Untapped Potential for Emergency Department Observation Unit Use: A National Hospital Ambulatory Medical Care Survey (NHAMCS) Study

Introduction Millions of people present to the emergency department (ED) with chest pain annually. Accurate and timely risk stratification is important to identify potentially life-threatening conditions such as acute coronary syndrome (ACS). An ED-based observation unit can be used to rapidly evaluate patients and reduce ED crowding, but the practice is not universal. We estimated the number of current hospital admissions in the United States (US) eligible for ED-based observation services for patients with symptoms of ACS. Methods In this cross-sectional analysis we used data from the 2011–2015 National Hospital Ambulatory Medical Care Survey (NHAMCS). Visits were included if patients presented with symptoms of ACS (eg, chest pain, dyspnea), had an electrocardiogram (ECG) and cardiac markers, and were admitted to the hospital. We excluded patients with any of the following: discharge diagnosis of myocardial infarction; cardiac arrest; congestive heart failure, or unstable angina; admission to an intensive care unit; hospital length of stay > 2 days; alteplase administration, central venous catheter insertion, cardiopulmonary resuscitation or endotracheal intubation; or admission after an initial ED observation stay. We extracted data on sociodemographics, hospital characteristics, triage level, disposition from the ED, and year of ED extracted from the NHAMCS. Descriptive statistics were performed using sampling weights to produce national estimates of ED visits. We provide medians with interquartile ranges for continuous variables and percentages with 95% confidence intervals for categorical variables. Results During 2011–2015 there were an estimated 675,883,000 ED visits in the US. Of these, 14,353,000 patients with symptoms of ACS and an ED order for an ECG or cardiac markers were admitted to the hospital. We identified 1,883,000 visits that were amenable to ED observation services, where 987,000 (52.4%) were male patients, and 1,318,000 (70%) were White. Further-more, 739,000 (39.2%) and 234,000 (12.4%) were paid for by Medicare and Medicaid, respectively. The majority (45.1%) of observation-amenable hospitalizations were in the Southern US. Conclusion Emergency department-based observation unit services for suspected ACS appear to be underused. Over half of potentially observation-amenable admissions were paid for by Medicare and Medicaid. Implementation of ED-based observation units would especially benefit hospitals and patients in the American South.


INTRODUCTION
Over six million adults present to the emergency department (ED) with chest pain in the United States annually. 1,2 While there are multiple etiologies of chest pain, including non-cardiac and benign disorders, accurate and timely risk stratification is important to identify potentially life-threatening conditions such as acute coronary syndrome (ACS). Several objective measures (ie, electrocardiography [ECG], cardiac biomarkers, noninvasive imaging of the myocardium) 3 and decision-support tools 4 have been developed for ACS risk stratification. Yet 2-4% of patients with ACS are inadvertently discharged from the ED. 5,6,7 Of those patients with chest pain admitted for further evaluation, less than half will be diagnosed with ACS. 8 One factor contributing to these discrepancies in the ED is that ACS symptoms are often non-specific. 9 Additionally, multiple non-ACS conditions are associated with elevated troponin levels. 10 Emerging evidence suggests that ED-based observation units (EDOU) for chest pain may overcome these limitations by enabling implement-ation of a rapid risk-stratification protocol (eg, cardiac biomarker testing, telemetry monitoring, stress testing, echocardiogram) over a short period of time. 1,11 The use of EDOUs has been described since the 1980s. In 2006 an Institute of Medicine report, The Future of Emergency Care in the United States Health System, supported the use of EDOUs as a tool to reduce ED crowding, improve patient care, and reduce cost. 12 Although these units are diverse, a defining feature is the use of protocolized care with the goal of rapidly discharging the patient back home within 24 hours. Despite documented financial and patient benefits, their adoption has not been universal. 13 Recent estimates suggest that 39% of EDs have a separate observation or clinical decision unit. 2 While the utility of EDOUs for chest pain has been reported, 4,14 it is not fully known to what degree EDbased observation services could expand in the United States.
In this study we used a publicly available, de-identified, and unlinked survey database of nationwide ED visits to determine the number of patients admitted to the hospital from the ED for symptoms of ACS who could potentially have been evaluated in an EDOU. We also sought to determine which patient factors were most associated with patients being admitted despite meeting our derived observationeligible criteria. Our goal was to corroborate the potential for more EDOUs nationwide as a means to significantly reduce unnecessary hospital admissions and related expenses.

Study Design and Data Source
We conducted a cross-sectional analysis using data from the National Hospital Ambulatory Medical Care Survey (NHAMCS) from 2011-2015. Visits were included if patients presented with symptoms of ACS (eg, chest pain, dyspnea), had an electrocardiogram (ECG) and cardiac markers, and were admitted to the hospital. We obtained data for this analysis from the publicly available NHAMCS dataset published on the US Centers for Disease Control and Prevention (CDC) website. The database includes information that is de-identified and unlinked to the patient encounter and aggregated solely for informative and research purposes.
The NHAMCS is an annual, nationally representative probability sample survey administered by the CDC's National Center for Health Statistics. Data is collected on visits to outpatient clinics and EDs of non-institutional, short-stay. and general hospitals in 50 states and the District of Columbia, excluding federal, military, and Veterans Affairs hospitals. The NHAMCS uses a four-stage probability sampling design including selection of primary sampling units (PSU), hospitals within PSUs, clinics within hospitals, and patient visits within clinics. The exact methods of the NHAMCS survey have been described in detail elsewhere. 15 Hospitals are selected based on geographic PSUs. For the years included, on average 411 hospitals were eligible annually, and 369 participated, giving an unweighted average hospital sampling response rate of 89.8%. Sixteen data collection groups randomly rotate across these hospitals through 13 four-week reporting periods throughout the year. Contractors for the NHAMCS (SRA International, Inc., Durham, NC) collect data from ED visit medical records while being monitored by NHAMCS field representatives. The NHAMCS staff members independently check 10% of the Untapped Potential for EDOU Use: NHAMCS Study Navas et al.
data for accuracy. Error rates are 0.3-0.9% for various items on the survey; the survey includes patient-level data, patient disposition, and hospital-level data.

Study Population
For this analysis we focused on visits to hospital EDs for symptoms of ACS ( Figure 1). Visits were included if a cardiac troponin and ECG were ordered in the ED, the patient was admitted to inpatient status into the hospital and was discharged with a length of stay shorter than two days. We excluded patients with any of the following: discharge diagnosis of myocardial infarction, cardiac arrest, congestive heart failure, or unstable angina; admission to an intensive care unit; hospital length of stay > 2 days; alteplase administration, central venous catheter insertion, cardiopulmonary resuscitation or endotracheal intubation; or admission after an initial ED observation stay. We further excluded any patient with a non-cardiovascular primary diagnosis upon hospital discharge.

Variables
The NHAMCS survey records demographic data, payment source, clinician types, procedures, prescriptions, laboratory and radiographic tests ordered for each visit, up to three reasons for visit (chief complaints), the ED diagnosis (International Classification of Diseases, 9 th Revision codes), and the final hospital discharge diagnosis for those patients who were admitted to the hospital. In addition, we extracted the following variables from the NHAMCS database: age; race/ethnicity; gender; insurance status; clinician type; hospital characteristics (geographic location at the level of state, academic status, metropolitan area, and ownership); and disposition from the ED (admission, discharge, and transfer). We used data from the NHAMCS files for 2011-2015.

Ethics
This study was exempted from full review by our institutional review board.

Data Analysis
We performed data analyses using SAS version 9.4 (SAS Institute Inc., Cary, NC). Sociodemographics, hospital characteristics, and dispositions from the ED were summarized using the median with interquartile range for continuous variables and percentage with 95% confidence intervals for categorical variables. We calculated unweighted percentages for "reasons for visits" and final diagnoses. Estimates of average annual national visits were derived using survey procedures with the weights, strata, and PSU design variables provided by the NHAMCS.

RESULTS
During 2011-2015, there were an estimated 675,883,000 ED visits nationwide in the US. Of these, 14,353,000 patients with symptoms of ACS who had an ECG in the ED and cardiac markers were admitted to the hospital. This number was calculated using raw percentages of selective cohorts and population ratio adjustment. We identified 1,883,000 visits that may have been amenable to observation services. Of these visits, 987,000 (52.4 %) were by males and 1,318,000 (70.0 %) identified as White (Table 1). Furthermore, 739,000 (39.2 %) and 234,000 (12.4 %) were paid for by Medicare and Medicaid, respectively ( Figure 2). The majority of these observationamenable hospitalizations were in the Southern US (Figure 3).
When comparing ED visits leading to observation-amenable admissions to overall proportions of ED visits, they occurred proportionally slightly more in females, Medicare patients, and in the Midwest and South. These types of admissions occurred less often in Medicaid patients and in the US Northeast and West.

DISCUSSION
Chest pain is the second most common reason for ED visits  In this study we determined the proportion of patients who were hospitalized for symptoms of potential ACS who could have been observed in an EDOU. Emergency department-based observation unit services for ACS appear to be underused. We identified that over half of observation-amenable admissions were paid for by Medicare and Medicaid. These findings appear to vary geographically within the US. This is the first study based on a nationally representative sample to evaluate the rate of use of EDOUs for chest pain. Previous work has shown that national EDOUs are growing in number, from 19% in 2003 to 39% in 2017, and that chest pain is the most common EDOU symptom requiring diagnosis. 2,14,16 Although the number of EDOUs is growing, there remains room for further expansion. Studies have shown that patients at low risk or intermediate risk of ACS are more common than patients with unstable angina or ST-elevation myocardial infarction. 17 Indeed, we identified nearly two million patients within this risk category who were not evaluated in an EDOU. Importantly, EDOUs have been shown to reduce admission rates from the ED for chest pain. 18 Another benefit of the EDOU is reduction in costs, with an estimated savings of $124 per patient for ED visits with chest pain. 19 The majority of patients in our analysis who missed an opportunity for EDOU evaluation were Medicare beneficiaries. Older patients (age >65 years) have a higher rate of EDOU use for chest pain. 20 Previous work has specifically determined how much Medicaid and Medicare have paid for observation services. 21 These studies show that patients treated in observation units for chest pain are less likely to have an adverse event within 30 days and Medicare payments for these services are nearly half what they are for inpatient services. 22 On the other hand, these services may cost patients more because they fall under Medicare Part B (outpatient-related services), which requires them to pay a $183 deductible and 20% cost-sharing for services. 22 We also note that observation-amenable admissions occur proportionally more frequently in Medicare patients and less frequently in Medicaid patients, suggesting that financial incentives may play a role. Thus, it is possible that although EDOUs reduce a hospital's health-related expenditures, one reason for their underutilization is patient finances. It is important to note, however, that for commercially insured patients, observation units reduce total out-of-pocket expenses. 23 Increasing the use of observation services may require further expanding the number of EDOUs in the US and evaluating payment options for Medicare beneficiaries.
Despite the growing use of EDOUs, no previous work has specifically commented on geographic variation of observation for ACS. 24 However, studies have shown over half of EDOUs generally are in an urban setting and 30% are in the South. 25 Our analysis revealed that 45% of the visits for ACS that could have been seen in an EDOU were in hospitals in the Southeast US, while just 10% were in the Northeast US. One explanation for these differences may be that there are a greater number of EDOUs or a shorter distance to a hospital with an ED in the Northeast -an important factor when a patient is experiencing symptoms of ACS. It is also possible that many of the EDs in the Northeast and West are in high-density population areas and have already adapted EDOUs as a means of dealing with hospital crowding. An investigation of ED-managed observation units also found that EDs in an area with a median income of <$32,000 were less likely to have an observation unit. 26 This is particularly important given research showing that patients with a lower socioeconomic status may report to EDs less frequently for chest pain or delay seeking treatment for chest pain. 26,27

LIMITATIONS
Our study has several limitations. First, this was a crosssectional analysis using retrospective data of ED visits across five years (2011)(2012)(2013)(2014)(2015). However, the NHAMCS is a nationally representative database that includes data on patient visits to the ED, demographic characteristics, symptoms, chief complaint, diagnoses, laboratory services, and medications. We recognize that we present a hypothetical construct of what an observationamenable admission is, but we believe it is a reasonable estimate. This database estimate cannot account for numerous factors that may have impacted needs for hospitalization and were not in the existing NHAMCS variables, including social determinants of health. Furthermore, it represents a conservative estimate of the number of hospitalizations potentially amenable to observation services.
We did not include patients hospitalized for other, atypical anginal symptoms, and it may have been clinically appropriate and reasonable for a certain proportion of the hospitalizations that we excluded to have started out  Figure 3. Geographic region of emergency department visits amenable to observation services.

Navas et al.
Untapped Potential for EDOU Use: NHAMCS Study as observation status. Next, the unweighted ED response rate of the NHAMCS was <80%, which could have biased results and limited generalizability. Nevertheless, the NHAMCS is the largest dataset to date with populationbased estimates of ED visits in the US. Finally, while our inclusion criteria may be subject to misclassification bias, we used a comprehensive algorithm to determine participants with ACS symptoms who were not evaluated in an EDOU. It is also important to note that not all hospitals can build and staff observation units easily.

CONCLUSION
We identified 1,883,000 visits for ACS symptoms that were amenable to EDOU services. We also found that over half of observation-amenable admissions were paid for by Medicare and Medicaid, and were more likely to occur in the Southern US. These data support the need to further expand the use of EDOU for patients with symptoms of ACS. Although the benefits of ED-based observation service have been previously modeled theoretically and demonstrated in local settings, ultimately, further research should determine the economic and patient-oriented impacts of expansion of ED-based observation services as it actually occurs nationwide.