‘Virtuous cycle’ enables HCA to make improvements in clinical care

Register now

When trying to illustrate how data streams into and out of electronic health record systems, health IT leaders typically end up with a diagram full of arrows pointing to and from various devices, interfaces, data warehouses and other technology pieces.

By contrast, Nashville-based HCA visualizes the data flow as a continuous circle.

“It begins and ends with the patient,” says Jonathan Perlin, MD, the company’s chief medical officer and president of clinical services. “Data generated as part of patient care is categorized as information in the clinical data warehouse, processed into knowledge, and then brought right back to patients where we can apply that knowledge to provide more precise, more personalized, and more compassionate care.”

Perlin is participating in a roundtable dialogue on “Architecting for 21st Century Health Systems” at the HTLH: The Future of Healthcare conference this week in Las Vegas. He’s emphasizing the importance of using data generated during patient encounters to inform and improve future care. Perlin sees this continuous process as a type of perpetual motion machine.

How does the concept of a perpetual motion machine apply to health technology and performance improvement?

Perlin: Since the 13th century, there’s been this notion of a perfect machine that never loses energy and, in fact, can create more energy from the energy that is put into it. The truth is that physics will never allow the creation of that sort of machine.

But, in healthcare today, the data generated through patient care presents an opportunity that is metaphorically like a perpetual motion machine. We have this virtuous cycle where the knowledge created as a byproduct of care helps to inform the care of that particular patient as well as the collective needs of all patients. [This analogy of] the perpetual motion machine is about closing the loop and not letting data dissipate as heat, or lost energy, but rather bringing it back as light to illuminate the care of patients.

Can you give an example of how data science is driving care improvement at HCA?

Perlin: We’ve turned back the clock on sepsis detection by 18 hours, using an algorithm that continuously collects and analyzes patient vital signs, labs, nursing surveillance reports and other data. It detects early sepsis with 100 percent sensitivity and 50 percent more specificity, compared with our clinicians. This is huge when you consider that mortality increases 4 to 7 percent for every hour that a sepsis diagnosis is delayed.

When we add prediction to the model, we’re able to push identification back another 5 hours —at the same level of sensitivity but only 20 percent better specificity. Right now, we’re focusing on detection because we need our clinicians to trust the model. But we’re pushing toward being able to predict sepsis and giving clinicians almost a day lead time.

How does the sepsis model work exactly?

Perlin: You need big data to do this. At HCA, we are privileged to have more than 30 million patient encounters every year. Data on tens of thousands of patients who develop sepsis are fed into the machine learning model so it can hone its understanding of the relationship among the different pieces of data and come up with the most sensitive and specific algorithm to detect and, ultimately, predict sepsis.

The decision model started with traditional sepsis variables and, in the process, found a few triggers that allow us to have a better understanding of when sepsis can occur. It’s not just about recognizing that blood pressure is decreasing or temperature is dropping, it’s about seeing nuances in the rates of decline and the relationships among the different variables.

There’s no clinician in the world who can spend 24/7 looking at every new data element at the moment it’s created and thinking about the trajectories of different data elements in conjunction to each other — except for one, and that’s the computer.

For reprint and licensing requests for this article, click here.