could be used to accurately predict the risk of delivery within 1 week of admission among women with preterm labor and minimal cervical dilatation. Study Design: We performed a case-control study of patients admitted to our institution with preterm labor and minimal cervical dilatation. A case patient was a patient who sought treatment with uterine contractions between 24 and 34 weeks’ gestation with cervical dilatation ≤2 cm, who received tocolysis with magnesium sulfate, and who was delivered within 7 days of admission. Control subjects were those who had the same presentation with preterm labor but were not delivered within the first 7 days after admission. The medical records of case patients and control subjects were then abstracted, and information on >70 potential predictors was recorded. Statistical analysis consisted of bivariate and multivariable methods. We also generated a multivariable clinical predictive model with the purpose of detecting a proportion as high as possible of those destined to be delivered within 1 week (high sensitivity). We estimated that we would need 50 case patients and 150 control subjects to detect an odds ratio of 2.5 for risk factors with a prevalence of 20%, an α error of .05, a β error of .20, and a control subject/case patient ratio of 3:1. Results: Three variables were eligible for inclusion in our logistic models according to the bivariate analyses—bleeding on admission, substance abuse, and admission white blood cell count ≥14,000 cells/μL. The simplest and most favorable model included only 2 variables, bleeding and substance abuse, and yielded a sensitivity of 46% and a specificity of 76%. The full 3-variable model had similar test characteristics. For no model were we able to achieve a sensitivity ≥50%. Conclusion: The results of this case-control study suggest that combinations of clinical factors do not yield an adequate level of discrimination to be used alone for predicting the likelihood of delivery within 1 week among patients with minimal degrees of cervical dilatation. (Am J Obstet Gynecol 1999;181:1414-8.)