Prediction of successful expectant management of first trimester miscarriage: Development and validation of a new mathematical model

Abstract

To generate and evaluate a new logistic regression model for the prediction of successful expectant management of first trimester miscarriage. Data were collected prospectively from women diagnosed with 1st trimester miscarriage. Clinical and ultrasonographic variables were recorded for multivariate analysis. Clinically stable women who were managed expectantly were followed up for two weeks until the outcome was established: success or failure. A multinomial logistic regression (MLR) model was developed on 186 training cases for the prediction of successful expectant management and tested prospectively on a further 126 cases. The performance of the model was evaluated using receiver operating characteristic (ROC) curve as well as in terms of sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV). Two thousand and forty eight consecutive first trimester women underwent TVS. Complete data from 312 (15.2%) women with miscarriage managed expectantly were included in the final analysis. The most important independent prognostic variables for the MLR model were as follows: type of miscarriage at primary scan, vaginal bleeding and maternal age. When developed retrospectively on a training data set, MLR model gave an area under the ROC curve (AUC) of 0.796. Prospective validation of MLR model on a new test data set resulted in an AUC of 0.803. We have developed and validated a new mathematical model to predict successful management of first trimester miscarriage.