Course Project for BM60130 taught at IIT Kharagpur
Unstructured clinical data such as nursing notes are very less used to build a predictive model for post-discharge mortality despite containing rich information. Our work examines a simple bag of words model approach for 7/30/180/365 day mortality prediction. We also explored syntactic sentiment dimensions from these nursing notes as a predictor of mortality, and report preliminary survival analysis results too. Our simple BOW model using XGBoost achieved 0.92 AUC for 30-day mortality.
Link to Project Report is here