Concordance between Emergency Department Triage Score and Resource Utilization for Children from Different Racial and Ethnic Groups

Pediatric emergency department (ED) triage is designed to prioritize unscheduled care based on acuity, expected resource need, and time of arrival.¹The most widely used triage system in the United States, the Emergency Severity Index (ESI), requires ED clinicians to first identify the sickest patients (levels 1 and 2). Stable patients are then sorted based on anticipated resource utilization (levels 3 – 5), with higher priority assigned to those likely to require greater resources.¹ Large investigations have recently demonstrated the subjectivity and limitations in existing triage systems and highlighted the discordance in resources predicted and resources utilized.^2,3

Developing consistent, high-quality, and equitable triage systems is a key research priority for pediatric emergency medicine.⁴ Triage inconsistencies may disproportionately impact children based on their race and ethnicity. However, existing research on the intersection of triage with race and ethnicity is limited, with most research conducted in single academic pediatric centers or on single illness presentations.^5–8

Therefore, we sought to evaluate whether child race and ethnicity is associated with discordance in triage resource prediction and ED utilization in a large nationally representative sample of clinically stable children receiving care. We hypothesized that there would be greater discordance for non-Hispanic Black and Hispanic children as compared to non-Hispanic White children given the potential for bias in clinical assessment and clinical tools.

We performed a cross-sectional analysis of ED visits for children <18 years of age who appeared in the 2018-2022 National Hospital Ambulatory Medical Care Survey (NHAMCS).⁹ We excluded visits from EDs which 1) contributed fewer than 10 visits per year because we were unable to ascertain if they were pediatric or general EDs, and 2) did not use the emergency severity index score (ESI). We also excluded visits with missing ESI scores, visits with ESI 1 or 2, given that these are assigned on the basis of illness immediacy as opposed to resource prediction, and visits with a disposition of missing or left against medical advice.

The primary exposure was a four-level categorical variable of race and ethnicity (Hispanic, Non-Hispanic Black, Non-Hispanic Other (inclusive of Asian, Native Hawaiian or Pacific Islander, American Indian or Alaska Native, or multiracial), and Non-Hispanic White), readily available in NHAMCS. This variable was assigned in NHAMCS based on surveyor review of the medical record and was imputed for 20% of the sample.⁹ This variable was chosen to evaluate the effect of interpersonal, structural, and systemic racism on concordance between triage resource prediction and ED utilization.

We evaluated higher than predicted (ED utilization > triage prediction, hereto referred as “high”), lower than predicted (ED utilization < triage prediction, hereto referred as “low”) and discordant (ED utilization ≠ triage prediction) resource utilization (Supplemental Table 1). Resources were defined according to ESI and assessed in NHAMCS using previously described methods.³

We used descriptive statistics, t-tests, and chi-square tests to report baseline characteristics and outcomes. We performed unadjusted and adjusted logistic regression analyses to evaluate the association between race and ethnicity and the three outcomes of high, low, and discordant resource utilization. The following covariates were evaluated for potential inclusion in adjusted models: age category (<1 year, 1-5 years, 6-12 years, 13-18 years), ≥1 chronic condition (e.g., asthma, obesity, cancer), sex, insurance (public or other), overnight presentation, weekend presentation, ambulance arrival, ED region (Midwest, Northeast, South, West), and pediatric (≥75 patients seen are <18 years) vs. general (<75% patients seen are <18 years) ED. Those covariates which were significantly associated (p<0.05) with race and ethnicity were included. All analyses were performed using encounter-level patient survey weights and R version 4.3.2 (R Foundation, Vienna, Austria). This study received an exempt determination from the study hospital’s Institutional Review Board.

The study cohort included 10,936 unique (91.2 million weighted) pediatric ED visits. Hispanic, Non-Hispanic Black, Non-Hispanic Other, and White children represented 29.0%, 23.5%, 4.1%, and 43.5% of the cohort, respectively. Of the 147.3 million weighted ED visits for children <18 years who appeared in the 2018-2022 NHAMCS survey, the exclusions included 39.4 million for missing ESI score, 9.4 million for ESI 2, 5.4 million for ED without ESI, 1.1 million for ESI 1, 0.4 million for missing disposition, 0.3 million for left against medical advice, and 0.2 million for ED with too low of volume. Thirty-six million (39%) were triaged as ESI 3, 48.4 million (53%) as ESI 4, and 6.8 million (8%) as ESI 5. The majority of children had public insurance (65%) and presented to a general ED (68%).

When stratified by race and ethnicity, high resource utilization (ED utilization > triage prediction) was most common in Non-Hispanic Black (11%), followed by Hispanic (9%), Non-Hispanic White (9%) and Non-Hispanic Other (6%). These differences were not statistically significant after adjustment for covariates associated with child race and ethnicity (Supplemental Table 2, Table 1). Low resource utilization (ED utilization < triage prediction) was common in all groups: for Hispanic (46%), Non-Hispanic Black (46%), Non-Hispanic Other (45%), and Non-Hispanic White children (42%), respectively. In adjusted analysis, we found that Hispanic (adjusted OR (95% CI): 1.23 (1.05-1.44)) and Non-Hispanic Black (adjusted OR (95% CI): 1.30 (1.11-1.52)) children had a higher odds of low resource utilization compared to Non-Hispanic White children. Similarly, Hispanic (adjusted OR (95% CI): 1.17 (1.02-1.36)) and Non-Hispanic Black (adjusted OR (95% CI): 1.28 (1.11-1.48)) children had a higher odds of discordant resource utilization than Non-Hispanic White children.

Table 1:

Association between race and ethnicity and resource concordance

	n (%*)	OR (95% CI)	aOR1 (95% CI)
High resource utilization (ED utilization > triage prediction)
Hispanic (n=26.5)	2.2 (9%)	0.98 (0.78-1.24)	1.00 (0.78-1.29)
Non-Hispanic Black (n=21.4)	2.4 (11%)	1.32 (1.02-1.71)	1.24 (0.97-1.59)
Non-Hispanic Other (n=3.7)	0.4 (6%)	0.72 (0.44-1.18)	0.74 (0.44-1.23)
Non-Hispanic White (n=39.7)	3.4 (9%)	Ref	Ref
Low resource utilization (ED utilization < triage prediction)
Hispanic (n=26.5)	12.3 (46%)	1.25 (1.08-1.44)	1.23 (1.05-1.44)
Non-Hispanic Black (n=21.4)	9.9 (46%)	1.34 (1.15-1.55)	1.30 (1.11-1.52)
Non-Hispanic Other (n=3.7)	1.7 (45%)	1.13 (0.84-1.51)	1.05 (0.80-1.39)
Non-Hispanic White (n=39.7)	16.8 (42%)	Ref	Ref
Discordant resource utilization (ED utilization ≠ triage prediction)
Hispanic (n=26.5)	14.5 (46%)	1.17 (1.02-1.34)	1.17 (1.02-1.36)
Non-Hispanic Black (n=21.4)	12.3 (57%)	1.31 (1.15-1.51)	1.28 (1.11-1.48)
Non-Hispanic Other (n=3.7)	2.1 (57%)	1.03 (0.79-1.33)	0.99 (0.77-1.28)
Non-Hispanic White (n=39.7)	20.2 (51%)	Ref	Ref

^*Row percentages

¹Covariates include ED type (pediatric vs. general), presence of chronic condition (yes vs. no), US census region, public insurance (yes vs. no), age category (<1 year old, 1-5 years, 6-12 years, 13-18 years)

In this national sample of clinically stable children receiving ED care, we found that, compared to Non-Hispanic White children, Hispanic and Non-Hispanic Black children were more likely to experience low resource utilization (ED utilization < triage prediction). This suggests that, for Non-Hispanic Black and Hispanic children, ED clinicians are less likely to apply the level of resources initially anticipated to be required in triage. Although the specific differences identified in this study are modest, our findings add to the substantial existing evidence that Non-Hispanic White children receive different care than Black and Hispanic children, even after accounting for negative social drivers of health that disproportionately impact Non-Hispanic Black and Hispanic children,.

On a surface-level assessment, our findings might suggest that Hispanic and Non-Hispanic Black children disproportionately experience the benefits of higher triage levels, such as lower wait times and expedited access to ED care. However, it is important to interpret our findings within the larger context in which Non-Hispanic Black and Hispanic children experience overall higher rates of disease morbidity and mortality compared to Non-Hispanic White children.⁴ The cumulative effect of systematically different care throughout the process of pediatric ED care delivery may be an important driver of inequitable health outcomes.

As we work to understand how and why these differences in care delivery arise, we hypothesize that our findings may be explained by two distinct timepoints during the visit: ED triage and the patient/provider ED interaction. We have intentionally described our findings as discordance (not mistriage) to acknowledge that differences in both triage and clinician medical decision-making throughout the ED visit may be driving the observed findings. There is inherent subjectivity in triage assessment, and the high rate of discordance across all children clearly demonstrates a need to improve our triage processes. The same is true for the patient/provider interaction wherein often subjective decisions are made to pursue diagnostic testing and treatment. Standardized disease-specific guidelines to guide this patient/provider interaction may reduce disparities but should be careful to limit opportunities for individual clinician discretion.¹⁰ Furthermore, improvement efforts across the spectrum of ED care must intentionally consider health equity so as not to inadvertently worsen disparities.¹¹ Critically necessary efforts to improve triage and ED resource utilization must be mindful of these considerations.

Our analysis is subject to limitations. First, we did not include children with ESI levels 1 and 2, who are assigned triage levels based on perceived clinical instability as this data was not available in NHAMCS. Second, the exposure variable of race and ethnicity was assessed through NHAMCS surveyor chart review and imputed, by NHAMCS, for the 20% of visits where the variable was missing. It is therefore possible that the child’s race and ethnicity was not self-reported and that children with missing race and ethnicity data may have different experiences in the ED than those for whom the variable was available. Third, the chronic condition indicator in NHAMCS is not comprehensive of all pediatric chronic conditions so may not fully capture a child’s medical complexity. Finally, NHAMCS does not include medication route which is used to assign ESI, so we used previously published methods to categorize medication route.³

In this large national sample of pediatric ED visits, we identified disparities by child race and ethnicity in concordance between resources anticipated at triage and ED resource utilization. Our work provides additional evidence of systematic differences, based on a child’s race and ethnicity, in pediatric ED care delivery. Future efforts to improve pediatric triage and ED resource utilization should be developed and tested in ways that consider the health and outcomes of all children seeking emergency care.