Flagship Projects

Risk Prediction

Recent Projects

EHR-based Bipolar Disorder Risk Prediction Models

To develop a basis for clinically-relevant early identification efforts in healthcare settings, we are training and cross-validating EHR-based algorithms that predict the risk of bipolar disorder (BPAD). Various machine-learning approaches (Naive Bayesian Classifier, Random Forest, XGBoost) are being trained and cross-validated in this multisite study.

EHR-Genetics Risk Prediction Models

Leveraging the genome-wide association study (GWAS) data available at various sites, we are assessing the impact of incorporating polygenic risk scores (PRS) into electronic health record (EHR)-based risk prediction model.

Asset 6