Dr. Hojjat Salmasian is chief data and analytics officer at Children's Hospital of Philadelphia, also known as CHOP. While he does not serve as a chief artificial intelligence officer – probably the hottest new position in health IT – he does co-lead the health system's AI strategy and serves as the co-chair of its AI governance committee.
Salmasian is a primary care physician by background and an assistant professor of medicine and biomedical informatics. His research focuses on healthcare quality and patient safety, and the role of health IT and artificial intelligence in those areas. He has coauthored more than 130 scientific publications.
All of this is how he came to be CHOP's AI leader.
Expanding a role into AI
"CHOP has a very strong data and analytics program," Salmasian said. "The person who was leading that effort as the vice president of data analytics is now my boss as the chief digital and informatics officer. When he transitioned into his new role, knowing the strong foundation of the analytics program we had, he was looking for a leader to take this role who not only could continue those strengths, but importantly, expand them specifically into the areas of data science and artificial intelligence.
"I was in a role prior to coming to CHOP that was focused exactly on using data science for helping that hospital in its mission," he continued. "It was a very natural fit, both in terms of what was the next step for CHOP's data analytics and AI journey as well as for me as a person who was an expert in that area and looking to broaden his impact into other areas."
CHOP was looking for an executive with a combination of the technical skills and the leadership skills required for leading the data analytics and AI change at the organization. It also wanted an executive with background as a clinician and a researcher. All of this to understand what it would take for CHOP to create platforms that vigorously enable the use of data and AI across the board, from research to clinical care operations to education.
"Because my background is in all three of those categories of leadership – data analytics, clinician, researcher – it helps me be that bridge role that is much needed at our organization," Salmasian noted. "And that is the role I now play."
The skills required to lead AI
Anyone looking to become an AI leader at a hospital or health system should have a particular set of skills, Salmasian explained.
"It's important to have a strong foundation in data and AI because a lot of the technologies in the space are still relatively new and pretty nuanced," he advised. "For you to be able to understand when to use what tool or how to enable its use and impact across your organization, you have to have a decent amount of familiarity with the conceptual frameworks, the technologies, the dependencies therein, and so forth.
"And you need to be a very strong leader in terms of matrix leadership and symbolic leadership," he continued. "AI, especially at academic medical centers, is not a single person's charge. There's typically a person who is nominally the leader for AI for the organization, but truly informatics, application teams, cloud teams, analytics teams, all of them, have to work together to support the clinical leaders and operational leaders who are the real people who actually bring the AI to their strategy and harness the value of it in the care of the patients and the efficiency of the work of the workforce."
An AI leader must be able to effectively lead through symbolic leadership and matrix leadership the work that teams are doing in an organization – that's a recipe for success for anybody who wants to be an AI leader at a health system, he added.
"Symbolic leadership means you need to symbolize through role modeling, through inspiration of people who don't even report up to you, what the future state is going to look like," he explained. "You need to be that symbolic leader who people look up to and say, 'He's not my boss, or my boss' boss, but what he says makes sense to me. I'm going to change my behavior because of the vision he's laying out.'
"Top-down leadership is not going to get you far in the AI space because most of the work in AI doesn't happen by a single team or a single department," he continued. "It is really across areas. You have to be able to influence the clinical leader and the clinical workforce on the frontline in how they are choosing to change their behavior and use AI in their work."
Wide-ranging influence
Further, an AI leader must be able to influence the technology and cloud workers to prioritize the work in a way that enables harnessing the value of AI on the frontline. That is never a single person or a single department's vision; it's across the board, he added.
"So obviously, there is an AI leader who sets the strategy and brings all the operating models together and makes sure the departments are working together well," he said. "But a lot of that is because those successful leaders in AI typically are very strong symbolic leaders."
Part of Salmasian's job at CHOP is to lead AI governance. He is the chair of the AI governance committee. As such, he is responsible, along with the partners on that committee, to develop the guidance and principles that ensure staff are using AI responsibly and effectively.
"I'm also leading the development of our AI strategy," he said. "I'm putting together what is the structure through which we want to identify and prioritize AI use cases and initiatives. What are the resources within my team and other teams we need to effectively evaluate and deploy AI and harness its value? What are the operating model changes we need to do across the organization to effectively use AI?
"Also, my team leads a lot of the AI literacy training of the staff," he added.
Day-to-day activities
Salmasian's day-to-day job is a mixture of strategic work, as well as tactical work. The strategic work usually focuses on how CHOP aligns across teams, the desired structure, the vision to pursue, and how the organization stays aligned with that vision as it meets challenges both within and outside of AI work.
"The tactical work is really about bringing together the components of a successful AI journey for us, which includes governance," he said. "So, I attend those meetings and participate in those decisions, which includes road-mapping and operating model work, which includes our work on literacy. Again, that's a component in which I contribute a significant amount.
"Then, there's platform work – deploying new platforms that allow for AI to become more of a democratically accessible tool for the organization," he continued. "But no two days are alike in my job, and it's filled with a lot of exciting challenges I'm happy to be able to help and solve. And that's what makes me a really proud member of CHOP, is to be able to constantly contribute toward our vision and mission."
CHOP uses artificial intelligence today in many ways.
"We use it in our clinical space, including, for example, a number of predictive models or generative AI models integrated into our electronic health record system," Salmasian explained. "We use it in our operations. For example, we have predictive models focused on processes or outcomes of the revenue cycle process.
"We use it in research," he continued. "For example, we have developed AI-enabled tools that support researchers in some of the processes they use. Also, we have enabled AI tools for the researchers to use for research into AI."
A variety of AI tools
CHOP also has AI tools deployed for labor productivity and general-purpose use. For example, it has what is called CHOP GPT, a tool designed to be a safe, secure and privacy-compliant alternative to public large language models like ChatGPT. Users at CHOP can go to CHOP GPT and benefit from the power of LLMs without any unwanted data exposure to a third-party vendor.
"We also have a lot of work that is more on the foundational level," Salmasian explained. "We have developed and implemented tools that make sure our data is AI-ready, as well as platforms that allow for AI development to be conducted efficiently. So new tool development using AI is supported through those activities."
One example of AI in use today at CHOP that Salmasian is very happy with is a feature in the EHR that allows for clinical teams to respond faster and in more detail to patient inquiries sent through the patient portal.
"We were among the first health systems to work with that company and try it in pediatrics, identify some of the gaps in the initial design that would make it difficult to be effected in pediatrics, and then be the first organization to actually successfully demonstrate value for it in pediatrics and deploy for all of our users," he noted.
"26,000 times, so far, we have responded to patient messages using AI," he continued. "There's always a human in the loop that reviews the content and decides if they want to use it, edit it or start from scratch. And what we're seeing is some of our users are saving between 20 and 40 seconds per message using this tool."
This is saving millions of seconds every month for users, which translates to a significant amount of time saved that can be reallocated to other more important activities – so CHOP has been able to demonstrate a significant amount of value from this AI tool, he concluded.
Click here to watch a brief video of bonus content where Salmasian discusses how healthcare artificial intelligence strategy at hospitals and health systems is a big challenge as the AI market and tech explode around providers. He offers advice for creating and driving successful strategy.
Follow Bill's health IT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.
WATCH NOW: How can an IT exec become a chief AI officer – and then work with the C-suite
