How Common Is Pediatric Emergency Mistriage?

Only one third of pediatric patients were correctly triaged at emergency departments (EDs) in a Northern California healthcare system, according to a multicenter retrospective study published in JAMA Pediatrics. Researchers also identified gender, age, race, ethnicity, and comorbidity disparities in those who were undertriaged.

The researchers found that only 34.1% of visits were correctly triaged, while 58.5% were overtriaged and 7.4% were undertriaged. The findings were based on analysis of more than 1 million pediatric emergency visits over a 5-year period that used the Emergency Severity Index (ESI) version 4 for triage.

“The ESI had poor sensitivity in identifying a critically ill pediatric patient, and undertriage occurred in 1 in 14 children,” wrote Dana R. Sax, MD, a senior emergency physician at the Permanente Medical Group in Northern California, and her colleagues.

“More than 90% of pediatric visits were assigned a mid to low triage acuity category, and actual resource use and care intensity frequently did not align with ESI predictions,” the authors wrote. “Our findings highlight an opportunity to improve triage for pediatric patients to mitigate critical undertriage, optimize resource decisions, standardize processes across time and setting, and promote more equitable care.”

The authors added that the study findings are currently being used by the Permanente system “to develop standardized triage education across centers to improve early identification of high-risk patients.”

Disparities in Emergency Care

The results underscore the need for more work to address disparities in emergency care, wrote Warren D. Frankenberger, PhD, RN, a nurse scientist at Children’s Hospital of Philadelphia, and two colleagues in an accompanying editorial.

“Decisions in triage can have significant downstream effects on subsequent care during the ED visit,” they wrote in their editorial. “Given that the triage process in most instances is fully executed by nurses, nurse researchers are in a key position to evaluate these and other covariates to influence further improvements in triage.” They suggested that use of clinical decision support tools and artificial intelligence (AI) may improve the triage process, albeit with the caveat that AI often relies on models with preexisting historical bias that may perpetuate structural inequalities.

Study Methodology

The researchers analyzed 1,016,816 pediatric visits at 21 emergency departments in Kaiser Permanente of Northern California between January 2016 and December 2020. The patients were an average 7 years old, and 47% were female. The researchers excluded visits that lacked ESI data or had incomplete ED time variables as well as those with patients who left against medical advice, were not seen, or were transferred from another ED.

The study relied on novel definitions of ESI undertriage and overtriage developed through a modified Delphi process by a team of four emergency physicians, one pediatric emergency physician, two emergency nurses, and one pediatric ICU physician. The definition involved comparing ESI levels to the clinical outcomes and resource use.

Resources included laboratory analysis, electrocardiography, radiography, CT, MRI, diagnostic ultrasonography (not point of care), angiography, IV fluids, and IV, intramuscular, or nebulized medications. Resources did not include “oral medications, tetanus immunizations, point-of-care testing, history and physical examination, saline or heparin lock, prescription refills, simple wound care, crutches, splints, and slings.”

Level 1 events were those requiring time-sensitive, critical intervention, including high-risk sepsis. Level 2 events included most level 1 events that occurred after the first hour (except operating room admission or hospital transfer) as well as respiratory therapy, toxicology consult, lumbar puncture, suicidality as chief concern, at least 2 doses of albuterol or continuous albuterol nebulization, a skeletal survey x-ray order, and medical social work consult with an ED length of stay of at least 2 hours. Level 3 events included IV mediation order, any CT order, OR admission or hospital transfer after one hour, or any pediatric hospitalist consult.

Analyzing the ED Visits

Overtriaged cases were ESI level 1 or 2 cases in which fewer than 2 resources were used; level 3 cases where fewer than 2 resources were used, and no level 1 or 2 events occurred; and level 4 cases where no resources were used.

Undertriaged cases were defined as the following:

• ESI level 5 cases where any resource was used and any level 1, 2, or 3 events occurred.
• Level 4 cases where more than 1 resource was used and any level 1, 2, or 3 events occurred.
• Level 3 cases where any level 1 event occurred, more than one level 2 event occurred, or any level 2 event occurred, and more than one additional ED resource type was used.
• Level 2 cases where any level 1 event occurred.

About half the visits (51%) were assigned ESI 3, which was the category with the highest proportion of mistriage. After adjusting for study facility and triage vital signs, the researchers found that children aged 6 and older were more likely to be undertriaged than those younger than 6, particularly those aged 15 and older (relative risk [RR], 1.36).

Undertriage was also modestly more likely with male patients (female patients’ RR, 0.93), patients with comorbidities (RR, 1.11-1.2), patients who arrived by ambulance (RR, 1.04), and patients who were Asian (RR, 1.10), Black (RR, 1.05), or Hispanic (RR, 1.04). Undertriage became gradually less likely with each additional year in the study period, with an RR of 0.89 in 2019 and 2020.

Among the study’s limitations were use of ESI version 4, instead of the currently used 5, and the omission of common procedures from the outcome definition that “may systematically bias the analysis toward overtriage,” the editorial noted. The authors also did not include pain as a variable in the analysis, which can often indicate patient acuity.

Further, this study was unable to include covariates identified in other research that may influence clinical decision-making, such as “the presenting illness or injury, children with complex medical needs, and language proficiency,” Dr Frankenberger and colleagues wrote. “Furthermore, environmental stressors, such as ED volume and crowding, can influence how a nurse prioritizes care and may increase bias in decision-making and/or increase practice variability.”

The study was funded by the Kaiser Permanente Northern California (KPNC) Community Health program. One author had consulting payments from CSL Behring and Abbott Point of Care, and six of the authors have received grant funding from the KPNC Community Health program. The editorial authors reported no conflicts of interest.

This article originally appeared on MDedge.com, part of the Medscape Professional Network.