Emergency departments (EDs) are the front lines of healthcare, but they are increasingly facing challenges with overcrowding and delays. When a patient arrives at the ED, the very first step is often triage – a crucial process where healthcare professionals rapidly assess patients to identify those with the most urgent needs. This initial evaluation determines the priority of treatment and significantly impacts patient flow and resource allocation within the department.
Alt: Overcrowded emergency department waiting area with patients and limited seating, highlighting the need for efficient triage.
“With the number of emergency department visits rising across the nation, capacity strains have resulted in unprecedented levels of crowding and subsequent delays in necessary medical attention,” explains Scott Levin, Ph.D., an expert in emergency medicine at Johns Hopkins University School of Medicine. “Therefore, emergency departments must swiftly determine which patients require immediate, critical intervention versus those who can safely wait for treatment.”
Currently, many EDs rely on the Emergency Severity Index (ESI) to categorize patients during triage. This system assigns a score ranging from Level 1, for the most critically ill individuals, to Level 5, for those with less urgent conditions. A patient’s ESI level dictates the treatment area within the ED, their position in the waiting queue, and influences clinical decisions throughout their care journey. However, as Levin points out, “This current triage algorithm is inherently subjective. Clinicians make rapid judgments about a patient’s waiting capability based primarily on their own clinical expertise.” Research indicates that a significant proportion of patients are often categorized at Level 3, a broad middle ground that lacks precise differentiation. “We hypothesized that the Level 3 patient group encompasses a wide range of patients, some severely ill and others less so, and we aimed to investigate whether these patients could be more effectively categorized,” Levin states.
To address the limitations of the subjective ESI system and enhance triage accuracy, Dr. Levin and his team at the Department of Emergency Medicine innovated an electronic triage tool. A recent study published in the Annals of Emergency Medicine evaluated this e-triage tool across multiple sites, encompassing nearly 173,000 emergency department visits. The findings revealed that the electronic tool performed as well as, or even better than, the ESI in predicting patient outcomes.
Alt: Doctor using a tablet with electronic triage tool interface in a busy emergency department setting, illustrating technology integration in patient care.
The study demonstrated significant discrepancies in patient priority assignments between e-triage and ESI. Notably, within the over 65% of visits initially triaged at ESI Level 3, the e-triage system identified approximately 10%, or over 14,000 patients, who might have benefited from being escalated to a higher priority level, such as Level 1 or 2. These patients flagged by e-triage were at least five times more likely to experience critical outcomes like death, ICU admission, or emergency surgery, and twice as likely to require hospital admission overall. Conversely, the e-triage tool also effectively identified patients who could be safely down-triaged to lower priority levels (Level 4 or 5). This capability is crucial for reducing wait times for low-acuity patients and optimizing the utilization of limited ED resources.
The electronic triage tool leverages a sophisticated algorithm to forecast patient outcomes. This algorithm employs systems engineering principles and advanced machine learning techniques to discern patterns and correlations between predictive data points and patient outcomes. “When a patient presents at the emergency department, and their information is entered, the e-triage tool analyzes that patient’s data against a vast dataset of comparable patients to predict their potential outcome,” Levin explains.
Alt: Headshot of Scott Levin Ph.D., emergency medicine expert and developer of electronic triage tool, emphasizing expertise and authority.
These data-driven methodologies are commonplace in industries like defense, transportation, and finance, but their adoption in healthcare, especially in emergency care, has been limited. “Machine learning-based approaches fully exploit the wealth of data within electronic health records, enabling a level of outcome prediction accuracy previously unattainable,” states Gabor Kelen, M.D., director of the Department of Emergency Medicine at Johns Hopkins. “This represents the future trajectory of healthcare, although some healthcare providers may initially exhibit reluctance. Clinical decision support tools powered by machine learning are also highly adaptable, allowing customization to align with the specific needs of an emergency department’s patient demographics and local healthcare delivery infrastructure.”
Alt: Professional portrait of Gabor Kelen M.D., Director of Emergency Medicine Department, highlighting his leadership and experience in the field.
Importantly, the e-triage tool is designed to function as a clinical decision support system, empowering clinicians to make more informed decisions about patient care. “The core principle behind this tool, and all effective clinical decision support tools, is that the synergy of the tool and the clinician leads to more accurate predictions and improved prognostic assessments compared to either entity operating in isolation,” Levin emphasizes.
By enabling more precise differentiation of patient priority levels, the electronic triage tool ultimately contributes to ensuring patients receive the appropriate level of care in a timely manner. “The overarching goal is to reduce patient wait times within the emergency department,” Levin concludes. “For patients at risk of requiring critical care, this technology is engineered to enhance their identification and expedite their access to treatment. Conversely, for patients with less acute conditions, e-triage aims to identify and streamline their pathway, minimizing unnecessary delays.”
Currently, e-triage is implemented at The Johns Hopkins Hospital and Howard County General Hospital, both integral parts of Johns Hopkins Medicine. Ongoing prospective evaluations are showing promising preliminary results, suggesting improvements in the detection of patients with critical outcomes and positive impacts on mitigating emergency department congestion. The continued development and refinement of electronic triage tools like this hold significant promise for transforming emergency care and optimizing patient outcomes.
