High Risk Pregnancy Prediction from Clinical Text

Patients with high risk pregnancies can benefit from case management and additional care. In order to provide patients with those services, they may be enrolled in case management programs. Machine learning and natural language processing (NLP) methods can be leveraged to automatically detect patients for referral to such programs. We describe initial experiments in predicting high risk pregnancies from clinical text data collected from electronic medical records (EMRs). We show that simple text features from unstructured records outperform baseline classification.