An Automated Patient Chart Error Detection System for Radiation Therapy

Project Summary/Abstract Every year, approximately 1,200 severe mistreatments happen in radiation therapy. Radiation therapy lawsuits rank in the top third of all medical specialties with an average of $313,000 per claim settled or litigated. The current method for detecting treatment errors is by a weekly patient chart check, where each treatment record is manually reviewed on a weekly basis. This labor-intensive and inefficient method prevents us from detecting the treatment error at an early stage. Here we propose a novel software system, ChartAlert, for automating patient chart checking. ChartAlert is a prospective real-time adaptive electronic checking system that can be configured to support different clinical workflows and perform “smart” check using artificial intelligence. It supports two major treatment databases (Elekta MOSAIQ and Varian ARIA) in radiotherapy. In Phase I project, we have successfully developed ChartAlert for MOSAIQ prototype that is under clinical testing in two treatment centers. Our preliminary results demonstrated the significant improvement of effectiveness in patient chart checking and the flexibility of supporting different workflows. In this Phase II proposal, we will continue the ChartAlert development. We will demonstrate the feasibility of the ChartAlert approach and its advantages over the standard manual checking method. We will develop a prospective checking module, develop the ARIA data translation module for ChartAlert for ARIA, design and implement an AI-based “smart” check module, and verify the proposed system at the partner sites. Successful completion of these aims will demonstrate the feasibility and commercial potential of the ChartAlert approach. Ultimately, this work will result in an intelligent patient chart checking software, which will increase patient chart check efficiency, save staff time, improve cancer patient treatment safety, and preventing potential lawsuits.Project Narrative Treatment errors in radiation oncology occur at a rate of 2% per patient, and radiation therapy lawsuits rank in the top third of all medical specialties with regard to claims made, claims paid, and damages. Current methods of detecting treatment errors are manual and inefficient. There is a critical need for efficient treatment error detection in order to improve patient safety and save cost. We propose to develop a scalable and comprehensive software system (ChartAlert) for automated patient chart error detection in radiation therapy. ChartAlert can be extended to other types of patient charts to check treatment and prescription consistency and improve patient safety.