Electronic Triage Tool Improves ED Patient Care

Electronic triage (e-triage) systems classify emergency department (ED) patient’s injury severity more accurately and supports triage decision-making, claims a new study.

Researchers at Johns Hopkins University (JHU, Baltimore, MD, USA), StoCastic (Baltimore, MD, USA), and other institutions conducted a multi-site, retrospective, cross-sectional study of 172,726 visits to urban and community EDs in order to evaluate if an e-triage system composed of a random forest model of triage data--vital signs, chief complaint, and active medical history--could predict the need for critical care, an emergency procedure, and inpatient hospitalization in parallel, and translate the risk to triage level designations.

The predicted outcomes and secondary outcomes of the study were elevated levels of troponin and lactate, which were evaluated and compared with the standard Emergency Severity Index (ESI). The results showed significant differences in patient priority levels between e-triage and ESI evaluations. For example, out of the more than 65% of ED visits triaged to ESI Level 3, e-triage identified about 10% of patients who may have benefitted from being up-triaged to a more critical priority level, such as level 1 or 2.

The up-triaged ESI level 3 patients were at least five times more likely to experience a critical outcome--such as death, admission to the ICU, or emergency surgery--and twice as likely to be admitted to the hospital. The e-triage tool was also able to increase the number of patients down-triaged to a lower priority level, such as Level 4 or 5, helping to minimize low-acuity patients from waiting and overusing scarce resources. The study was published on September 6, 2017, in Annals of Emergency Medicine.

“The ultimate objective is that patients should be waiting less in the emergency department,” said lead author Scott Levin, PhD, of JHU. “For patients at risk of having a critical care need, this technology is designed to detect them better and make sure they are seen quicker. For patients who are less sick, e-triage should detect those patients and put them on an expedited track, so they don’t need to wait as long.”

“Machine-based learning takes full advantage of electronic health records and allows a precision of outcomes not previously realizable,” concluded senior author Professor Gabor Kelen, MD, director of the JHU department of emergency 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.”

Random forests are an ensemble learning method for classification, regression, and other tasks that operate by constructing a multitude of decision trees at training time and outputting the class that is the mode of the classes (classification) or mean prediction (regression) of the individual trees.

Related Links:
Johns Hopkins University
StoCastic