Artificial Intelligence (AI) has the potential to transform healthcare. It is already helping to reduce clinician workloads and speed the analysis of large populations of patients. There is, however, a trust gap that is developing which, if left unaddressed, may hold back the widespread implementation of AI. Bridging that gap requires transparency, working closely with clinicians, and operating within strict guidelines. SAS is working towards all these goals.
Healthcare IT Today recently had the opportunity to speak one-on-one with Dr. Stephen Kearney, Global Medical Director at SAS, a world leader in analytics. We asked him to talk about how SAS is working with healthcare organization on the data side of AI as well as the AI algorithms themselves.
AI Trust Gap
SAS is a founding member of the Coalition for Health AI (CHAI) which has a goal of developing a framework, with health equity in mind, that addresses algorithmic bias. SAS works alongside other founding members including Dr. John Halamka from Mayo Clinic and Dr. Michael Pencina from Duke Health.
Together with their Coalition partners, SAS is creating model cards which help organizations understand the inner workings of AI algorithms and applications.
“We need guidelines, parameters and transparency,” said Dr. Kearney. “The model cards help with that. The model cards that SAS developed are like what’s on the side of a cereal box. These cards help you understand what the ingredients are so that you can make good choices. It’s really about transparency.”
That transparency is key to gaining the trust of clinicians and administrator who may be hesitant to implement AI. Also key is understanding the opportunity for bias in the datasets that are used to train AI algorithms. Dr. Kearney was quick to point out that is it impossible to completely eliminate bias in data, but being aware of what bias may be present can help identify the operating parameters (limitations) of an AI model.
According to Dr. Kearney, just because there is bias in the training data does not mean an AI tool developed with it is unusable. It just means that organizations need to be careful on where and how that AI is applied.
Clinical Validation
To further build trust, SAS is working with partners like the Erasmus Medical Center in the Netherlands. Together, these organizations are publishing their AI and analytics algorithms so that clinicians from Erasumus can validate them.
“They [Erasumus clinicians] validated those models with the same rigor of a clinical trial,” explained Dr. Kearney. “They asked questions. The algorithms were published. People understood them. When they used those models, everyone across their health system trusted the results because it had the rigor behind it.”
Watch the interview with Dr. Stephen Kearney to learn:
- How SAS is helping to make interoperability easier for clients with a library of pre-built data connectors
- Eliminating bias in data is critical to SAS’s AI efforts in healthcare
- Why bridging the data trust gap is as important as implementing the right analytics technology
Learn more about SAS at https://www.sas.com/
Listen and subscribe to the Healthcare IT Today Interviews Podcast to hear all the latest insights from experts in healthcare IT.
And for an exclusive look at our top stories, subscribe to our newsletter.
Tell us what you think. Contact us here or on Twitter at @hcitoday. And if you’re interested in advertising with us, check out our various advertising packages and request our Media Kit.
