Improving Patient Safety Through AI-Assisted Incident Review

Michigan Medicine

Problem

As a large hospital system, Michigan Medicine files a large number of patient safety incident reports, which are then reviewed by an internal team. This team frequently faces capacity constraints in performing these reviews due to the high volume. 

Audience

Michigan Medicine patient safety report incident reviewers

Outcome/Impact

Michigan Medicine implemented a two-layer AI review system powered by Maizey for patient safety incident reports. Incident reports, which are often a mix of free text and structured data, are first processed through an extraction layer to clean and standardize inputs before being analyzed by Maizey. The system then determines whether an incident requires further review and classifies its severity and type. Achieving 98–99% accuracy, the solution dramatically reduces manual workload while maintaining high-quality oversight. By introducing intelligent filtering, teams can focus on the most critical cases, demonstrating how Maizey enables scalable, high-stakes operational workflows without requiring perfect automation.

Relevant Topics