This document discusses leveraging electronic health records for predictive modeling. It provides an overview of the size and types of data available from different sources in a hospital's electronic health records. It also describes how models can be developed to predict outcomes like readmission or complications at different time intervals. The document outlines the process used to develop datasets for pre-surgery, 30-day post-surgery, and surgery completion models, including matching data, adding features from other sources, and handling outliers. Finally, it mentions some potential applications of these predictive models, like mobile apps to identify problems and reduce clinician effort.