In part one of this installment of our Chief AI Officers in Healthcare series, Dr. Zafar Chaudry, chief digital officer and chief AI and information officer at Seattle Children’s, discussed his journey to becoming a CAIO and described the background and expertise most suited for the role.
Today Chaudry will discuss exciting AI implementations at Seattle Children’s and the huge results already being achieved by the clinical teams. And some tips for IT leaders looking to become Chief AI Officers themselves.
Q. Please talk at a high level about where and how Seattle Children’s is using artificial intelligence today.
A. We’ve been using forms of artificial intelligence for a couple of years now. The difference we now have is we’ve moved from internal data and language models to much larger language models. We use AI in multiple ways. I’ll give you an example.
We are using artificial intelligence to determine how we can put patients to sleep. One of the problems with kids is they can get addicted to opioids. You traditionally use opioids to put them to sleep. Then when you send them away from the hospital with pain medication, that pain medication can have opioids in it.
The risk is there can be addiction. We looked at our data and worked with our clinicians to ask whether there’s another way? Can we find another cocktail of medication that isn’t opioid-based that can manage pain but also put patients to sleep? I can tell you, using the AI tools, we’ve been very successful in that space. 100% of our outpatient surgeries are now opioid-free. 50% of our inpatient surgeries are opioid-free.
And so we’re a leader in that particular space, as an example. Another example I can give you is our neurosurgeons have spent some time figuring out if you do brain surgery on a child, when the child comes to the ICU, there’s a high risk of the child having a stroke. And if a child has a stroke, then the outcome is not good, as you can imagine, for survival.
And so we used, again, our data, our algorithms. I have about five data scientists on my team. They write the algorithms. They work with, in this case, the neurosurgery team. We’ve been able to take all the data – if you imagine a child in ICU and all these monitors plugged into the child, there’s so much data being collected on that child.
By using the data, creating algorithms and applying some AI, we have a very high prediction rate before the stroke happens so we can intervene and the child doesn’t have a stroke. Those are some clinical use cases as to how AI can truly help have better outcomes. That’s really what we’re looking for.
Q. How do you as the Chief AI Officer oversee AI work?
A. So, in both of those examples, there’s a few steps. My job is to make sure we are collecting the data correctly. Is the data in the right format? Is the data being stored in a safe environment? Am I providing the infrastructure, the computing horsepower, to actually translate that data and analyze that data?
There’s an infrastructure component my team and I are responsible for. There is the actual writing of the algorithms, which sits within my analytics team. That’s why I have data scientists. There’s the security component, which sits within my security team. It’s got to be secure, it’s got to be safe. It’s got to be clean because bad data doesn’t help any algorithm or any AI.
So, my responsibility is to make sure those pieces work. My other responsibility is to make sure I’m working with the clinicians to make sure the funds are correct. There’s a budgetary component to all of this. But beyond that, my job as an AI chief is to facilitate the problems the clinicians need to solve.
Because you first need to know the problem you want to solve, and then it’s my team that needs to figure out what the pieces are in that puzzle, put those pieces together, which then translates into better outcomes.
So, I spend my day making sure those things happen. I also am the liaison between the clinical teams because I’m also a clinician, I can translate a lot of the hardcore technical speak into what makes sense for what you’re actually describing.
Because truth be told, physicians and nurses don’t really care what the technology is. They just want to consume it, and they don’t want to hear about the complexities of it. One thing we’re not good at in the technology realm is trying to explain what we do at an eighth-grade level.
My job also is translation between the teams to make sure everybody is on the right page. I always tell my team that being in the technology space, 20% of our job is sales because sometimes you have to go and convince people the technology will do what they need it to do, but they need to put some time into it because my team puts time into this, and it would never be successful.
AI won’t be successful if you don’t have protected time from the clinicians that you’re trying to help. If they don’t put time into this project and their subject-matter expertise, we won’t be successful.
Because you can build an algorithm, but they’re the ones who have to validate whether the algorithm is telling the story correctly or incorrectly. That’s why when you’re using AI, we always tell people to double-check what it’s saying because it might not be correct, it might hallucinate. I can’t do that without the subject-matter experts. My job is multifaceted in this space.
Q. What are a couple of tips you can offer to other IT executives looking to become a Chief AI Officer at a hospital or health system?
A. Start with the basics. If you don’t understand the newest technologies in the AI space, spend some time studying those areas because it’s a very fast-moving environment. Think about how you will be able to create an infrastructure or a partnership with a technology vendor where you can actually afford to do this.
So, think about the business metrics and spend time heavily with the people you serve, the clinicians, the doctors, the nurses, allied health professionals, because you are going to have to be the bridge to collaborate to actually make this all work.
Take time doing these three pieces. You may already have one or two of those skill sets, so just top up the rest, and then you should be in good shape. Let’s be truthful, it’s not rocket science. It’s a people business we are in. If you can win people over – influencing skills are really important. If you can do that, then the technology pieces will fall into place.
The good news in the AI space is there are some very solid partners that can help you deliver the technology. The other thing to think about is, Do you have the skill sets around you on your own team?
It’s not a standard analytics person who will lead in the AI space. There are some new skills. Maybe take some time as well to get existing teams you have and provide them additional training. Prompt engineering in AI isn’t intuitive because we don’t ask ourselves questions the same way as human beings.
I spent a good year learning prompt engineering. It was awfully confusing at first. You have to be very specific. I would say get some training in that arena or get access to an AI in your environment and spend a lot of your personal time testing it, asking it, so you understand what the key words are to ask to get to where you want to be.
Editor’s Note: To get BONUS CONTENT not found in this story, click on the video below.
Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication
WATCH NOW: The Chief AI Officer showcases the technology at Seattle Children’s