On the HIMSS AI in Healthcare Discussion board in San Diego subsequent week, one analytics chief will talk about the bogus intelligence use instances he is discovered most success with – and can provide suggestions and perspective on greatest practices for AI deployments in scientific settings.
Dr. Luis Ahumada, director of well being knowledge science and analytics at John Hopkins All Youngsters’s Hospital – talking on a panel with Sumit Nagpal, CEO of Cherish Well being – will talk about his expertise as far as an early adopter of AI and machine studying fashions.
He’ll provide some real-world examples of how Johns Hopkins is utilizing AI-enabled instruments to enhance its care-delivery processes – automating and expediting scientific diagnostics and liberating up clinicians to focus extra on affected person care. He’ll describe overcoming obstacles to AI adoption, greatest practices for integrating it into current workflows and far more.
“AI has been serving to,” says Ahumada, nevertheless it’s “an ongoing challenge.”
At John Hopkins, “we’re making an attempt to know extra the place the principle advantages will likely be,” he stated, as a result of “sources aren’t accessible to everybody and AI is dear.”
Ahumada sees worth in two sorts of AI, with totally different traits of dimension, form and scope.
“One is LLMs – however they are not going to unravel every part,” he stated. “The opposite one is what we name conventional machine studying: creating the fashions for prediction and high-risk calculators, issues like that. There will likely be a hybrid in some unspecified time in the future between the 2. However for the previous 10 or 15 years, we have been specializing in the second.”
Each are necessary for one use case John Hopkins All Youngsters’s Hospital has been keenly centered on: scientific documentation.
“We’ve high-risk calculators for every part below the solar: readmissions or dangers for surgical procedure, problems, issues like that. However on the identical time, loads of these issues are attributable to inefficiencies throughout the system, within the documentation course of. And for my part, we have to clear up that first. And that is the place the present instruments that we’ve at our disposal are most likely aimed for.”
Knowledge integrity is the “keystone and the bedrock for every part that we do with LLMs and ML.”
Dr. Luis Ahumada, John Hopkins All Youngsters’s Hospital
Towards that objective, the well being system has been centered on “getting higher knowledge,” he stated. “We are able to use machine studying, we are able to use LLMs, we are able to use loads of various things to gather this knowledge higher.”
Sure, there are additionally large challenges round lacking knowledge, validation, knowledge integrity, says Ahumada. “Knowledge will not be excellent. It must be, however it’s not. That is an issue as a result of we use that knowledge that we acquire every single day, each second, to create fashions.”
One other fundamental hurdle has to do with “amassing and placing all of that knowledge collectively,” he stated. “Conventional machine studying loves enormous volumes of information.” However it could actually usually be a problem to “put collectively even a small registry for a whole lot of hundreds of sufferers,” he defined.
However regardless of these challenges, Johns Hopkins is pushing ahead with an array of various generative AI and machine studying use instances that Ahumada will describe in additional element in San Diego.
Knowledge integrity is the “keystone and the bedrock for every part that we do with LLMs and ML,” stated Ahumada.
However he is additionally involved additionally about price, and about serving to be sure that well being techniques of any dimension – not simply these with the sources of Johns Hopkins – have entry to the sorts of AI instruments that may assist them enhance their scientific and administrative processes.
Simply this week, a report from KLAS was revealed displaying that enormous well being techniques are already making hay with generative AI however smaller ones have been extra restricted of their embrace.
“AI is meant to be open to everybody, nevertheless it’s not true as a result of it’s costly,” he stated. “You should utilize superior LLMs, however even with that, you possibly can arrange an LLM in your store [but] individuals want to know that to have the ability to try this, it’s important to have people who know methods to do it, and so they’re not cheap. So sure, there are loads of totally different advantages. However it’ll price cash.”
Attend this session on the HIMSS AI in Healthcare Discussion board happening on December 14-15, 2023, in San Diego, Calif. Be taught extra and register right here.