Increasing Length of Stay Among Adult Visits to U.S. Emergency Departments, 2001–2005

Authors

  • Andrew Herring MD,

    1. From the Highland General Hospital Department of Emergency Medicine/Alameda County Medical Center (AH), Oakland CA; Harvard Medical School (AW, DUH, SW, DFMB, CAC), Boston, MA; the Department of Medicine, Cambridge Health Alliance (AW, DUH, SW), Cambridge, MA; and the Department of Emergency Medicine, Massachusetts General Hospital (JAE, DFMB, CAC), Boston, MA.
    Search for more papers by this author
  • Andrew Wilper MD,

    1. From the Highland General Hospital Department of Emergency Medicine/Alameda County Medical Center (AH), Oakland CA; Harvard Medical School (AW, DUH, SW, DFMB, CAC), Boston, MA; the Department of Medicine, Cambridge Health Alliance (AW, DUH, SW), Cambridge, MA; and the Department of Emergency Medicine, Massachusetts General Hospital (JAE, DFMB, CAC), Boston, MA.
    Search for more papers by this author
  • David U. Himmelstein MD,

    1. From the Highland General Hospital Department of Emergency Medicine/Alameda County Medical Center (AH), Oakland CA; Harvard Medical School (AW, DUH, SW, DFMB, CAC), Boston, MA; the Department of Medicine, Cambridge Health Alliance (AW, DUH, SW), Cambridge, MA; and the Department of Emergency Medicine, Massachusetts General Hospital (JAE, DFMB, CAC), Boston, MA.
    Search for more papers by this author
  • Steffie Woolhandler MD, MPH,

    1. From the Highland General Hospital Department of Emergency Medicine/Alameda County Medical Center (AH), Oakland CA; Harvard Medical School (AW, DUH, SW, DFMB, CAC), Boston, MA; the Department of Medicine, Cambridge Health Alliance (AW, DUH, SW), Cambridge, MA; and the Department of Emergency Medicine, Massachusetts General Hospital (JAE, DFMB, CAC), Boston, MA.
    Search for more papers by this author
  • Janice A. Espinola MPH,

    1. From the Highland General Hospital Department of Emergency Medicine/Alameda County Medical Center (AH), Oakland CA; Harvard Medical School (AW, DUH, SW, DFMB, CAC), Boston, MA; the Department of Medicine, Cambridge Health Alliance (AW, DUH, SW), Cambridge, MA; and the Department of Emergency Medicine, Massachusetts General Hospital (JAE, DFMB, CAC), Boston, MA.
    Search for more papers by this author
  • David F.M. Brown MD,

    1. From the Highland General Hospital Department of Emergency Medicine/Alameda County Medical Center (AH), Oakland CA; Harvard Medical School (AW, DUH, SW, DFMB, CAC), Boston, MA; the Department of Medicine, Cambridge Health Alliance (AW, DUH, SW), Cambridge, MA; and the Department of Emergency Medicine, Massachusetts General Hospital (JAE, DFMB, CAC), Boston, MA.
    Search for more papers by this author
  • Carlos A. Camargo Jr. MD, DrPH

    1. From the Highland General Hospital Department of Emergency Medicine/Alameda County Medical Center (AH), Oakland CA; Harvard Medical School (AW, DUH, SW, DFMB, CAC), Boston, MA; the Department of Medicine, Cambridge Health Alliance (AW, DUH, SW), Cambridge, MA; and the Department of Emergency Medicine, Massachusetts General Hospital (JAE, DFMB, CAC), Boston, MA.
    Search for more papers by this author

  • Presented at the Society for Academic Emergency Medicine Annual Conference, Washington, DC, May 30, 2008.

Address for correspondence and reprints: Andrew Herring, MD; e-mail: andrew.a.herring@gmail.com.

Abstract

Background:  Emergency departments (EDs) are traditionally designed to provide rapid evaluation and stabilization and are neither staffed nor equipped to provide prolonged care. Longer ED length of stay (LOS) may compromise quality of care and contribute to delays in the emergency evaluation of other patients.

Objectives:  The objective was to determine whether ED LOS increased between 2001 and 2005 and whether trends varied by patient and hospital factors.

Methods:  This was a retrospective analysis of a nationally representative sample of 138,569 adult ED visits from the National Hospital Ambulatory Medical Care Survey (NHAMCS), 2001 to 2005. ED LOS was measured from registration to discharge.

Results:  Median ED LOS increased 3.5% per year from 132 minutes in 2001 to 154 minutes in 2005 (p-value for trend < 0.001). There was a larger increase among critically ill patients for whom ED LOS increased 7.0% annually from 185 minutes in 2001 to 254 minutes in 2005 (p-value for trend < 0.01). ED LOS was persistently longer for black/African American, non-Hispanic patients (10.6% longer) and Hispanic patients (13.9% longer) than for non-Hispanic white patients, and these differences did not diminish over time. Among factors potentially associated with increasing ED LOS, a large increase was found (60.1%, p-value for trend < 0.001) in the use of advanced diagnostic imaging (computed tomography [CT], magnetic resonance imaging [MR], and ultrasound [US]) and in the proportion of ED visits at which five or more diagnostic or screening tests were ordered (17.6% increase, p-value for trend = 0.001). The proportion of uninsured patients was stable throughout the study period, and EDs with predominately privately insured patients experienced significant increases in ED LOS (4.0% per year from 2001 to 2005, p-value for trend < 0.01).

Conclusions:  Emergency department LOS in the United States is increasing, especially for critically ill patients for whom time-sensitive interventions are most important. The disparity of longer ED LOS for African Americans and Hispanics is not improving.

Throughout the nation, emergency departments (EDs) are being called upon to take care of more and more patients as national rates of ED utilization steadily climb. In the United States from 1993 to 2003, there was a 26% increase in the number of ED visits, while 425 EDs closed their doors (9% decline), and hospitals reduced the total number of staffed inpatient beds by 198,000 (17% decline).1,2 Although many EDs that remained open expanded their capacity, most emergency physicians continue to report working crowded conditions.3 In their report on the future of emergency medicine, Hospital-based Emergency Care: At the Breaking Point, the Institute of Medicine called attention to ED crowding as a national threat to public health because of its association with decreased patient dissatisfaction,4–6 poor clinical outcomes,7–9 ambulance diversion,5,10,11 and diminished regional disaster response capacity.4,11–16

Emergency department length of stay (LOS) is a widely accepted factor in ED crowding and directly associated with a number of quality-of-care measures including decreased patient satisfaction, increased hospital LOS, increased morbidity among ventilated patients, and increased mortality among critically patients.7,9,17,18 Normally defined as the time elapsed from when a patient registers until the patient physically leaves the ED, ED LOS is not a direct measure of ED crowding, but is widely used as an easily quantifiable surrogate marker of ED crowding and an important component of ED quality assurance monitoring.7

In this study, we used nationally representative data on ED visits to investigate whether ED LOS is increasing and whether patient and hospital factors are associated with increasing ED LOS. We also examined whether black/African American and Hispanic patients experience longer ED LOS than non-Hispanic white patients and if these racial/ethnic disparities are changing over time.

Methods

Study Design and Data Collection Protocols

We analyzed data from the National Hospital Ambulatory Medical Care Survey (NHAMCS), an annual survey of U.S. ED visits conducted by the Centers for Disease Control and Prevention’s National Center for Health Statistics (CDC/NCHS). The NHAMCS uses a four-stage probability design to collect a nationally representative sample of all visits to EDs based in general and short-stay hospitals, excluding federal hospitals, hospital units of institutions, and hospitals with fewer than six staffed beds. U.S. Census Bureau workers train hospital staff to collect data on a random sample of ED patient visits during a randomly assigned 4-week period each year. The NHAMCS methodology is described in detail elsewhere.19 We used data for the years 2001–2005 (prior to 2001, ED LOS was not collected). Our analysis was exempted from human subjects review by the Massachusetts General Hospital institutional review board.

Study Setting and Population

We analyzed ED visits by adult patients 18 years of age and older. Many children are seen at dedicated pediatric EDs or dedicated areas within a given ED, and their ED LOS is dependent on factors distinct from adult patients. As a consequence, children under 18 years of age were excluded from our analyses. Data collection forms included information on patient disposition. NHAMCS included information on the ultimate patient disposition including admission to the hospital or admission to an intensive care or critical care unit (ICU/CCU). We designated a patient as “critically ill” if the ED visit resulted in admission to the ICU/CCU. Patients who left against medical advice or left without being seen were included in our analysis and considered as discharged.

Outcome Measures: Year of Visit and ED LOS

The NHAMCS records information on the day, month, and year of the patient visit. To prevent hospitals from collecting data during the same month each year, the sample of hospitals is randomly divided into 16 subsets and assigned 1 of 16 four-week reporting periods that rotate across each survey year. Consequently, each hospital is surveyed approximately once every 15 months.19

For NHAMCS, the CDC/NCHS calculated ED LOS for each visit from the time of a patient’s registration and the time of the patient’s discharge from the ED. ED LOS data were missing from 18,700 visit records (10.3% of all visits). Visits without information about ED LOS showed small but statistically significant differences in race/ethnicity, region, ownership, and metropolitan statistical area (MSA) designation from visits with information on ED LOS (data not shown). We excluded visits with missing ED LOS data from all subsequent analysis.

Covariates

Data collection forms include patient demographics, reasons for the visit, health care provider type, diagnostic/screening tests and imaging ordered, discharge diagnoses, and disposition. For NHAMCS, hospital staff classified patient race (white, black or African American, Asian/Pacific islander, Native American, other, multiple) and ethnicity (Hispanic, non-Hispanic) according to their observations and the medical record. We analyzed race/ethnicity using four mutually exclusive categories: non-Hispanic white, non-Hispanic black/African American, Hispanic, and other.

We considered a patient to be uninsured if his or her expected payment source was self-pay or no charge and analyzed payer status using five categories: private insurance, Medicare, Medicaid, uninsured, and other. We included whether the physician provider was an intern or resident in our analysis as an indicator of whether visit occurred at a teaching hospital. The initial health care provider (usually a triage nurse) seeing each patient assigned triage acuity using four levels of immediacy with which the patient should be seen: emergent (less than 15 minutes), urgent (15–59 minutes), semiurgent (60–119 minutes), and nonurgent (2–24 hours). We categorized computed tomography (CT), magnetic resonance (MR), and ultrasound (US) imaging as advanced imaging in our analyses. The use of diagnostic and screening testing was analyzed in two categories: zero to four ordered tests and five or more ordered tests.

We quantified safety-net burden by stratifying EDs according to their ED patient payer mix.20 In brief, we used the classification developed by Gardner et al.20 that categorizes EDs by the percentage of total visits for which Medicaid or uninsured was the payer status. EDs were categorized into three strata: <20% Medicaid/uninsured visits are considered “non–safety net,” 20 to 50% Medicaid/uninsured visits are considered “secondary safety net,” and >50% Medicaid/uninsured visits are considered “safety net.” NHAMCS obtains hospital characteristics such as ownership, region of the country, and MSA designation from a commercially available database updated annually by Verispan (Yardley, PA).

Data Analysis

We performed all analyses with STATA Version 10.0 (StataCorp, College Station, TX) and the svy package of commands using the probability weights and sample design variables (strata and primary sampling unit) provided by the CDC/NCHS for the NHAMCS. To describe the distribution of the untransformed data, we computed the median ED LOS and the interquartile range (IQR). To best describe the central tendency of ED LOS over time, we natural log-transformed ED LOS prior to regression analyses to account for data skewing due to outliers with very long ED LOS. The linear regression beta coefficient is reported as percent change in ED LOS (based on the mean of the log-transformed data). Predictors of ED LOS were analyzed by sequentially introducing covariates into our multivariate linear regression model, using natural log-transformed ED LOS as the outcome. We preformed separate multivariate linear regressions for all patients and for critically ill patients

We analyzed changes in ED LOS times between 2001 and 2005 using bivariate linear regressions with the year of visit as a continuous predictor of natural log-transformed ED LOS time. Differences in time trends between racial/ethnic groups, men and women, urban and nonurban EDs, and patients by expected payer source were individually analyzed by creating interaction terms (defined as the product of the visit year and covariate of interest) that were introduced with natural log-transformed ED LOS in our linear regression models. We analyzed time trends in the proportional distribution of selected covariates found to be strongly associated with change in ED LOS using logistic regression models. Results are presented as percentages with respect to a reference group with 95% confidence intervals (CIs). A two-tailed p < 0.05 was considered statistically significant.

Results

Between 2001 and 2005, NHAMCS collected data for 138,569 adult ED visits; 21,665 (16%) of these unweighted visits resulted in a hospital admission, including 2,160 (1.7%) admissions to an ICU or CCU. Using the survey weights, we estimate that there were a total of 420 million adult ED visits over the 5-year period, including 65.1 million ED visits resulting in admission to the hospital and 6.9 million ED visits leading to admission to an ICU or CCU. ED visits resulting in admission to an ICU or CCU were more likely to occur at a teaching hospital, more likely to have health insurance (especially Medicare), and more likely to be assigned a triage acuity of urgent emergency. Differences in ED visit characteristics between the two groups were otherwise small (Table 1).

Table 1. 
Patient and Hospital Characteristics of Adult U.S. ED Visits, NHAMCS 2001–2005
Characteristic All Patients (n = 138,569)*Critically Ill Patients† (n = 2,160)*
%Median ED LOS (IQR)‡%Median ED LOS (IQR)
  1. CCU = critical care unit; ICU = intensive care unit; IQR = interquartile range; LOS = length of stay; NHAMCS = National Hospital Ambulatory Medical Care Survey.

  2. *= unweighted number of observations in the data set for years 2001–2005. Proportions are calculated using survey weights.

  3. †ED visits that resulted in admission to an ICU or CCU.

  4. ‡Self-pay or no charge.

  5. §See Methods.

All patients100.0146 (84–242)1.7222 (144–328)
Patient age, yr
 18–2516.3130 (75–218)17.1222 (144–323)
 26–6463.9154 (90–257)33.1223 (146–330)
 65 and older19.8183 (112–287)49.7220 (139–328)
Patient sex
 Female44.0151 (87–249)53.2225 (148–328)
 Male56.0140 (80–233)47.8216 (140–330)
Patient race/ethnicity
 White, non-Hispanic67.1140 (80–230)74.8213 (139–315)
 Black or African American, non-Hispanic20.0158 (92–261)14.7250 (166–390)
 Hispanic10.3167 (95–282)7.3248 (166–390)
 Other2.7151 (82–251)3.2200 (133–293)
Patient insurance type
 Private insurance40.0129 (75–212)28.5224 (148–312)
 Medicare22.6175 (106–279)46.0208 (134–313)
 Medicaid17.2124 (70–210)14.8249 (151–430)
 Uninsured18.1132 (75–225)8.8191 (110–265)
 Other/unknown4.6110 (66–185)1.9256 (160–256)
Clinician type
 Resident or intern9.2215 (128–343)17.8255 (137–387)
Patient triage acuity
 Less than 15 minutes18.3157 (91–257)53.8200 (133–305)
 15 to 60 minutes34.9154 (89–253)28.6265 (182–375)
 61 to 120 minutes18.8142 (82–236)4.0304 (160–385)
 More than 2 hours11.0123 (70–206)1.5237 (187–535)
 Unknown17.1140 (76–232)12.1180 (120–280)
Patient disposition
 Discharged85.5135 (78–220)  
 Admitted to hospital15.5253 (161–380)  
Clinician ordered diagnostic tests
 Advanced imaging12.5249 (169–363)23.9258 (162–384)
 Five or more diagnostic tests30.1226 (153–335)82.7222 (145–330)
Hospital setting
 Rural16.8105 (62–166)15.7174 (120–257)
 Urban83.2157 (90–259)84.3232 (148–341)
Hospital owner
 Nonprofit72.5148 (85–243)77.1235 (150–338)
 Government17.9143 (80–217)14.8182 (119–310)
 Proprietary9.6140 (82–228)8.1198 (130–255)
Hospital location
 Northeast20.2151 (84–260)20.6253 (155–406)
 Midwest22.7139 (79–230)25.4188 (124–298)
 South38.0144 (84–235)34.8232 (151–314)
 West18.2157 (90–258)19.2216 (148–351)
ED payer mix§
 Non–safety net20.4147 (82–231)20.7223 (147–318)
 Secondary safety net65.4142 (83–231)67.0211 (137–310)
 Safety net14.2168 (94–290)12.3254 (156–465)

For all patients, median ED LOS time was 146 minutes (Table 1). Median ED LOS increased with advancing age and women had longer median ED LOS than men. The median ED LOS was shorter among whites than among black/African Americans or Hispanics. Patients seen in urban EDs had longer median ED LOS than those seen in rural EDs. The median ED LOS was longer for admitted patients (253 minutes) than for discharged patients (135 minutes). For critically ill patients, the median ED LOS was 76 minutes longer than for the all patient group (Table 1). Although the mean age of critically ill patients was higher than that of the all-patient group (62.6 years vs. 45.3 years), we did not find large variations in median ED LOS by age among critically ill patients. As with the all patient group, the median ED LOS of critically ill patients was shorter for whites than for black/African Americans or Hispanics. We found longer median ED LOS for critically ill patients in the Northeast and South than in the West and Midwest.

In multivariate analysis of all patients (Table 2), longer ED LOS was found among black/African American (10.6% longer) and Hispanic (13.9% longer) patients, compared to non-Hispanic white patients. ED visits at which advanced imaging was ordered had a 37.6% longer ED LOS than those during which no advanced imaging was ordered. ED LOS was 46.4% longer for ED visits with five or more diagnostic or screening tests ordered. Urban EDs (29.6%) and safety net EDs (10.3%) had longer ED LOS. Longer ED LOS for black/African American and Hispanic patients persisted among patients admitted to CCUs and ICUs. Among critically ill patients, ED LOS was 14.3% longer for black/African American patients and 17.9% longer for Hispanic patients than for white patients. In contrast to the all-patient group, neither ED location in an urban setting nor ED safety net status was associated with change in ED LOS for critically ill patients.

Table 2. 
Multivariate Adjusted Percentage Change in ED LOS by Selected Patient and Hospital Characteristics, NHAMCS 2001–2005
CharacteristicAdjusted Percentage Change
All PatientsCritically Ill Patients*
% (95% CI)p-value% (95% CI)p-value
  1. CT = computed tomography; ICU = intensive care unit; IQR = interquartile range; LOS = length of stay; MR = magnetic resonance; NHAMCS = National Hospital Ambulatory Medical Care Survey; US = ultrasound.

  2. *ED visit that resulted in admission to an ICU or CCU.

  3. †Self-pay or no charge.

  4. ‡CT, MR, or US.

  5. §See Methods.

Patient age, yr
 18–25[Reference] [Reference] 
 26–648.1 (6.6 to 9.6)<0.001−4.4 (−18.6 to 9.9)0.54
 65 and older9.1 (6.7 to 11.5)<0.001−1.3 (−16.6 to 14)0.87
Patient sex
 Female5.0 (3.7 to 6.2)<0.001−0.4 (−11 to 10.1)0.94
 Male[Reference] [Reference] 
Patient race/ethnicity
 White, non-Hispanic[Reference] [Reference] 
 Black or African American, non-Hispanic10.6 (8.1 to 13.1)<0.00114.3 (0.8 to 27.9)0.04
 Hispanic13.9 (10.6 to 17.2)<0.00117.9 (0.6 to 35.2)0.04
 Other−1.2 (−8.5 to 6)0.745.8 (−14.4 to 25.9)0.58
Patient insurance type
 Private insurance[Reference] [Reference] 
 Medicare3.8 (1.5 to 6.2)0.001−12.1 (−25.7 to 1.5)0.08
 Medicaid1.9 (−0.5 to 4.2)0.027.2 (−11 to 25.4)0.44
 Uninsured†0.9 (−0.8 to 2.6)0.30−29.2 (−53 to −5.3)0.02
 Other/unknown−8.8 (−12.4 to −5.1)<0.0015.8 (−26.6 to 38.1)0.73
Clinician type
 Resident or intern24.8 (19.7 to 29.9)<0.00112.8 (−2.1 to 27.7)0.092
Patient triage acuity
 Less than 15 minutes[Reference] [Reference] 
 15 to 60 minutes5.2 (1.9 to 8.5)0.00220.0 (6.2 to 33.7)0.004
 61 to 120 minutes5.5 (1.7 to 9.3)0.00517.5 (−5.0 to 40)0.13
 More than 2 hours−2.6 (−7.3 to 2.1)0.2824.0 (−0.5 to 48.6)0.06
 Unknown2.1 (−3 to 7.2)0.42−4.3 (−24.0 to 15.4)0.67
Patient disposition
 Discharged[Reference]   
 Admitted29.6 (25.6 to 33.6)<0.001  
Clinician ordered imaging
 No advance imaging‡[Reference] [Reference] 
 Advanced imaging 37.6 (35.6 to 39.7)<0.00119.1 (9.4 to 28.8)<0.001
Clinician-ordered tests
 Four or fewer[Reference] [Reference] 
 Five or more46.4 (44.2 to 48.5)<0.00111.3 (−5.5 to 28.1)0.19
Hospital setting
 Rural[Reference] [Reference] 
 Urban29.6 (24 to 35.2)<0.0014.6 (−14.4 to 23.7)0.632
Hospital owner
 Nonprofit[Reference] [Reference] 
 Government−0.5 (−6.5 to 5.4)0.86−26.7 (−44 to −9.4)0.002
 Proprietary−4.2 (−10.5 to 2)0.18−15.9 (−33.4 to 1.6)0.07
Hospital location
 Northeast6.1 (0.7 to 11.4)0.0321.6 (1.9 to 41.3)0.03
 Midwest[Reference] [Reference] 
 South7.9 (2.1 to 13.7)0.0115.7 (−3.6 to 35)0.112
 West7.4 (1.1 to 13.8)0.0214.2 (−9.1 to 37.4)0.232
ED payer mix§
 Non–safety net[Reference] [Reference] 
 Secondary safety net0.9 (−3.2 to 5.1)0.66−7.2 (−23.5 to 9.0)0.38
 Safety net10.3 (0.4 to 16.5)0.00111.3 (−14.7 to 37.3)0.39

Change in ED LOS Over Time

As shown in Figure 1, the median ED LOS increased over time for all patients and for critically ill patients. Unadjusted bivariate and adjusted multivariate annual percentage change in ED LOS by selected patient groups is presented in Table 3. Overall, the unadjusted increase in ED LOS was 3.5% per year, from 132 minutes in 2001 to 154 minutes in 2005. The ED LOS for critically ill patients increased 7.0% per year from 185 minutes to 254 minutes. For patients admitted to non-CCUs, there was no significant ED LOS time trend. ED LOS for discharged patients increased 3.5% per year. When we introduced interaction terms into our linear regression models, we found no differentials in ED LOS time trend by use of diagnostic or screening testing, sex, race/ethnicity, hospital location, or hospital ownership. However, uninsured patients admitted to the hospital had a 5.6% more rapid ED LOS increase of 6.1% per year (p-value for interaction = 0.03).

Figure 1.

 Median emergency department length of stay (ED LOS) by patient type, National Hospital Ambulatory Medical Care Survey (NHAMCS) 2001–2005.

Table 3. 
Annual Percentage Change in ED Length of Stay for Selected Patient Groups, NHAMCS 2001–2005
Patient GroupMedian ED LOS (IQR)
20012005Annual % change*p-value for trend
  1. IQR = interquartile range; LOS = length of stay; NHAMCS = National Hospital Ambulatory Medical Care Survey.

  2. *Annual percentage change calculated by linear regression as described under Methods.

  3. †ED visits that resulted in admission to an ICU or CCU.

All visits132 (75 to 220)154 (88 to 259)3.5 (1.6 to 5.3)<0.001
Discharged125 (72 to 202)143 (83 to 236)3.5 (1.6 to 5.3)<0.001
Admitted230 (147 to 356)262 (164 to 390)1.3 (−1.4 to 4.1)0.34
Critically ill†185 (125 to 291)254 (160 to 378)7.0 (2.1 to 11.9)0.005

We explored potential explanations for the observed increases in ED LOS and found that the proportion of ED visits at which advanced imaging was ordered increased from 9.8% (95% CI = 9.0% to10.7%) in 2001 to 15.8% (95% CI = 14.6% to 17.1%) in 2005 (a 60.1% increase, p-value for trend = 0.001). The proportion of ED visits at which five or more diagnostic or screening tests were ordered increased from 26.7% (95% CI = 24.9% to 28.6%) to 31.4% (95% CI = 29.0% to 33.8%) over the study time period (a 17.6% increase, p-value for trend = 0.001). Finally, the proportion of ED visits by Hispanic patients increased by 42.7%, from 8.7% (95% CI = 7.0% to 10.4%) in 2001 to 12.4% (95% CI = 10.3% to 14.5%) in 2005. There was no change in the proportions of patients admitted to the hospital, triage acuity, or hospital setting, nor did we find significant change in the proportions of visits either by uninsured patients or by patients over 65 years old. Safety net EDs with a payer mix of greater than 50% uninsured or Medicaid had a high (168 minutes) but stable ED LOS, whereas non–safety net EDs with less than 20% uninsured or Medicaid patients had an increase in their ED LOS 4.0% (95% CI = 0.72% to 7.3%) per year, from 127 minutes in 2001 to 162 minutes in 2005.

Discussion

Although it is a widely used indicator of the quality of ED-based care and the presence of ED crowding,20–22 we did not identify any previous study that analyzed national data on time trends in ED LOS. Our results show that the LOS for patients presenting to U.S. EDs increased by 22 minutes (17%) between 2001 and 2005 and that patients admitted to ICUs or CCUs had an increase that was more than twice as large (69 minutes, 37%). The concerning racial/ethnic disparity of longer median ED LOS among black/African American and Hispanic patients than among white patients remains large and is not improving.

Prolonged ED LOS is directly associated with patient dissatisfaction,6 ambulance diversion,10 poor patient outcomes,23 increased adverse events for patients with non-ST elevation myocardial infarction,14 increased inpatient stays,18 and increased mortality.9,16,24,25 The worse outcomes associated with prolonged ED LOS may arise from crowded ED conditions, long waits to be seen by a physician, or insufficient medical and nursing attention to admitted patients awaiting transfer to an inpatient hospital bed. Moreover, EDs commonly lack the capacity to provide the focused one-on-one care often required to optimally manage critically ill patients for extended periods of time, leading to delays in the initiation of definitive treatment and more frequent adverse events.5,26,27 Our finding that the trend of increasingly long stays for patients in U.S. EDs disproportionately affects critically ill patients is consistent with reports that a frequent source of ED crowding is the lack of staffed, inpatient beds for critically ill patients and comes at a time when the U.S. population is aging and the need for critical care initiated in the ED is increasing.5,14,27–30 Longer ED LOS patients among patients ultimately admitted to an ICU/CCU increases the exposure of these vulnerable patients to the potentially higher risk clinical environment of a busy ED. Because most EDs are neither equipped nor staffed to provide extended critical-level care, delays in transport to an inpatient critical care bed could lead to increased morbidity and mortality among critically ill patients presenting to U.S. EDs.

In the present study, we identified several factors that may be contributing to increasing ED LOS, including a significant trend toward ordering more diagnostic or screening tests and more frequent use of advanced imaging techniques (CT, MR, US). Although the observed increase in the utilization of imaging is particularly striking (60%), it is not possible from our data to determine if this increase is a result of patients spending more time in the ED because of crowding or is causally related to increasing ED LOS. Nevertheless, there is a clear trend toward more diagnostic testing and imaging occurring during U.S. ED visits, and the efficiency with which these services are integrated into ED-based care could potentially have a significant impact on ED LOS.

We found a trend toward an increasing proportion of ED visits by Hispanic patients who also tend to have longer ED LOS. The explanation for this finding is unclear. Some authors have proposed that the need for additional language services such as interpreters may tend to increase ED LOS for monolingual Spanish-speaking patients. It may also be that EDs that serve predominantly minority patients experience more crowded conditions, a possibility that we were not able to adequately control for due to limitations in our data.20,31 Aging of the American population likely did not have a significant impact on the observed increases in ED LOS, as we observed no significant change in the proportion of ED visits by patients over 65 years.

Our data suggest that increasing lack of health insurance in the general population likely did not have a significant direct impact on ED LOS increases. We did not find an increase in the proportion of ED visits by patients without health insurance during the study period, nor did we find that safety net EDs with large numbers of uninsured patients have been disproportionately affected by ED LOS increases. To the contrary, increases in ED LOS were largest at non–safety net EDs with predominately insured patients where ED LOS appears to be steadily “catching up” to the longer ED LOS common at safety net EDs. Because of limitations in our data, socioeconomic status was not directly available for analysis. This limitation notwithstanding, it is reasonable to assume that non–safety net EDs with a primarily privately insured payer mix generally serve patients of higher socioeconomic status with a smaller proportion of nonpoor patients. The larger increases in ED LOS that we observed at non–safety net EDs are consistent with recent evidence that from 1997 to 2003 there was a disproportionate increase in ED utilization among nonpoor individuals with a usual source of primary care.32

The ED plays a crucial role in U.S. society, both providing care to acutely ill and injured patients and serving as a major point of entry into the health care system for uninsured patients. In 1986, the U.S. Congress enacted the Emergency Medical Treatment & Labor Act (EMTALA) that requires EDs to provide stabilizing care to all presenting patients, regardless of their ability to pay. Unfortunately, EMTALA does not include provisions to adequately compensate hospitals for this care, leading some authors to describe EMTALA as an unfunded mandate.33,34 EDs are a major point of access to health care for the more than 47 million uninsured. In fact, more than 60% of hospitalizations of uninsured patients originate in the ED.35,36 Uncompensated or insufficiently compensated care is a financial liability for many health care institutions and may create a disincentive to offer or expedite ED services.

Limitations

Diversion status of the ED at the time of visit is not available in the NHAMCS public use file, nor does NHAMCS offer information on other direct indicators of ED crowding such as the percentage occupancy of ED beds. NHAMCS does not document “boarding” per se, making it impossible to distinguish between time spent receiving appropriate evaluation from time spent awaiting an inpatient bed. For 10% of all patients, ED admission and/or discharge time was not recorded. However, differences in the patient and hospital characteristics of the missing records from those with ED LOS recorded were small, making a major bias in our analysis unlikely. Finally, we were not able to evaluate the effect of several important socioeconomic indicators (such as income) on ED LOS.

Conclusions

We observed a steady increase both in the amount of time adult patients spend in the ED and in the volume of diagnostic testing done for each patient. Our results should raise concerns for the capacity of U.S. EDs to adequately cope with simultaneous trends toward an increased volume of patients seeking care in the ED, patients spending more time in the ED with each visit, and patients receiving increasingly more diagnostically intensive care. Indeed, our results concur with a recent analysis that found a 4.1% annual increase in wait times to see an ED physician from 1997 to 2004; together these findings suggest increasingly crowded conditions in U.S. EDs.37 Reforms that reallocate national resources to increase funding for care originating in the ED are urgently needed to maintain access to this essential population health resource.

Ancillary