UPMC is on the forefront of using big data to improve patient outcomes. The system started collecting patient data more than 30 years ago and now has a robust platform to support predictive analytics and better curate care.
So, what's next?
Chris Carmody, senior vice president of UPMC information technology division and chief technology officer of UPMC, joined the Becker's Healthcare Podcast to talk about the system's digital strategy, predictive analytics and how artificial intelligence can make care "smarter and better."
Note: The conversation below is lightly edited for clarity.
Question: What are the opportunities you're most excited about right now with digital technology in healthcare?
Chris Carmody: The number one thing that's really driving us is our clinical analytics program. It's an area I got involved with about six years ago with our chief analytics officer Dr. Oscar Marroquin. He's a wonderful human being and one of the great thought leaders at UPMC. He's a cardiologist by training and so passionate about delivering life-changing medicine. We've built a program that has about 1,300 predictive models based upon a lot of digital data that we've been able to produce over 32 years; that's one of our larger database platforms and goes back to 1991 when it was first conceived and built.
We leverage that data today and those various models that have really created the UPMC DNA in how we deliver care and provide better insights to our care teams and clinicians who are engaging with patients and our health plan members. In the analytics space, technology has evolved a lot. Within the last five or six years, we've bursted the compute and processing capabilities out into the cloud and are in the process of re-platforming the different data marts and models that we've built to help us go exponentially to a new scale and further leverage technologies like machine learning, which is what you need before jumping into artificial intelligence.
Q: How are you applying artificial intelligence and predictive analytics to clinical care today?
CC: We've used artificial intelligence to extract out useful data elements from a lot of the unstructured data that helps inform those predictive models. Any time a patient is scheduled for a surgical procedure at UPMC, there's close to 2 million of those per year, the patient's information is run through our predictive modeling to help identify any risk factors that we need to take preventative action on before we bring the in for a surgical procedure.The goal being the best possible outcome for every individual patient coming in. We use those models to flag certain areas of concern and that allows clinicians to focus on addressing some of those concerns before the procedure.
Similarly on the back end, when you have a patient that's being discharged from one of our facilities, we run through our predictive modeling from a readmissions perspective and identify how we can help them get back home and stay on course to their path of recovery.
Q: Where is the technology headed?
CC: We are working on improving upon and having access to some of the data that's not apparent in your traditional electronic health record like symptoms and social determinants of health. We've been using an AI model from Microsoft called Text Analytics to help identify those nontraditional unstructured data that we've generated over all these years. A lot of the discharge summaries and other reports we scan through are scoured for that information to help patients put the right support system in place when they leave our four walls at UPMC.
We've seen a drastic improvement across the board in reducing readmission rates. That's really what it's all about. It's using technology, data and the combination of that with our clinical expertise to continually create this learning environment that we get smarter and better, and provide those insights to our clinicians with the ultimate goal for every human being that we touch: that we create the best possible outcome.
Q: What are your thoughts on ChatGPT? Has UPMC started using large language models in a meaningful way?
CC: I'm sure everyone's abuzz around ChatGPT and OpenAI, and the large language models. We are absolutely exploring those technologies in a safe, secure and ethical way as we move forward to see how we can best utilize them. There's many more initiatives and products we have going on at UPMC, but the core foundation, from a cybersecurity perspective, is healthcare has a large bullet on our back and we always have to be smart.
We have to continually be grounded with good IT hygiene where we apply the best practices and controls, and safeguards to protect our patient and our members' information and data. That's always at the forefront of everything we do here at UPMC to move the needle. That's where our focus is right now.