NUMBER TWO: MORE TIMELY TRIAGE
When patients walk into an emergency department, the staff’s first responsibility is to do a rapid triage assessment: Is this person at immediate risk of dying? Is the patient actually so well that he should have just stayed home and taken an aspirin? Or somewhere in between?
Scott Levin, an associate professor of emergency medicine at the School of Medicine who holds a joint appointment in Civil Engineering and directs the Systems Institute at the Whiting School, believes far too many emergency-room patients are classified in a vague in-between zone. If hospitals developed more accurate predictions at the time of triage, he says, they would be better able to allocate their resources—and ultimately save more lives.
In typical clinical practice, more than half of emergency patients are classified in the middle level—3—of the five-level Emergency Severity Index. “That’s a problem,” Levin says, “because it doesn’t set up an appropriate queue. If you look at what actually happens to these patients, it turns out that the ‘3’ level mixes patients who are very sick with patients who are very well.”
Over the last two years, Levin and his colleagues have developed a new triage algorithm known as the HopScore. To build the system, Levin analyzed hundreds of thousands of past patients’ electronic records across multiple emergency departments. His goal was to find a mix of clinical indicators that retrospectively “predicted” which patients ultimately experienced a critical event (death, intensive care admission, emergent surgery, or catheterization) or required hospital admission.
Now Levin and his team think they have found a simple algorithm that outperforms the predictive power of classical triage systems. The significant components turned out to include the patient’s chief complaint, vital signs, age, gender, and whether the patient arrived at the ER by ambulance, all routinely collected at triage in emergency departments. The HopScore is now being calculated for each patient who arrives at the Johns Hopkins Hospital or Howard County General Hospital Emergency Department—but it isn’t yet affecting clinical practice.
“If it turns out that these scores have good predictive value, and we very much expect that they will,” Levin says, “then we’ll begin to use the HopScore to make rapid decisions about where to place patients and how to allocate staff resources.”
This kind of collaboration between engineers and clinicians is crucial to the future of medicine, Levin believes. When he earned his PhD in biomedical engineering roughly 15 years ago, he was supported by a National Science Foundation grant that required deep interdisciplinary study. “This is how I was trained, and this is how I love to work,” he says. “We shouldn’t be separated by method or by subject area—we should work together around real-world objectives like patient safety.”