ACADEMIC EMERGENCY MEDICINE 2010; 17:801–808 © 2010 by the Society for Academic Emergency Medicine
Objectives: This study assessed whether sociodemographic differences exist in triage assignment and whether these differences affect initial diagnostic testing in the emergency department (ED) for patients presenting with chest pain.
Methods: A nationally representative ED data sample for all adults (≥18 years) was obtained from the National Hospital Ambulatory Health Care Survey of EDs for 1997–2006. Weighted logistic regression was used to examine the associations between race and presenting symptom, triage assignment, and test ordering, adjusting for patient and hospital characteristics.
Results: Over 10 years, an estimated 78 million visits to the ED presented with a complaint of chest pain. Of those presenting with chest pain, African Americans (odds ratio [OR] = 0.70; 99% confidence interval [CI] = 0.53 to 0.92), Hispanics (OR = 0.74; 99% CI = 0.51to 0.99), Medicaid patients (OR = 0.72; 99% CI = 0.54 to 0.94), and uninsured patients (OR = 0.65; 99% CI = 0.51 to 0.84) were less likely to be triaged emergently. African Americans (OR = 0.86; 99% CI = 0.70 to 0.99), Medicaid patients (OR = 0.70; 99% CI = 0.55 to 0.88), and uninsured patients (OR = 0.70; 99% CI = 0.55 to 0.89) were less likely to have an electrocardiogram (ECG) ordered. African Americans (OR = 0.69; 99% CI = 0.49 to 0.97), Medicaid patients (OR = 0.67; 99% CI = 0.47 to 0.95), and uninsured patients (OR = 0.66; 99% CI = 0.44 to 0.96) were less likely to have cardiac enzymes ordered. Similarly, African Americans and Hispanics were less likely to have a cardiac monitor and pulse oximetry ordered, and Medicaid and uninsured patients were less likely to have a cardiac monitor ordered.
Conclusions: Persistent racial, sex, and insurance differences in triage categorization and basic cardiac testing exist. Eliminating triage disparities may affect “downstream” clinical care and help eliminate observed disparities in cardiac outcomes.
Disparities in cardiac care are well documented.1,2 Evidence demonstrating racial and ethnic disparities in invasive procedures is especially robust,3 with disparities in catheterization4–6 and revascularization procedures, including thrombolysis, angioplasty, and coronary artery bypass.2,5,7 Additionally, women are less likely to be referred for cardiac catheterization8 and undergo invasive cardiac procedures.6,9,10
A patient’s presenting symptoms and initial emergency department (ED) triage assignment could contribute to observed disparities in cardiac care. The initial triage decision in the ED is an important clinical decision step determining the perceived level of acuity and possible diagnosis of the patient. Therefore, it often determines how quickly the patient is assessed by a physician. In conjunction with presenting symptoms, the ordering of an electrocardiogram (ECG) will guide clinical decisions regarding pharmacologic interventions, the need for catheterization, and revascularization. The American College of Cardiology/American Heart Association (ACC/AHA) guidelines specify that an ECG should be obtained and interpreted within 10 minutes of suspicion of acute coronary syndrome (ACS) on arrival to an ED.11,12
Prior studies have shown age and sex triage differences, with younger patients and women being less likely to be emergently triaged.13,14 Some research has demonstrated that triage nurses in the ED perceived middle-aged men as more likely to need urgent care and are more likely to consider a cardiac cause of symptoms for middle-aged men than matched females presenting with the same symptoms.13 In addition, prior research has also demonstrated errors in triage of chest pain.14–16
Given that approximately 6 million patients with chest pain are evaluated in U.S. EDs each year,17 it is important to assess whether ED triage is an important clinical factor affecting “downstream” cardiac care. Using nationally representative ED data from the National Hospital Ambulatory Health Care Survey of Emergency Departments (NHAMCS-ED) for 1997–2006, we conducted a study to assess whether sex, racial, or ethnic differences in presentation exist among patients with a final ED diagnosis of acute myocardial infarction (AMI). In addition, we analyzed whether sociodemographic differences exist in triage assignment and initial cardiac diagnostic testing in the ED.
This was an analysis of representative ED data for all adults (≥18 years) from the NHAMCS-ED for 1997–2006. This study was deemed by the institutional review board as exempt from formal review.
Study Setting and Population
The NHAMCS-ED is a cross-sectional survey of patient visits to EDs of noninstitutional general and short-stay hospitals, exclusive of federal, military, and Veterans Administration hospitals. The NHAMCS selects patient visits for inclusion based on a four-stage probability design. Data from medical records are extracted by specially trained hospital staff and coded at a central location. Up to three reasons for the visit, as stated by the patient, are recorded and coded using a standard classification system. A detailed description of the history, design, data collection, and publicly accessible data for NHAMCS-ED is available at the website of the National Center for Health Statistics (NCHS).18
We analyzed patient demographic characteristics: age, sex, and race/ethnicity (African American, Hispanic, white, other race). We defined insurance status in four categories: private, Medicare, Medicaid, or uninsured. We defined uninsured as those patients with payment listed as no charge or self-pay. Patient-reported reason for the visit is collected in the survey instrument. We considered a patient to have chest pain if his or her presenting complaint was coded as chest pain, discomfort, pressure, tightness, or heaviness (includes chest pressure); burning sensation in the chest; or heart pain (includes anginal pain, heart distress, and pain over the heart). Other presenting symptoms analyzed in this study, among those patients with a final ED diagnosis of AMI, include breathing difficulty, abdominal pain, and anxiety.
The NHAMCS uses the following triage categories: emergent (should wait less than 15 minutes), urgent (can wait 15–60 minutes), semiurgent (can wait 1–2 hours), and nonurgent (can wait 2–24 hours). For analytic purposes, we dichotomized triage category into “urgent” (including both emergent and urgent) and “nonemergent” (including both semiurgent and nonurgent). We analyzed the evaluation tests ordered: ECG, cardiac monitor, pulse oximetry, and cardiac enzymes (cardiac enzymes available only for 2005–2006).
We established the ED diagnosis of AMI using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes for diagnosis of AMI (410.xx). The ED diagnosis could be further specified as tentative or provisional indicating an admission for chest pain requiring further time for observation and evaluation (i.e., “rule-out AMI”).
We analyzed hospital characteristics (ownership, region of the country, and urban vs. rural), day of the week on which the patient arrived to the ED, and mode of arrival (ambulance, walk-in, or public service nonambulance such as arrival by police, social services, etc.). We categorized hospital ownership as nonprofit, for-profit, and government owned. Finally, we created a linear variable representing the survey year (ranging from 0 to 9).
Due to the complex design of NHAMCS, we used SUDAAN statistical software (Research Triangle Park, NC) for all analyses. We included special sample weights provided by the NCHS in the analysis to adjust for sampling bias, nonresponse bias, and population weighting adjustment. Our study consists of two analyses. First, prior research has documented sex and ethnic differences in AMI presentation.19–23 Therefore, we analyzed the presenting symptoms among patients who received a final ED diagnosis of AMI to determine possible differences in cardiac presentation. We stratified the chi-square test by race/ethnicity and sex to compare the distribution of presenting symptoms (i.e., chest pain, breathing difficulties, abdominal pain, and anxiety) by demographic characteristics, mode of arrival, day of arrival, triage category assignment, evaluative cardiac tests ordered, ED diagnosis, hospital characteristics, and survey year. Weighted logistic regression models were fitted to examine the multivariable-adjusted odds ratios (ORs) with 99% confidence intervals (CIs) to account for the large number of reported results. We constructed a model for each of the presenting symptoms as an outcome variable adjusting for all of the previously mentioned variables. All of these variables were included as predictors in the regression models because of prior published clinical and research associations.
Our second analysis focused on those adults presenting with chest pain to an ED to assess for possible differences in triage and cardiac testing. We used the chi-square test to compare the distribution of study variables among patients who presented with chest pain by demographic characteristics, mode of arrival, day of arrival, triage category assignment, evaluative cardiac tests ordered, ED diagnosis, and hospital characteristics. We used weighted logistic regression to examine the multivariable-adjusted ORs with 99% CIs. We constructed a model for each of the following outcomes: triage assignment, ECG, pulse oximetry, and cardiac monitor ordering. We included all of the variables in Table 1 as covariates in each model including day of arrival to the ED. All of these variables were included as predictors in the regression models because of prior published clinical and research associations. A linear survey year term was included in all of the models to assess changes over time. A separate analysis, similar to those previously mentioned, was performed with cardiac enzyme ordering as the dependent term because this outcome was available only for 2005–2006. Finally, because the risk of AMI rises with increasing age, we a priori decided to repeat the analyses for triage assignment and cardiac testing using only those patients greater than 41 years of age.
|Characteristics||% White (n = 12,426)||% African American (n = 4,022)||% Hispanic (n = 1,934)||% Other Race (n = 603)|
|Mode of arrival|
|Public service nonambulance†||1||2||3||1|
|Diagnostic tests ordered/received|
|Provisional diagnosis of AMI—yes||5||3||3||3|
|Diagnosis of AMI—yes||8||4||5||8|
The Hosmer-Lemeshow test statistic was used to test the goodness of fit of our logistic regression models.24 All of our logistic regression models had a nonsignificant Hosmer-Lemeshow test statistic, ruling out a gross lack of model fit. All p-values reported are for two-tailed tests, and a value of <0.05 or less was considered statistically significant.
The total sample consisted of 235,117 ED visits. Of these, 21,629 presented with a complaint of chest pain, representing 78 million such visits nationally over the 10-year period. African American and Hispanic patients tended to be younger than 65 years of age, have Medicaid insurance or be uninsured, and receive care at an urban hospital compared to whites (Table 1).
Of Hispanics with a diagnosis of AMI, 65% presented with chest pain, 22% with breathing difficulties, 11% with abdominal pain, and 2% with anxiety. Of African Americans with a diagnosis of AMI, 56% presented with chest pain, 29% with breathing difficulties, 14% with abdominal pain, and none with anxiety. Whites with a diagnosis of AMI presented primarily with chest pain (65%); the remaining presented as follows: 21% with breathing difficulties, 13% with abdominal pain, and 2% with anxiety. Among patients with AMI who identified themselves as other race, 70% presented with chest pain, 11% with breathing difficulties, 14% with abdominal pain, and 5% with anxiety.
Women of any race with a diagnosis of AMI presented less often to the ED with chest pain compared to men (41% vs. 59%, p = 0.0002). In multivariate adjusted models, no significant differences were found in presenting symptoms for African American and Hispanic patients who had a diagnosis of AMI compared to whites. However, women with a diagnosis of AMI were still less likely than men to present with chest pain.
Of all patients presenting with chest pain, African Americans (0.70; 99% CI = 0.53 to 0.92) and Hispanics (0.74; 99% CI = 0.51 to 0.99) were less likely to be categorized in the emergent triage category compared to whites in multivariate adjusted models (Table 2). In addition, the uninsured (0.65; 99% CI = 0.51 to 0.84) and those with Medicaid (0.72; 99% CI = 0.54 to 0.94) were less likely to be categorized in emergent triage categories compared to those with private insurance. Patients with chest pain in the southern states of the United States, patients at urban hospitals, or those who walked into the ED compared to those who arrived by ambulance were less likely to be triaged emergently. Patients with a provisional diagnosis of AMI were less likely to be emergently triaged (0.31; 99% CI = 0.10 to 0.90) while those with a diagnosis of AMI were more likely to be emergently triaged (4.64; 99% CI = 1.75 to 12.31).
|Characteristics||Adjusted OR (99% CI)|
|African American||0.70 (0.53–0.92)|
|Hispanic, any race||0.74 (0.51–0.99)|
|Mode of arrival|
|Public service nonambulance†||0.59 (0.18–1.93)|
|Provisional diagnosis of AMI||0.31 (0.10–0.90)|
|Diagnosis of AMI||4.64 (1.75–12.31)|
African Americans, patients with Medicaid, and those without insurance were less likely to have an ECG ordered when presenting with chest pain (Table 3). African Americans, Hispanics, Medicaid patients, and uninsured patients were less likely to have a cardiac monitor ordered (Table 3). African Americans and Hispanics were less likely to have pulse oximetry ordered (Table 3). Men and older patients were more likely to have all three tests ordered. Compared to all modes of arrival, those who walked into the ED were less likely to receive a cardiac monitor and pulse oximetry. Patients with a provisional diagnosis of AMI were more likely to receive all tests. Cardiac monitoring was more likely to be ordered in those patients with a diagnosis of AMI. In 2005–2006, African Americans, Medicaid patients, and uninsured patients were less likely to have cardiac enzymes ordered (Table 4). Older patients, those with a diagnosis of AMI, or those in the emergent triage category were more likely to have cardiac enzymes ordered.
|Predictor||ECG, Adjusted OR (99% CI)||Cardiac Monitor, Adjusted OR (99% CI)||Pulse Oximetry, Adjusted OR (99% CI)|
|41–65||2.22 (1.86–2.66)||1.97 (1.67–2.31)||1.37 (1.18–1.60)|
|>65||2.63 (1.96–3.53)||2.44 (1.89–3.16)||1.38 (1.11–1.71)|
|African American||0.86 (0.70–0.99)||0.66 (0.52–0.82)||0.76 (0.61–0.94)|
|Hispanic, any race||0.98 (0.73–1.32)||0.69 (0.51–0.92)||0.71 (0.53–0.95)|
|Other||0.70 (0.36–1.34)||0.85 (0.49 –1.46)||0.87 (0.55–1.39)|
|Male||1.25 (1.10–1.48)||1.16 (1.00–1.35)||1.20 (1.05–1.38)|
|Medicare||0.96 (0.71–1.31)||0.89 (0.70–1.14)||1.08 (0.89–1.32)|
|Medicaid||0.70 (0.55–0.88)||0.86 (0.69–0.95)||1.12 (0.91–1.38)|
|Uninsured||0.70 (0.55–0.89)||0.65 (0.52–0.82)||1.01 (0.81–1.27)|
|Mode of arrival|
|Public service nonambulance†||0.68 (0.31–1.54)||0.87 (0.39–1.96)||0.79 (0.39–1.60)|
|Walk-in||0.85 (0.68–1.00)||0.80 (0.69–0.93)||0.88 (0.76–1.00)|
|Emergent||5.49 (3.80–7.92)||4.36 (2.77–6.86)||1.36 (0.91–2.05)|
|Urgent||2.98 (2.11–4.20)||2.71 (1.71–4.29)||1.08 (0.73–1.61)|
|Semiurgent||1.55 (1.10–2.25)||1.72 (1.10–2.93)||0.99 (0.63–1.56)|
|Provisional ED diagnosis of AMI||3.75 (1.13–9.35)||1.76 (1.10–3.03)||2.17 (1.30–3.63)|
|ED diagnosis of AMI||1.58 (0.75–3.33)||1.69 (1.15–2.50)||1.24 (0.85–1.81)|
|Northeast||1.12 (0.81–1.54)||0.69 (0.48–1.00)||0.77 (0.54–1.11)|
|South||1.03 (0.77–1.38)||0.76 (0.54–1.06)||0.90 (0.64–1.25)|
|West||0.87 (0.63–1.21)||0.60 (0.40–0.90)||1.04 (0.69–1.58)|
|Nonprofit||1.08 (0.80 – 1.46)||1.02 (0.71–1.46)||1.22 (0.84–1.77)|
|Government||0.99 (0.67 – 1.46)||0.89 (0.56–1.41)||0.76 (0.46–1.26)|
|Urban||1.25 (1.00 – 1.59)||1.02 (0.74 – 1.39)||1.25 (0.86–1.82)|
|Cardiac Enzyme Testing, Adjusted OR (99% CI)|
|African American||0.69 (0.49–0.97)|
|Hispanic, any race||0.73 (0.47–1.08)|
|Mode of arrival|
|Public service nonambulance†||0.44 (0.14–1.36)|
|Provisional diagnosis of AMI||0.97 (0.23–4.06)|
|Diagnosis of AMI||2.39 (1.05–6.45)|
These triage and test ordering differences persisted in analyses that included only patients older than 41 years of age with chest pain. Adjusting for year of survey and day of the week on which the patient arrived made no changes to our findings. Of those presenting with chest pain, minority patients were not more likely to have a provisional diagnosis of AMI. There were no significant differences in triage category or test ordering by patient or hospital characteristics among patients with a diagnosis of myocardial infarction in the ED.
Our nationally representative study demonstrates that patient sociodemographic factors may influence the ED triage process. First, we found no differences in presenting symptoms for African American and Hispanic patients who had an ED diagnosis of AMI compared to whites. However, women with a diagnosis of AMI were less likely than men to present with chest pain. Second, of those presenting with chest pain, African Americans, Hispanics, those with Medicaid, and the uninsured were less likely to be triaged into the emergent category. Third, these groups of patients were less likely to receive basic cardiac diagnostic testing and the ordering of cardiac enzymes in the ED. Finally, we found no change in these racial/ethnic differences in test ordering over the 10-year period.
Our finding that minority patients and whites presented their AMI symptoms similarly should be taken in context of conflicting evidence in the literature. Some studies have found higher rates of atypical presentations among minorities,19,20 while others have found no difference.25–27 One study has even found a lower risk of atypical presentations among nonwhite patients compared to whites.21 However, our study is in agreement with prior studies that document higher rates of atypical presentations among women compared with men.22,23
Insurance status appears to be associated with triage decisions. Medicaid and uninsured patients were less likely to be triaged emergently in our study. This is consistent with previous studies demonstrating the influence of insurance status on ED decision-making.28,29 For example, one study showed that uninsured patients presenting to the ED with an injury were less likely to be hospitalized compared to those privately insured, after controlling for injury severity.29 Medicaid and uninsured patients with coronary artery disease enrolled in clinical trials are significantly less likely to undergo coronary catheterization,30 and patients enrolled in managed care plans are less likely to undergo angiography relative to those in traditional fee-for-service private plans.22
Nonclinical factors may influence triage decisions.31 Prior studies have shown age and sex triage differences with younger patients and women less likely to be emergently triaged.13,14 To our knowledge, our paper is the first to find triage differences by race in a nationally representative patient sample. In our study, minority patients presenting with chest pain were less likely to be triaged emergently. It is not clear why these differences exist, but the reasons are likely to be as complex and multifactorial as the triage decision-making process itself. Human factors may contribute to differential triaging.
Research supports the idea that triage nurses may be influenced by conscious or, more likely, unconscious biases about race/ethnicity, sex, and other sociodemographic characteristics as they make triage decisions.13 Triaging requires complex judgment based on a brief nurse-patient encounter, in which intra- and interpersonal characteristics are significant and differences are likely.32–34 In addition, the triage process occurs in a busy environment with high levels of complexity and uncertainty. These environmental characteristics have been associated with conditions that promote reliance on stereotypes, which can lead to biases in clinical decision-making.8,35,36 Stereotypes are known to be most activated by visible characteristics including race/ethnicity, age, and sex. Prior research demonstrates that triage nurses make triage decisions with the belief that patients do not present their symptoms objectively.32 For example, nurses doubted that patients with chest pain cried and believed that patients who were Asian, Hispanic, Oriental, and Middle Eastern were more demonstrative of their symptoms than patients of other cultural backgrounds who were perceived as more stoic.37
We found that minority patients, those with Medicaid, and the uninsured were less likely to receive basic cardiac diagnostic testing and the ordering of cardiac enzymes. This is consistent with prior studies demonstrating similar differences in test ordering in these patient groups.6,38–42 Pezzin et al.,43 using NHAMCS-ED data between 1995 and 2000, found similar results: that African Americans, women, the uninsured, and self-pay patients are less likely to receive early cardiac diagnostic tests when presenting to the ED with chest pain. Our study expands the time period of analysis from 1997 to 2006, allowing for analyses of changes in testing rates over a 10-year period. We found no change in racial or ethnic differences in test ordering over the 10-year period, similar to the 6-year analysis by Pezzin et al.
The ECG is an important first diagnostic test in the management of chest pain. In conjunction with presenting symptoms, it guides clinical decisions regarding pharmacologic interventions, the need for percutaneous coronary intervention (PCI), and hospitalization. The ACC/AHA guidelines specify that an ECG should be obtained and interpreted within 10 minutes of suspicion of ACS or arrival to an ED.11,12 In addition, a cardiac monitor and oxygen saturation tests allow for the evaluation of other noncoronary causes of chest pain, such as pulmonary embolism and pneumothorax. Delayed ECG acquisition has been shown to be associated with increased mortality at 30 days in ST-segment elevation AMI patients due to delays in receiving fibrinolysis or PCI.40
Hospital characteristics are important to triage decision-making as well. We found that patients with chest pain at urban hospitals were less likely to be triaged emergently. This is consistent with prior research that acute cardiac syndrome patients presenting to EDs in public hospitals are more likely to be discharged after triage than in privately funded hospitals after adjustment for patient and clinical factors.44
Our data are cross-sectional, and thus associations found are not proof of causality. Second, triage decision-making is a complex process involving a large number of clinical and nonclinical data. We were unable to control for all clinical comorbidities or other factors such as vital signs at presentation that could influence triage assignment. The number of cardiovascular risk factors a patient has, such as diabetes, prior AMI, or presence of cardiac stents, certainly should affect triage decisions. In addition, laboratory information, such as cardiac enzymes, is not available in NHAMCS-ED for all survey years included in our analysis. Also, we were not able to account for traumatic causes of chest pain, and we did not have a hospital discharge diagnosis. As a result, we are unable to differentiate whether the cause of differential triage is errors and misclassification of patients or if there is an inherent difficulty in identifying acute AMI patients among certain patient subgroups. Third, we were unable to account for hospital case mix or severity, which could influence clinical judgment. Our data on hospital characteristics do not provide potentially important facets of the delivery of care, such as ED volume and crowding,16 bed availability, and other clinical care structures such as the use of “critical care pathways.” Finally, we are unable to link triage and testing differences to clinical outcomes in a direct way, limiting our ability to assess the direct consequences of the observed differences in clinical processes.
Triage assessments are an important modifiable factor influencing clinical decisions and timing of care. In summary, we found persistent racial, sex, and insurance differences in triage categorization and cardiac testing over a 10-year period in a national sample of patients presenting to EDs. Eliminating triage disparities may affect “downstream” clinical care and help eliminate observed disparities in cardiac outcomes. Our findings indicate an important point of intervention for the use of quality improvement strategies focused on implementing and increasing adherence to protocol-driven triage and management algorithms that could eliminate disparities.45 In addition, the use of continuous quality monitoring with stratification of results by race/ethnicity should be implemented. Given the importance of cardiac care and the increase in racial and ethnic diversity, focusing efforts to reduce differences in triage for chest pain will be an important quality and equity-focused intervention.
Dr. López acknowledges the support of the Aetna Foundation (Grant 1200 205107). Dr. Wilper acknowledges the resources and the use of the facilities at the Boise VA Medical Center.