Characteristics and temporal trends of “left before being seen” visits in US Emergency Departments, 1995–2002

The Journal of Emergency Medicine, Vol. 32, No. 2, pp. 211–215, 2007 Copyright © 2007 Elsevier Inc. Printed in the USA. All rights reserved 0736-4679/07 $–see front matter

doi:10.1016/j.jemermed.2006.05.045

Administration of Emergency Medicine

CHARACTERISTICS AND TEMPORAL TRENDS OF “LEFT BEFORE BEING SEEN” VISITS IN US EMERGENCY DEPARTMENTS, 1995–2002 Benjamin C. Sun,

MD, MPP,*

Emily Spilseth Binstadt, MD, MPH,† Andrea Pelletier, and Carlos A. Camargo Jr, MD, DrPH†‡

MSc, MPH,‡

*Department of Medicine, West Los Angeles Veterans Affairs Medical Center, Los Angeles, California; †Harvard Affiliated Emergency Medicine Residency and ‡Department of Emergency Medicine, Massachusetts General Hospital, Boston, Massachusetts Address for Correspondence: Benjamin Sun, MD, MPP, West Los Angeles Veterans Affairs Medical Center, Office 3214A, Building 500, 11301 Wilshire Blvd, Los Angeles, CA 90073

e Abstract—The purpose of this study was to describe nationally representative characteristics and temporal trends in “left before being seen” (LBBS) visits in US emergency departments (EDs). The ED portion of the federal National Hospital Ambulatory Medical Care Survey, 1995–2002, was analyzed. Of the 810.6 million ED visits during the 8-year study period, an estimated 11.4 million (1.41%, 95% confidence interval [CI] 1.30 –1.52) had an LBBS disposition. The number and proportion of LBBS visits have increased over time, from 1.1 million visits in 1995 (1.15%, 95% CI 0.95–1.35) to 2.1 million visits in 2002 (1.92%, 95% CI 1.67–2.17). LBBS patients were more likely to be younger, non-White, Hispanic, urban, and uninsured compared to non-LBBS patients. The number and proportion of LBBS visits have increased over time. LBBS visits disproportionately affect vulnerable populations. These findings suggest that recent strains on the US ED system are adversely affecting healthcare quality and access. © 2007 Elsevier Inc.

adversely affects all patients by delaying emergency care, but it also affects another group of ED patients: those who “leave before being seen” (LBBS) by a physician. Single center studies suggest that LBBS patients have high levels of morbidity, and LBBS seems to increase with ED crowding (1– 6). Furthermore, vulnerable groups, such as minority and uninsured patients, have higher rates of LBBS compared to other patients (1,2,5). Thus, LBBS may be important as a quality of care measure, a proxy for the effects of ED overcrowding, and an indicator of health care access for vulnerable populations. Recent strains on the US ED system suggest that the proportion of LBBS visits may have increased over time. The existing LBBS literature is based on single-center/ regional, single time-point studies, and we are unaware of nationally representative data that address temporal trends in LBBS visits (1– 8). Using nationally representative data from the 1995– 2002 National Hospital Ambulatory Medical Care Survey (NHAMCS), we describe characteristics and temporal trends in LBBS visits. We hypothesized that the proportion of LBBS visits has increased over time, and that this finding cannot be explained by increases in low-acuity LBBS visits. We also hypothesized that mi-

e Keywords—Left before being seen; NHAMCS; ED crowding

INTRODUCTION Emergency department (ED) crowding has become a national problem in the past several years. Crowding

Administration of Emergency Medicine is coordinated by Eugene Kercher, MD, of Kern Medical Center, Bakersfield, California and Richard F. Salluzzo, MD, of Conemaugh Meridian Health Group, Johnstown, Pennsylvania

RECEIVED: 28 February 2005; FINAL ACCEPTED: 22 May 2006

SUBMISSION RECEIVED:

11 October 2005; 211

212

B. C. Sun et al.

ple of non-institutional, general, and short-stay hospitals (excluding federal, military, and Veterans Affairs Hospitals) in the 50 US States and the District of Columbia.

nority and uninsured patients would have disproportionately high LBBS rates.

METHODS

Patient Selection

Study Design

A four-stage sampling strategy was used, covering geographic Primary Sampling Units (PSUs), hospitals within PSUs, EDs within hospitals, and patients within EDs. Hospitals were stratified by ownership, size, and region, and they were sampled with a probability proportional to ED volume. Over 400 hospitals per year were sampled.

We performed a secondary analysis of the ED component of the 1995–2002 NHAMCS. This cross-sectional survey contains data on 209,000 ED visits, which represent approximately 811 million ED visits over the 8-year period. Analysis of the NHAMCS dataset was approved by the Human Research Committee.

Measurements Study Setting and Population Data forms included up to three patient reasons for the visit, disposition (which included LBBS), and demographic information including age, gender, race, ethnicity, region of the United States, urban status, insur-

NHAMCS is directed by the Centers for Disease Control and Prevention, and the National Center for Health Statistics (NCHS) (9). It is an annual, stratified, probability sam-

Table 1. Characteristics of LBBS and Non-LBBS Patients, 1995–2002 95% CI

Age (years) ⬍20 20–29 30–39 ⬎40 Gender Female Male Race White Black Other Ethnicity Non-Hispanic Hispanic Region Northeast Midwest South West Urban Status Urban Non-Urban Insurance Private Public Other Self-pay Unknown Illness severity† Urgent Non-urgent

95% CI

% Total of LBBS patients*

Low

High

% Total of non-LBBS patients*

Low

High

30.8% 23.7% 19.0% 26.4%

27.9% 20.4% 16.3% 24.4%

33.8% 27.0% 21.7% 28.5%

29.2% 16.5% 15.5% 38.8%

28.9% 16.2% 15.2% 38.4%

29.6% 16.8% 15.8% 39.2%

52.3% 47.7%

49.0% 44.9%

55.6% 50.5%

52.9% 47.1%

52.5% 46.7%

53.3% 47.5%

64.8% 32.4% 2.8%

62.3% 28.9% 2.0%

67.3% 35.8% 3.6%

76.3% 20.9% 2.7%

76.0% 20.6% 2.6%

76.7% 21.3% 2.8%

65.7% 13.1%

63.1% 10.9%

68.2% 15.3%

70.9% 9.6%

70.6% 9.4%

71.3% 9.8%

17.6% 19.4% 46.9% 16.1%

15.3% 17.1% 43.8% 13.8%

20.0% 21.6% 50.1% 18.3%

19.4% 25.6% 37.2% 17.9%

19.1% 25.2% 36.8% 17.6%

19.7% 25.9% 37.6% 18.2%

88.0% 12.0%

85.5% 10.4%

90.6% 13.5%

77.9% 22.1%

77.6% 21.8%

78.3% 22.4%

27.3% 26.7% 9.7% 24.5% 11.7%

24.7% 23.6% 5.3% 22.3% 9.8%

30.0% 29.8% 14.1% 26.7% 13.7%

39.7% 30.6% 9.0% 15.8% 4.8%

39.3% 30.3% 8.7% 15.5% 4.6%

40.1% 31.0% 9.3% 16.1% 5.0%

28.1% 71.9%

24.9% 69.6%

31.2% 74.3%

50.0% 50.0%

49.5% 49.7%

50.4% 50.4%

* Strata for variables may not sum to 100% due to missing data. † Coding changed after 1997; please see text for details.

213

110,200,000 2,116,129 1.92 1.12

810,727,341 11,416,901 1.41 1.28

ance source, and year of presentation. Visits were analyzed by the acuity of the medical problem (“urgent/ emergent” or “walk-in”) recorded at triage in the 1995– 1996 data. NHAMCS coding for acuity changed in 1997. We coded visits from 1997–2002 as “urgent/emergent” if the field “immediacy with which the patient should be seen” was recorded as “less than 15 minutes” or “15– 60 minutes,” and as “non-urgent” if recorded as “⬎1–2 hours.” We also examined ED visit rates by time of presentation, day of week, and season (January–March, April–June, July–September, October–December). The US Bureau of the Census regional staff oversees data collection. Visit sampling and collection were performed by hospital staff, and review of data collection was performed by a Bureau of Census field supervisor. Data abstraction was performed centrally by experienced NCHS coders. Quality control included computer checks to identify inconsistencies with value ranges, subsample assessment for accuracy in medical and drug coding, and adjudication by NCHS for ambiguous or illegible responses for fields including reasons for visit and diagnosis. The non-response rate for items was ⬍5%, and error rates were less than 0.1% for items requiring non-medical coding.

108,000,000 1,870,543 1.73 0.89 102,800,000 1,337,027 1.3 1.01

107,500,000 1,593,355 1.48 1.67

2000 1999

2001

2002

All years

Left before Being Seen

100,400,000 1,371,738 1.37 1.13 * Weighted chi-square for trend; p ⬍ 0.05. † Weighted chi-square for trend; p ⫽ 0.28.

90,346,724 1,019,959 1.13 1.24 96,545,042 1,109,766 1.15 1.37

94,935,575 998,384 1.05 1.87

1996 1995

Total ED visits* Total LBBS %LBBS %Missing disposition†

Table 2. Temporal Trends

1997

1998

Data Analysis We determined point estimates and 95% confidence intervals (CI) for ED frequency of LBBS patients. Nationally representative estimates were determined using NCHS-assigned patient weights, which adjust for selection probability and non-response rates. Relative standard errors were estimated using methodology suggested by the NCHS (personal communication, Schappert S, NCHS). US population estimates for calculating annual population frequency were based on US Bureau of Census data. Characteristics of LBBS and non-LBBS patients were compared using weighted chi-square analysis. Temporal trends were assessed with weighted chisquare test for trend. Data management and analysis were performed with Stata software (version 7.0, Stata Corporation, College Station, TX).

RESULTS Over the 8-year study period, there were 3020 LBBS visits, which represented an estimated 11,400,000 (95% CI 10,500,000 –12,300,000) LBBS visits to United States Emergency Departments. This represents 5.2 LBBS visits (95% CI 4.8 –5.6) per 1000 persons in the general population and 14.1 (95% CI 13.0 –15.2) per 1000 ED visits. Disposition was missing in 1.3% (95% CI 0.5–2.1) of visits.

214

B. C. Sun et al.

Figure 1. ED visits and % LBBS over time.

Characteristics of LBBS and non-LBBS patients are provided in Table 1. We assessed for demographic differences by inspecting 95% confidence intervals for each characteristic. For example, there are non-overlapping 95% confidence interval estimates for patients in the South who were LBBS (44 –50%) and non-LBBS (37– 38%), suggesting that LBBS visits are more frequent in the South compared to non-LBBS visits. Compared to non-LBBS patients, LBBS patients tended to be younger, non-White, Hispanic, treated in the South, urban, uninsured, and triaged as “non-urgent.” There were no differences among LBBS and non-LBBS patients in presenting complaint, or presentation by hour, day, or season (data not shown). Temporal trends in total ED visits, LBBS, and missing disposition data are displayed in Table 2. Both total ED visits and percent of ED visits that were LBBS have increased between 1995 and 2002. These trends are graphically displayed in Figure 1. The percentage of patients with missing disposition data has not changed over time (p ⫽ 0.28). There were no important changes over time in triage severity, race, age, insurance status, urban status, or Hispanic ethnicity in any of the subgroups over the study period (data not shown). DISCUSSION To our knowledge, this is the only national, populationbased study describing the characteristics and temporal

trends of ED LBBS visits. Using recent data from a federal survey, we demonstrate that LBBS visits have increased as a proportion of total ED visits over time. We also note that the proportion of LBBS patients with an “urgent/emergent” triage designation has not changed over time, and that LBBS visits are disproportionately represented by minority and uninsured patients. These findings have two important health policy implications. First, these data suggest that US EDs may fail to provide care to increasing numbers of patients with acute illness. We estimate an increase of 1 million LBBS visits from 1995 to 2002. Although we were unable to assess clinical outcomes, two single center studies suggest that a subset of LBBS patients have significant morbidity and require urgent inpatient care (1,2). Second, we demonstrate that minority and uninsured patients disproportionately represent LBBS visits. These findings suggest that vulnerable groups have reduced access to timely ED care compared to other patients. Other studies in public hospitals and metropolitan areas with a high proportion of uninsured patients report substantially higher rates of LBBS than noted in our report (1,2,5). Further research is needed to determine if a subset of these patients could be safely treated in a primary care setting. NHAMCS is the only nationally representative dataset that provides descriptive information about US ED visits, and it is a valuable tool for studying the epidemiology of emergency conditions. However, it has several limitations.

Left before Being Seen

First, we are unable to cross-validate the accuracy of disposition coding. However, LBBS coding for NHAMCS remained constant during the study period, and biases in coding are unlikely to affect our finding that the proportion of LBBS patients increased over time. Second, there are missing disposition data for some visits. Missing data may introduce bias into our LBBS estimates. However, we found no changes in proportion of patients with missing disposition data over time, nor did we find changes in patient characteristics with missing disposition data over time. These sensitivity analyses support our conclusion that the proportion of LBBS patients has increased over time, and that this finding is unlikely to be biased by missing disposition data. Third, there was a change in coding of illness severity after 1997, which may confound our ability to assess trends in severity of LBBS visits. In a sensitivity analysis, however, we found no change in illness severity of LBBS patients from 1997–2002. Fourth, LBBS may be related to wait times. We did not have data on wait times, and future investigations should explore this potential causeeffect relationship. Finally, the cross-sectional design of NHAMCS does not allow us to track clinical outcomes in LBBS patients. Our findings suggest that nationally representative studies to prospectively track outcomes in LBBS patients are urgently needed. Despite limitations in the NHAMCS cross-sectional design, our findings strongly suggest an increasing proportion of LBBS visits over time. When considered in the context of national ED crowding, shortage of on-call specialists, and ambulance diversions, our results provide further evidence of overload in the US ED system.

215

The national increase in LBBS visits has adverse implications for patient safety and healthcare access, and the design of solutions deserves the urgent attention of policymakers and researchers. Acknowledgment—Dr. Camargo is supported, in part, by an Emergency Medicine Foundation Center of Excellence Award (Dallas, TX).

REFERENCES 1. Baker DW, Stevens CD, Brook RH. Patients who leave a public hospital emergency department without being seen by a physician. Causes and consequences. JAMA 1991;266:1085–90. 2. Bindman AB, Grumbach K, Keane D, Rauch L, Luce JM. Consequences of queuing for care at a public hospital emergency department. JAMA 1991;266:1091– 6. 3. Hobbs D, Kunzman SC, Tandberg D, Sklar D. Hospital factors associated with emergency center patients leaving without being seen. Am J Emerg Med 2000;18:767–72. 4. Kyriacou DN, Ricketts V, Dyne PL, McCollough MD, Talan DA. A 5-year time study analysis of emergency department patient care efficiency. Ann Emerg Med 1999;34:326 –35. 5. Stock LM, Bradley GE, Lewis RJ, Baker DW, Sipsey J, Stevens CD. Patients who leave emergency departments without being seen by a physician: magnitude of the problem in Los Angeles County. Ann Emerg Med 1994;23:294 – 8. 6. McNamara KJ. Patients leaving the ED without being seen by a physician: is same-day follow-up indicated? Am J Emerg Med 1995;13:136 – 41. 7. Fernandes CM, Daya MR, Barry S, Palmer N. Emergency department patients who leave without seeing a physician: the Toronto Hospital experience. Ann Emerg Med 1994;24:1092– 6. 8. Weissberg MP, Heitner M, Lowenstein SR, Keefer G. Patients who leave without being seen. Ann Emerg Med 1986;15:813–7. 9. 2000 NHAMCS Micro-Data File Documentation. Hyattsville, MD: National Center for Health Statistics; 2000.