Using ROC surface to predict preterm delivery based on hemoglobin level in the first trimester of pregnancy

Receiver Operating Characteristics (ROC) curves have numerous applications for identifying a cut-off point in diagnostic tests. Nonetheless, given that sometimes two cut-off points have to be specified simultaneously, the ROC curve can be used to identify such points. The Volume under the ROC Surface (VUS) serves as a criterion for the accuracy of diagnostic tests. One of the unfortunate outcomes in pregnancy is pre-term delivery; it has been noted that an increase in the level of hemoglobin in the first trimester of pregnancy could result in preterm delivery in weeks 34 to 37 and that an ongoing hemoglobin increase could result in the delivery of a premature fetus before the 34th week of pregnancy. In this regard, in order to separate three groups of on-time delivery, pre-term delivery and immature delivery two cut-off points have to be identified, simultaneously. A suitable measure to identify such points is the ROC surface. In the current study, the hemoglobin information of the first trimester of pregnancy and delivery time of 623 pregnant ladies referring to Milad Hospital in Tehran in 2009-2010 was obtained. ROC surface was adopted to draw two ideal cut-off points for the first trimester of pregnancy. The optimal points for hemoglobin of the first trimester computed with the ROC surface were 12.54 and 13.2. While a hemoglobin rate less than 12.54 indicated an on-time delivery, a rate between the two cut-off points referred to pre-term delivery and hemoglobin more than 13.2 showed a premature fetus. The three-dimensional ROC surface is a useful tool that can visually summarize the ability of a biological marker to classify individuals between more than two groups.