Longitudinal Discriminant Analysis with Random Effects for Predicting Preeclampsia using Hematocrit Data

Preeclampsia is the third leading cause of death in pregnant women. This study was conducted to evaluate the ability of longitudinal hematocrit data to predict preeclampsia and to compare the accuracy in longitudinal and cross-sectional data. In a prospective cohort study from October 2010 to July 2011, 650 pregnant women referred to the prenatal clinic of Milad hospital in Tehran were selected. The hematocrit level was measured in the first, second and third trimester of pregnancy and the participants were followed-up to delivery. The preeclampsia cases were recorded. The Covariance pattern and linear mixed effects models were applied for discriminant analysis of the longitudinal data. Statistical analyses were performed in the SPSS-20 and SAS-9.1. The prevalence rate of preeclampsia was 7.2% (47 out of 650 women). The women with preeclampsia had a higher meanhematocrit values (difference=0.99; P=0.014). The sensitivities for longitudinal data and cross-sectional data in three trimesters were 91%, 54%, 72%, 51% and the specificities were 61%, 51%, 51%, and 47%, respectively. The positive predictive values were 70%, 52%, 59%, 49% and the negative predictive values were 87%, 53%, 64%, and 49%, respectively. The levels of hematocrit can be used to predict preeclampsia and to monitor the pregnant women. Measuring the hematocrit during the three trimesters regularly can help to identify women at risk for preeclampsia.