Using Explainable AI to Improve Healthcare (Audio Described Version)

NIH/NLM K99/R00 Awardee, Dr. Siru Lu, has developed a data-driven process to generate suggestions for improving alert criteria using explainable artificial intelligence (XAI) techniques. Her approach identifies improvements regarding clinical decision support (CDS) that might be overlooked or delayed in manual reviews. It also unveils a secondary purpose for the XAI: to improve quality by discovering scenarios where CDS alerts are not accepted due to workflow, education, or staffing issues.
Non-Audio Described Version - • Using Explainable AI t...
reporter.nih.gov/search/8ckk8...
academic.oup.com/jamia/advanc...
Video transcript:
[Peter J. EMBI, MD]
When I'm seeing a patient in clinic,
I sometimes don't remember all of the things that I need to do. And and so the, the most basic explanation of what decision support is, is it's an alert that actually reminds me, as a doctor, what it is I'm supposed to be doing to take better care of my patient, whether that's ordering a test or providing a treatment.
[Adam Wright]
We in fact unfortunately like this so much
that we sometimes create
such a large volume of alerts and suggestions that nurses and doctors can feel frustrated.
[Allison McCoy]
Alert fatigue happens when clinicians continuously get these alerts - especially alerts that are not relevant to them or their patients. So they continue to override them.
But oftentimes that advice is important for patient safety.
[Siru Liu]
In my project, of using explainable AI
to optimize clinical decision support,
I'm trying to identify ways to improve alert logic and also identify potential improvements in workflows.
[McCoy]
Traditionally, when we evaluate clinical decision support alerts, we have to look through thousands of alert firings every day.
But what Siru is doing is using artificial intelligence, which is really good at looking at a lot of data.
[EMBI]
I really think NLM's funding and support
was key to this research.
There's so much that we need to do in order to fill the important gaps in our research knowledge and be able to advance the field.
[McCoy]
This is a grant program, from the National Library of Medicine and
other agencies, that allows faculty members to start off working on their research projects under some mentorship for a few years and then transition into an independent award to do the research that they want to do.
[Liu]
Without the NIH-NLM funding, I cannot do this project.
It is very helpful for my research and also for my career.
[Wright]
Siru has brought some really cutting edge machine learning techniques to find opportunities to improve our decision support.
We've already made tremendous strides in improving the acceptability, reducing alert fatigue, and making alerts more usable for our users.
[EMBI]
I think this research is going to help us really
maximize and optimize the ways in which we really can unlock the potential of AI for improving health and health care.
@NIHgov #ai #machinelearning #healthcare #datadriven