We’ve Got You Covered!
Artificial Intelligence FAQ for your healthcare site with Dr. Kathryn Todd and Arun Dixit
When you think of Artificial Intelligence (AI), scenes from movies like The Terminator may come to mind – and it may be scary to imagine the idea of robots running your healthcare site.
As our guests in this episode of Talk with HIROC explain, it’s not that scary. AI isn’t meant to take away jobs or replace humans entirely. AI, when implemented properly, can instead open a world of possibility for your healthcare site.
Today’s guests: Dr. Kathryn Todd - Vice President of Clinical Excellence with Alberta Health Services, and Arun Dixit, Digital and Innovation Strategist at HIROC.
Don’t forget – HIROC can help you with this. For more information about AI in your healthcare site, visit HIROC.com/resources and check out our new guide, Artificial Intelligence: Risk Management in Healthcare.
WHAT IS AI?
Before we get to the do’s and don’ts – we need to know what AI is.
Dr. Kathryn Todd borrows her definition from American Cardiologist Dr. Eric Topol, who wrote a popular book on AI
Paraphrasing Dr. Topol, Todd says AI is:
“The science of creating intelligent machines that have the ability to achieve goals in a similar way that humans would, using a whole constellation of technology.”
Dr. Eric Topol, American cardiologist and author of Deep Medicine: How Artificial Intelligence Can Make Healthcare Human Again
In short, artificial intelligence is essentially computers that train themselves, learning like humans do. These computers analyze massive amounts of data – like the human brain does – and condense it into output that can be useful to us.
“In the near future, AI is going to be an integral part of how we move forward in healthcare,” says Dr. Todd.
WHAT CAN IT DO FOR YOUR SITE?
While AI is still new, it has potential for:
- Supporting decisions made by clinicians, providers, and even policymakers
- Imaging recognition. Radiologists, or those working with skin lesions, etc. could benefit from this. Some studies have shown AI can recognize images more accurately than humans, and there’s potential for AI to help detect cancer in some scans.
- Serving patients directly. Chat bots, etc. This could be helpful when there is some sort of barrier to communication.
Dr. Kathryn Todd and Arun Dixit agree there’s even more potential applications for AI when thinking about logistics. These include:
- Staff scheduling
- Deploying resources like ambulances and aircraft
- Monitoring privacy breaches – detecting when a patient file is accessed, and determining if it’s suspicious or routine.
WHAT ARE THE RISKS?
AI has some real promise – but as our guests mentioned, it’s not perfect. Risks include:
The last one is especially concerning for Dr. Todd. She says due to the sheer amount of data AI has to process for it to teach itself, a lot of which being sensitive patient information, there’s a risk of a breach.
“This puts us in a bit of a predicament,” she says. “We have to work through that, and I think that there’s a couple promising solutions.”
Another risk: AI, like the human brain it mimics, can make mistakes.
“I’ve seen one study where AI analyzing an images was thrown off by a single pixel that was changed,” says Arun Dixit.
In Alberta, where Dr. Todd works, she’s noticed two promising solutions are coming together to keep patient data safe for use by AI:
A “data haven”. Alberta is looking to create a safe haven for data, which allows for patient information to be stored securely in one place.
Synthetic data. Since AI uses massive quantities of patient data looking for trends and patterns, there is not always a need for the data to be that of real patients.
“This is data that mimics the real data, but is not specific to any individual patient,” says Dr. Todd.
The best way to manage risk in AI? Understand its flaws.
“There’s often a lot of ‘hype’ around the types of output AI systems can produce,” says Arun Dixit. “But the truth is, it’s not perfect.”
He says it should be viewed as a supplementary technology – not a replacement for the human mind.
“AI should really be thought of as a tool that can help us answer questions,” he said. “It should be used in collaboration with a wide range of stakeholders: clinicians, staff, and of course patients and families as well.”
Importantly, he stresses it’s not meant to replace jobs in healthcare – but to take away some of the menial tasks that could help free up even more time.
Dr. Todd agrees.
“I think we can demonstrate that that’s what it can do. It can take over some of the very routine tasks, and allow great minds of humans to do what those great minds need to do,” she said. “And keep the mundane, repetitive tasks allocated to machine learning algorithms. It frees up a space in a human’s mind.”
So on top of understanding the best way to use AI – what are keys to a good AI system in healthcare? It comes down to four pillars:
Another consideration Dixit warns us of: alarm fatigue. He says there are two easy questions we can ask ourselves to mitigate this:
- How does this fit into existing workflows?
- Are we adding another alarm signal into an already challenging workplace?
He says: “the question is not ‘can we build an early warning system’, it’s ‘how does an effective early warning system fit into existing clinical workflows appropriately and safely?’”
In short, AI won’t be much use if all it does is create a new alarm sound that could confuse or overwhelm staff.
Dixit says at the end of the day, AI is being developed with creating better healthcare systems in mind.
“My belief is that AI-based research is done with the end goal of improving outcomes,” he says. ”Improving processes and increasing safety.”
If you would like any more information about Artificial Intelligence in healthcare, please visit HIROC.com/resources and view our guide to AI It was created specifically to help support leaders in healthcare, and it’s a free resource that anyone can access.
By Julian Abraham, Communications and Marketing Associate, HIROC