Anomalous Adverse Event Detection
Shorter, more accurate trials by accelerating anomaly detection in clinical trials. Our goal for a clinical trial management company was to reduce the manual effort & duration required to find anomalies in clinical trials, in order to allow for shorter duration, more accurate trials, and to flag safety issues early.
Discovery into how clinical trials data was being reviewed manually by a medical data review team for issues related to data entry, protocol variation across sites, etc. based on various rules developed over time. Develop ML models that combine various modalities of data — including labs, drug exposure, and previously reported adverse events — to develop various multivariate anomaly detection methods. Developed data app to present potential anomalies to medical data reviewers along with explainability about why a data point was flagged.
Improvement in speed and efficiency for anomalies being identified manually.