Mistriaged – How Good Is ESI?

May 15, 2023

Written by Jason Lesnick

Spoon Feed
The Emergency Severity Index (ESI) mistriaged roughly 1 of every 3 patients who presented to 21 EDs in the Kaiser Permanente health system. 

Why does this matter?
ESI is used by >70% of EDs in the country to triage our patients. It is important for us to know how well ESI performs, how often mistriage occurs, and what characteristics are associated with mistriage.  With this knowledge, we can develop ways to limit undertriage, better optimize resource allocation, and promote equity. 

Just how good is ESI at estimating severity of illness?
This retrospective cohort study of 21 EDs within the Kaiser Permanente Northern California system looked at 5,315,176 adult patient encounters from January 2016 until December 2020. 

First, we will review how ESI works. The ESI assigns a value I-V with I being highest acuity and V being lowest acuity. The algorithm is pictured below. 

The authors developed definitions to measure and classify under- and overtriage to assess triage accuracy of ESI in a large, diverse adult patient population and created an algorithm that would review patient visit data to determine whether patient visits were triaged correctly or not.  

The definitions are summarized in the below figure.

The authors found 3,262,047 encounters (61.4%) were assigned a midlevel triage category (ESI III), while 33,491 (0.6%) were ESI I, 929 555 (18.1%) were ESI II, 1,046,806 (19.7%) were ESI IV, and 43,277 (0.8%) were ESI V. Applying study-developed definitions, they estimate that mistriage occurred in 1,713,260 ED encounters (32.2%). Undertriage and overtriage occurred in 176,131 (3.3%) and 1,537,129 (28.9%) cases, respectively.

The authors found that the sensitivity of ESI for high-acuity patients (correctly assigned ESI I or II patients who had a life-stabilizing intervention) was 65.9% and specificity was 83.4%. As far as correctly assigning ESI IV or V to patients who used <2 resources and had no critical interventions, the sensitivity of ESI was 50.0% with a specificity of 96.8%. Of patients who had a level 1 intervention, 60.9% of patients were undertriaged with 46.8% being assigned ESI II and 14.1% being assigned ESI III-V.  

The authors also looked for what variables could be associated with mistrage and found:

• More likely to be mistriaged in both directions:
  • Younger, male, and Black patients were mistriaged in both directions more often compared to older, female, and White patients 
  • Patients living in poorer neighborhoods were also more likely to be mistriaged in both directions
  • Patients arriving during non-office hours 
• More likely to be undertriaged:
  • Patients arriving by ambulance
  • Patients taking insulin or sulfonylureas 
  • Patients with high comorbidity burdens 
  • Patients with a recent ICU admission 

Kudos to these authors for this incredible effort to attempt to evaluate how well ESI performs. Triage is an incredibly important part of our jobs as we manage limited resources, and these authors have identified multiple opportunities for improvement in the most commonly used triage system in the country. It will be interesting to see future studies and interventions about how to minimize mistriage, the tradeoffs between excess over- and undertriage, and how this algorithm assesses mistriage in other populations. This is a dense paper and worth a full read if you want to learn more!

Source
Evaluation of the Emergency Severity Index in US Emergency Departments for the Rate of Mistriage. JAMA network open vol. 6,3 e233404. 1 Mar. 2023, doi:10.1001/jamanetworkopen.2023.3404