The healthcare industry is facing a wide variety of challenges—and solutions aren’t always straightforward. Each month, Modern Healthcare asks leaders in the field to weigh in on their approaches to the sector’s thorny issues.
Dr. Jeffrey Hoffman, chief medical information officer at Nationwide Children’s Hospital, and Dr. Michael Oppenheim, senior vice president of clinical digital solutions at Northwell Health, discuss how patient data can be used most effectively.
What’s a recent development regarding data or technology that you’re really excited about?
Hoffman: Generative artificial intelligence tools like ChatGPT get all the attention right now, and rightly so. A powerful tool just now coming into its own is applying these advanced analytical techniques to create predictive algorithms. Predictive models are everywhere these days. Unfortunately, at least in the clinical space, off-the-shelf predictive models—even those from large vendors—haven’t always lived up to their promise.
Oppenheim: Even with technologies that are totally not AI, at the end of the day, AI enables almost everything. One example is the growing use of biosensors for remote patient monitoring. … When you look at biosensors and almost every technology, the data that come out often have an unacceptable noise-to-signal ratio. So you need a layer of intelligence on top to help make sense of it all.
Electronic health records and other technologies have created large sets of patient data. Where has that information made a major impact?
Hoffman: There’s certainly a huge amount of data. We have discrete data as well as information derived from unstructured data like clinical notes. One of the big impacts over the past several years, as a result of having all these data at our fingertips, is really the instantaneous and relatively seamless access clinicians now have to a patient’s full medical story, no matter where the care was delivered.
Not a Modern Healthcare subscriber? Sign up today.
Oppenheim: Everybody’s looking to use various types of predictive models. Data have mostly benefited the research side, which has then translated to the clinical side. I don’t know that having this mass of data has helped me at the bedside directly as much as it has provided the fodder for scientists, researchers and predictive modelers to build models and algorithms that we can then apply to treatment.
How would you say these available data are being underutilized?
Hoffman: We have so much access to data these days. Whether they're clinical or operational, it’s really hard at times to separate the signal from the noise and really figure out what’s meaningful. It takes a skilled set of individuals or significant investments in technology to turn those data into actionable information, and to present it to the decision-makers in an understandable, digestible way.
Oppenheim: Organizing the relevant information to have all the key pieces in front of me will point me in a better direction toward diagnosis. But if I’m spending all my time hunting and gathering and not synthesizing and integrating, then I’m not getting value. So it’s not one or two pieces of data or one particular thing, but the ability to pull out everything, organize it and present it.
