Understandably, most of the discussion around AI use in healthcare focuses on data analytics for population health management and predictive analytics. Given the massive scale of the data we’re collecting, that’s no surprise.
In fact, one could argue that using AI technologies has gone from an interesting idea to an increasingly established parto the health IT mix. After all, few human beings can truly understand what’s revealed by terabytes of data on their own, even using well-designed dashboards, filters, scripting and what have you. I believe it takes a self-educating AI “persona,” if you will, to glean advanced insights from the eternity of information we have today.
That being said, I believe there’s other compelling uses for AI-fueled technologies for healthcare organizations. If we use even a relatively simple form of interpretive intelligence, we can improve health IT workflows for clinicians.
As clinicians have pointed out over and over, most of what they do with EMRs is repetitive monkey work, varied only by the need to customize small but vital elements of the medical record. Tasks related to that work – such as sending copies of a CT scan to a referring doctor – usually have to be done in another application. (And that’s if they’re lucky. They might be forced to hunt down and mail a DVD disc loaded with the image.)
Then there’s documentation work which, though important enough, has to be done in a way to satisfy payers. I know some practice management systems that integrate with the office EMR auto-populate the patient record with coding and billing information, but my sense is that this type of automation wouldn’t scale within a health system given the data silos that still exist.
What if we used AI to make all of this easier for providers? I’m talking about using a predictive intelligence, integrated with the EMR, that personalizes the way data entry, documentation and follow-up needs are presented. The AI solution could automatically queue up or even execute some of the routine tasks on its own, leaving doctors to focus on the essence of their work. We all know Dr. Z doesn’t really want to chase down that imaging study and mail it to Albany. AI technology could also route patients to testing and scans in the most efficient manner, adjusted for acuity of course.
While AI development has been focused on enterprise issues for some time, it’s already moving beyond the back office into day-to-day care. In fact, always-ahead-of-the-curve Geisinger Health System is already doing a great deal to bring AI and predictive analytics to the bedside.
Geisinger, which has had a full-featured EMR in place since 1996, was struggling to aggregate and manage patient data, largely because its legacy analytics systems couldn’t handle the flood of new data types emerging today.
To address the problem, the system rolled out a unified data architecture which allowed it to integrate current data with its existing data analytics and management tools. This includes a program bringing together all sepsis-vulnerable patient information in one place as they travel through the hospital. The tool uses real-time data to track patients in septic shock, helping doctors to stick to protocols.
As for me, I’d like to see AI tools pushed further. Let’s use them to lessen the administrative burden on overworked physicians, eliminating needless chores and simplifying documentation workflow. And it’s more than time to use AI capabilities to create a personalized, efficient EMR workflow for every clinician.
Think I’m dreaming here? I hope not! Using AI to eliminate physician hassles could be a very big deal.