At hospitals throughout the United States the emergency severity index (ESI) is used during triage to assign a score from level 1 for patients who are the most critically sick, to level 5 for patients who are the least sick. This index can be viewed as subjective and may not always improve patient outcomes.
Researchers at Johns Hopkins University School of Medicine have developed an electronic triage system to improve this process. The tool offers scope for better identifying critically ill patients and assigning priority treatment levels.
“Nurses and physicians make a quick assessment on whether the patient can wait solely based on their clinical judgment,” says Scott Levin, Ph.D., associate professor of emergency medicine. In most cases, researchers say patients are assigned to a level 3 and not entirely differentiated.
“We thought that level 3 patient group included a large mix of patients who are pretty sick and others who weren’t, and our goal was to determine whether these patients could be sorted out.”
The e-triage system demonstrated equal or improved identification of patient outcomes compared to ESI based on a multi-site retrospective study of nearly 173,000 emergency department visits. Of the more than 65 percent of visits triaged to ESI level 3, e-triage identified about 10 percent, or more than 14,000, ESI Level 3 patients who may have benefitted from being up-triaged to a more critical priority level, such as Level 1 or 2. These patients were at least five times more likely to experience a critical outcome, such as death, admission to the ICU or emergency surgery and two times more likely to be admitted to the hospital. The number of patients down-triaged to a lower priority level was also increased by the new system, which helps minimize low-acuity patients from waiting and overusing scarce resources.
An algorithm predicts patient outcomes based on a systems engineering approach and advanced machine learning methods to identify relationships between predictive data and patient outcomes.
“When a patient comes in, and we collect the patient’s information, the e-triage tool is comparing that patient to hundreds of other like patients to make a prediction on the patient’s outcome,” Levin says.
These methods are common in other industries, such as defense, transportation and finance, but rarely, if ever, are those implemented in health care.
“Machine-based learning takes full advantage of electronic health records and allows a precision of outcomes not previously realizable,” says Gabor Kelen, M.D., director of the department of emergency medicine and professor of emergency medicine at the Johns Hopkins University School of Medicine. “It is the wave of future health care, although some providers may be hesitant. Decision aids that take advantage of machine-learning are also highly customizable to meet the needs of an emergency department’s patient population and local health care delivery systems.”