Visualization of Radiology Errors
Working with a resident from Emory University Hospital, we had the opportunity to develop tools that display attributes of errors made by residents over the past two years. The tool supports immediate analysis while remaining flexible so that data can continue to be collected and uploaded into the future.
This was an exciting and important topic as it can greatly impact the healthcare provided to patients. Human error can cause serious, sometime fatal, outcomes; one of the first steps to reducing errors is to understand when and why these errors occur. Few visualization tools exist to help radiology physicians at this facility learn from their mistakes so we were excited to be able to provide such a tool to some of the Emory residents.
Our Users and context
We have worked very closely with our client to understand the future users of this tool. Our users are physicians practicing at Emory University School of Medicine. Our users have two central needs for a diagnostic error visualization display.
First, the tool will be used as part of a quality assurance program. For instance, residents will be able to use the application to see what common errors are made by their peers of similar experience level.
Secondly, the visualization will serve as an educational tool for physicians; the data can be used to focus educational lectures on commonly made errors. These needs indicate that this tool must be a combination of exploratory as well as directed information display. In addition the tool must be able to accept more data as future errors are made and recorded so that users may monitor changes over time.
These needs indicate that this tool must be a combination of exploratory as well as directed information display. In addition the tool must be able to accept more data as future errors are made and recorded so that users may monitor changes over time.
We have opted for a dashboard interface. Multiple interfaces will better allow our users to navigate the data and draw their own conclusions. This flexibility encourages our users to explore the data in ways that best supports their learning and understanding. The visualizations are modeled to best mine the supplied dataset which focuses on errors and details of those errors that have been made in the recent past.