Non-invasive prediction of clinical infection in women with preterm labor / 대한산부인과학회지

ABSTRACT

OBJECTIVE: To develop a model based on non-invasive variables to predict the probability of subsequent maternal and/or neonatal clinical infection in women admitted to hospital for preterm labor with intact membranes. METHODS: Transvaginal ultrasound for measurement of cervical length was performed and maternal blood was collected for the determination of white blood cell (WBC) count at admission in 165 consecutive women with preterm labor (between 20.0 and 35.0 weeks). Clinical infection was defined as the presence of clinical chorioamnionitis at delivery or early onset neonatal sepsis. Receiver operating characteristic (ROC) curves and logistic regression analysis were used for statistical analyses. RESULTS: The prevalence of clinical infection was 5% (8/165). Women who developed clinical infection had a significantly lower median gestational age at admission, a lower shorter median cervical length, and a higher median WBC count as compared to those who did not develop clinical infection. Logistic regression analysis was performed and a final model was chosen, which included maternal blood WBC, cervical length, and gestational age as the best predictors of clinical infection. A risk score was calculated containing these 3 variables for each patient. The model was shown to have an adequate goodness of fit (P=0.202), and the area under the ROC curve was 0.822, indicating reasonably good discrimination. CONCLUSION: In women admitted to hospital for preterm labor with intact membranes, the risk for the subsequent maternal and/or neonatal clinical infection can be predicted non-invasively with a risk score based on cervical length at admission, maternal blood WBC, and gestational age.