Anesthesiology providers are believed to be at least as likely as the general population to suffer from non-medical opioid use. Currently, there is a lack of data-driven methods to efficiently identify providers in need of support. This project applies machine learning models to electronic health records to detect changes in provider practice over time.
Over the course of the project, the team has constructed a two-phase approach. The first phase involves constructing a pipeline that models opioid usage and calculates the excess opioid usage over what is expected for an anesthesia provider. The second phase involves flagging shifts in excess usage using a change point detection model, particularly focusing on providers who exhibit multiple changes within the last 180 days of their service. We identified a significant change in practice in six providers out of 1488. The Duke Health team manually validated these results as either notable documentation errors or apparent changes in treatment practice.
