- Top of page
Objectives: The objective was to estimate the national left-without-being-seen (LWBS) rate and to identify patient, visit, and institutional characteristics that predict LWBS.
Methods: This was a retrospective cross-sectional analysis using the National Hospital Ambulatory Medical Care Survey (NHAMCS) from 1998 to 2006. Bivariate and multivariate analyses were performed to identify predictors of LWBS.
Results: The national LWBS rate was 1.7 (95% confidence interval [CI] = 1.6 to 1.9) patients per 100 emergency department (ED) visits each year. In multivariate analysis, patients at extremes of age (<18 years, odds ratio [OR] = 0.80, 95% CI = 0.66 to 0.96; and ≥65 years, OR = 0.46, 95% CI = 0.32 to 0.64) and nursing home residents (OR = 0.29, 95% CI = 0.08 to 1.00) were associated with lower LWBS rates. Nonwhites (black or African American (OR = 1.41, 95% CI = 1.22 to 1.63) and Hispanic (OR = 1.25, 95% CI = 1.04 to 1.49), Medicaid (OR = 1.47, 95% CI = 1.27 to 1.70), self-pay (OR = 1.96, 95% CI = 1.65 to 2.32), or other insurance (OR = 2.09, 95% CI = 1.74 to 2.52) patients were more likely to LWBS. Visit characteristics associated with LWBS included visits for musculoskeletal (OR = 0.70, 95% CI = 0.57 to 0.85), injury/poisoning/adverse event (OR = 0.65, 95% CI = 0.53 to 0.80), and miscellaneous (OR = 1.56, 95% CI = 1.19 to 2.05) complaints. Visits with low triage acuity were more likely to LWBS (OR = 3.59, 95% CI = 2.81 to 4.58), whereas visits that were work-related were less likely to LWBS (OR = 0.19, 95% CI = 0.12 to 0.29). Institutional characteristics associated with LWBS were visits in metropolitan areas (OR = 2.11, 95% CI = 1.66 to 2.70) and teaching institutions (OR = 1.33, 95% CI = 1.06 to 1.67).
Conclusions: Several patient, visit, and hospital characteristics are independently associated with LWBS. Prediction and benchmarking of LWBS rates should adjust for these factors.
Patients who present to an emergency department (ED) and leave without being seen (LWBS) by a physician are a safety concern.1–6 These patients may be severely ill and experience adverse events as a result of lacking or delayed ED treatment.7–10 One study found that 11% of LWBS patients required hospitalization within the next week, including some who underwent emergent surgery.8 High rates of LWBS also reflect lower patient satisfaction, potential violation of the Emergency Medical Treatment and Labor Act (EMTALA) by failure to perform a medical screening examination, and potential lost revenue.1
LWBS rates can be used as one indicator of ED quality.4,11 However, to do so, a national LWBS rate estimate should be established as a benchmark for comparisons. Published estimates of LWBS rates from single site institutions vary widely, from 0.84%12 to 15%.7,13 This likely represents underlying differences in the patient and hospital characteristics of the EDs reporting these rates. A recent national study suggests that the average LWBS rate was 1.4% during 1997–2004 and continues to increase over time.14 To make valid comparisons, LWBS rates should be compared among EDs that are similar in characteristics that significantly influence LWBS rates. The objective of this study was to estimate the national LWBS rate and to identify patient, visit, and institutional characteristics that predict LWBS rates.
- Top of page
From 1998 to 2006, there were an estimated 1.8 million (95% CI = 1.7 to 2.1 million) patients per year who LWBS in EDs across the United States, translating to a LWBS rate of 1.7 (95% CI = 1.6 to 1.9) patients per 100 ED visits (Table 1). On multivariate analysis (adjusted OR), patients who LWBS were less likely to be at the extreme of ages: age <18 years (OR = 0.80, 95% CI = 0.66 to 0.96) and ≥65 years (OR = 0.46, 95% CI = 0.32 to 0.64). Patients who LWBS were more likely to be nonwhite: black or African American (OR = 1.41, 95% CI = 1.22 to 1.63) and Hispanic (OR = 1.25, 95% CI = 1.04 to 1.49). Nursing home residents (OR = 0.29, 95% CI = 0.08 to 1.00) were less likely to LWBS. Compared to those with private insurance, those with Medicaid (OR = 1.47, 95% CI = 1.27 to 1.70), self-pay (OR = 1.96, 95% CI = 1.65 to 2.32), or other (OR = 2.09, 95% CI = 1.74 to 2.52) payment sources were more likely to LWBS.
Table 1. Patient Characteristics of ED Visits by Whether or Not the Patient Left Without Being Seen
|Patient Characteristics||Did Not LWBS, % (n = 283,907)||LWBS, % (n = 5,172)||Unadjusted OR of LWBS (95% CI)||Adjusted OR of LWBS (95% CI)||LWBS Rate (per 1,000 Visits)|
|Overall|| || || || ||17.3|
|Age, mean ± SE (yr)||36 ± 0.2||29 ± 0.4*|| || || |
| <18||25||25*||0.76 (0.68–0.84)*||0.80 (0.66–0.96)*||17.7|
| 18–39||34||46*||Reference 1.0||Reference 1.0||23.2|
| 40–64||26||23*||0.65 (0.60–0.72)*||0.75 (0.65–0.86)*||15.3|
| ≥65||15||6*||0.27 (0.23–0.32)*||0.46 (0.32–0.64)*||6.4|
| Female||54||54||Reference 1.0||Reference 1.0||17.4|
| Male||46||46||0.98 (0.91–1.06)||0.96 (0.87–1.07)||17.1|
| Black or African American||22||32*||1.67 (1.49–1.88)*||1.41 (1.22–1.63)*||24.9|
| Asian||2||2*||0.90 (0.70–1.14)||0.92 (0.61–1.39)||13.5|
| Others||1||1*||1.75 (1.10–2.78)*||1.66 (0.77–3.59)||26.0|
|Hispanic||13||16*||1.36 (1.18–1.56)*||1.25 (1.04–1.49)*||22.5|
| Nursing home resident||4||1*||0.19 (0.10–0.36)*||0.29 (0.08–1.00)*||3.3|
| Non–nursing home resident||96||99||Reference 1.0||Reference 1.0||17.1|
| Walk-In||83||92*||Reference 1.0||Reference 1.0||18.5|
| Ambulance||16||7*||0.40 (0.33–0.50)*||0.65 (0.50–0.84)*||7.5|
| Public service||2||1*||0.65 (0.46–0.92)*||0.47 (0.31–0.72)*||12.1|
| Private||37||26*||Reference 1.0||Reference 1.0||12.2|
| Medicare||15||7*||0.63 (0.53–0.74)*||1.07 (0.82–1.41)||7.7|
| Medicaid||20||23*||1.58 (1.41–1.76)*||1.47 (1.27–1.70)*||19.1|
| Self-pay||15||24*||2.23 (1.95–2.57)*||1.96 (1.65–2.32)*||26.8|
| Other||12||20*||2.51 (2.17–2.91)*||2.09 (1.74–2.52)*||30.0|
On bivariate analysis, several visit characteristics were associated with LWBS (Table 2). On multivariate analysis, visits for miscellaneous complaints were more likely to LWBS compared to general complaints (OR = 1.56, 95% CI = 1.19 to 2.05). Visits with the lowest triage acuity had higher risk of LWBS (OR = 3.59, 95% CI = 2.81 to 4.58) compared to visits with the highest triage acuity, while those with moderate pain scores had lower risk of LWBS (OR = 0.58, 95% CI = 0.47 to 0.72) compared with those without pain. Visits that were work-related also had lower risk of LWBS (OR = 0.19, 95% CI = 0.12 to 0.29). Patients who presented for a follow-up visit were less likely to LWBS (OR = 0.58, 95% CI = 0.36 to 0.92). Visits where the patient received a diagnostic (OR = 0.17, 95% CI = 0.15 to 0.20), procedural (OR = 0.05, 95% CI = 0.04 to 0.07), or medication intervention (OR = 0.03, 95% CI = 0.02 to 0.03) were all associated with lower risk of LWBS. The mean length of visit for patients who LWBS was 160 (SE ± 10) minutes.
Table 2. ED Visit Characteristics by Whether or Not the Patient Left Without Being Seen
|Visit Characteristics||Did Not LWBS, % (n = 283,907)||LWBS, % (n = 5.172)||Unadjusted OR of LWBS (95% CI)||Adjusted OR of LWBS (95% CI)||LWBS Rate (per 1,000 Visits)|
| General||16||14*||Reference 1.0||Reference 1.0||15.4|
| Psychiatric||2||3*||1.33 (1.01–1.75)*||1.26 (0.85–1.87)||20.4|
| Neurologic||7||7*||1.20 (1.02–1.42)*||1.18 (0.93–1.50)||18.5|
| Cardiovascular||1||1*||0.94 (0.69–1.30)||1.19 (0.75–1.88)||14.6|
| Eyes/ears/nose/throat||3||4*||1.14 (0.91–1.44)||0.83 (0.61–1.12)||17.6|
| Pulmonary||12||10*||0.95 (0.81–1.12)||0.84 (0.68–1.04)||14.7|
| Gastrointestinal||13||15*||1.28 (1.10–1.49)*||1.11 (0.92–1.35)||19.7|
| Genitourinary||4||5*||1.46 (1.21–1.77)*||1.01 (0.79–1.31)||22.4|
| Dermatologic||3||4*||1.44 (1.13–1.84)*||0.95 (0.68–1.33)||22.0|
| Musculoskeletal||14||11*||0.94 (0.81–1.09)||0.70 (0.57–0.85)*||14.5|
| Injury/poisoning/adverse||20||14*||0.82 (0.71–0.94)*||0.65 (0.53–0.80)*||12.6|
| Miscellaneous||5||11*||2.64 (2.22–3.15)*||1.56 (1.19–2.05)*||39.8|
| Should be seen in <15 minutes||21||8*||Reference 1.0||Reference 1.0||5.7|
| Should be seen in 15–60 minutes||42||32*||2.08 (1.75–2.47)*||1.64 (1.32–2.03)*||11.7|
| Should be seen in 1–2 hours||23||33*||3.85 (3.23–4.59)*||2.93 (2.38–3.61)*||21.5|
| Can be seen in >2 hours||14||27*||5.22 (4.16–6.55)*||3.59 (2.81–4.58)*||28.9|
|Presenting pain level|
| None||24||31*||Reference 1.0||Reference 1.0||15.9|
| Mild||26||26*||0.78 (0.65–0.93)*||0.71 (0.58–0.86)*||12.5|
| Moderate||30||24*||0.64 (0.52–0.78)*||0.58 (0.47–0.72)*||10.2|
| Severe||19||19*||0.80 (0.63–1.02)||0.69 (0.53–0.90)*||12.8|
|Visit related to|
| Alcohol||3||3||1.18 (0.85–1.62)||1.22 (0.60–2.50)||15.0|
| Work injury||5||2*||0.34 (0.24–0.48)*||0.19 (0.12–0.29)*||5.1|
|Previous visit within 72 hours||3||3||0.99 (0.78–1.26)||0.86 (0.59–1.25)||15.5|
|Follow-up visit||6||5||0.78 (0.60–1.01)||0.58 (0.36–0.92)*||11.4|
|Received diagnostic studies||87||46*||0.13 (0.11–0.14)*||0.17 (0.15–0.20)*||9.2|
|Received procedure||47||3*||0.04 (0.03–0.05)*||0.05 (0.04–0.07)*||1.1|
|Received medication||77||8*||0.03 (0.02–0.03)*||0.03 (0.02–0.03)*||1.8|
|Length of visit, mean ± SE (minutes)||192 ± 3||160 ± 10*|| || || |
| <1 hours||16||30*||2.44 (2.08–2.88)*||1.86 (1.42–2.44)*||22.6|
| 1–2 hours||27||24*||1.12 (0.95–1.32)||0.87 (0.69–1.12)||10.5|
| 2–4 hours||34||27*||Reference 1.0||Reference 1.0||9.4|
| >4 hours||23||18*||1.01 (0.80–1.27)||1.05 (0.78–1.42)||9.4|
Several institutional characteristics were associated with LWBS on bivariate analysis (Table 3). Patients seen in teaching institutions (OR = 1.49, 95% CI = 1.20 to 1.86) and metropolitan areas (OR = 2.35, 95% CI = 1.83 to 3.05) had higher risk of LWBS than those seen in nonteaching institutions or nonmetropolitan areas. After adjusting for patient (age, race, insurance status) and visit characteristics (mode of arrival, triage acuity; multivariate model), the risk of LWBS remained high for both teaching institutions (OR = 1.33, 95% CI = 1.06 to 1.67) and metropolitan areas (OR = 2.11, 95% CI = 1.66 to 2.70).
Table 3. Institutional Characteristics of ED Visits by Whether or Not They Left Without Being Seen
|Institutional Characteristics||Did Not LWBS, % (n = 283,907)||LWBS, % (n = 5,172)|| Unadjusted OR of LWBS (95% CI)|| Adjusted OR of LWBS (95% CI)||LWBS Rate (per 1,000 Visits)|
|Nonteaching institution||93||89*||Reference 1.0||Reference 1.0||16.7|
|Teaching institution||7||11*||1.49 (1.20–1.86)*||1.33 (1.06–1.67)*||24.7|
|Nonmetropolitan area||19||9*||Reference 1.0||Reference 1.0||8.3|
|Metropolitan area||81||91*||2.35 (1.82–3.05)*||2.11 (1.66–2.70)*||19.4|
| Voluntary, nonprofit||73||71||Reference 1.0||Reference 1.0||16.8|
| Government, nonfederal||18||19||1.11 (0.92–1.35)||0.88 (0.72–1.08)||18.7|
| Proprietary||10||10||1.06 (0.83–1.36)||0.87 (0.64–1.18)||17.9|
- Top of page
Patients who present to the ED and then LWBS represent a significant problem for many EDs around the country. These patients are at risk for progression of their disease, developing complications, and even death. Prior literature from single institutions suggests that the rate of LWBS is anywhere from 0.84% to 15%.7,12,13 We have shown the national rate of LWBS to be 1.7% using the NHAMCS. This may be used as a benchmark for EDs around the country to evaluate their LWBS rate.
However, we found that many patient, visit, and institutional characteristics affected this national LWBS rate. Young adults (18–39 years old) were the most likely to LWBS. Minorities (black or African Americans and Hispanics) and those who lack private insurance were also associated with a significantly higher risk of LWBS. This is consistent with other single-institution studies suggesting that the disadvantaged are at increased risk of LWBS.1,14,21,23–25 These data are concerning because the patients who tend to LWBS are more likely to lack access to alternative primary care and be more at risk for adverse events.3,24–26 This population may be using the ED as a source of primary care, given the finding that most LWBS visits are of low acuity. On the other hand, a significant percentage of LWBS patients (26%) do have private insurance. This may represent a valuable source of “recoverable” revenue for EDs hoping to improve operational efficiency.
We also found certain visit characteristics to be associated with LWBS. Higher-acuity visits (high triage priority, high pain level, ambulance arrival) were less likely to LWBS, which is consistent with other studies.21,24,27 This suggests that patients who LWBS are of lower illness acuity and possibly at less risk of complications. Therefore, LWBS may be less of a patient safety concern than previously suggested. However, 8% of patients who LWBS are initially triaged in the highest acuity (should be seen in less than 15 minutes) and thus may be still be at significant risk of harm. Alcohol-related visits were not associated with LWBS. Patients seen for work-related visits were much less likely to LWBS. This may represent a work requirement to complete the visit or perhaps a desire of EDs to evaluate these paying patients sooner.
We found that institutional factors influenced the risk of LWBS. Teaching institutions were associated with higher risk of LWBS after controlling for other patient, clinical, and hospital characteristics. This may be due to longer treatment and wait times associated with the education of trainees. We also found a difference between urban and rural areas. Even after adjusting for patient and visit characteristics, those seen in urban hospitals are more likely to LWBS than those seen in rural or suburban hospitals. This may represent the unavailability of alternative health care facilities for those patients seen in nonurban areas. The next closest ED for these patients may be too far away for them to consider LWBS. Another explanation may be increased crowding conditions found among EDs in metropolitan areas, which has been documented in the literature.1,7–9,23,24,28–31 Patients in rural/suburban EDs may be seen more quickly, have decreased waits to be seen, and have lower length of stays. All of these findings have implications when comparing EDs across the country for quality. Adjustments to LWBS rates should be made for the demographic makeup of the patient population and for metropolitan and teaching facilities.
Furthermore, by knowing what factors influence the decision to leave before completing the medical evaluation, EDs can implement changes to improve the LWBS rate. For example, we found that patients who had any type of diagnostic or therapeutic intervention were significantly less likely to LWBS. This has profound clinical importance, as more and more EDs are moving diagnostic testing and interventions earlier in the visit (at triage). EDs may be able to decrease their LWBS rates by performing diagnostic or therapeutic interventions on patients earlier. By altering factors that are known to affect incomplete patient evaluations, LWBS rates may be improved.
- Top of page
The first limitation is the potential biases of the NHAMCS database. There is the potential that the data may be inaccurately collected or that the sampling is not nationally representative. However, NHAMCS has been in existence for over 15 years, and the data have been validated in over 100 peer-reviewed articles. The second limitation is that we treated all of the EDs in the sample throughout the 9-year study period as independent, but many of the EDs in the study were repeatedly sampled, and we did not account for the autocorrelation within EDs. This may have resulted in an overestimate of the LWBS rate by different patient and hospital characteristics. The third potential limitation is in the definition of LWBS. Even though the NHAMCS definition of LWBS was “the patient left the hospital after being triaged, but before receiving any medical care,” almost half of LWBS patients received a diagnostic evaluation, while others received treatment. In some institutions, there are variations on patients who leave before completing their treatment: triaged and left, screened by physician and left, left without completion of service, etc. In some of these cases, patients may receive diagnostic testing (x-rays, blood draws, etc.) and procedures (nebulizers, intravenous fluids) in triage. This likely represents the trend toward pushing evaluation and treatment earlier in the patient stay.32 Therefore, these patients may have received some treatment before being seen by a physician. The NHAMCS survey does not distinguish these subcategories, so we were unable to distinguish these subcategories of LWBS. The fourth potential limitation is that the quality of data availability may be different for those patients who LWBS than for those who complete their care. Patients who complete their care may have higher-quality data because their charts are more complete. However, the data elements that were used in this study were either demographic or available at triage, both of which should be available early in the visit. Fifth, we utilized the first “reason for visit” in the model. The NHAMCS survey includes up to three reasons for visit. We assume that the first is the most important, although we cannot be certain of this in all cases. Finally, we may not have included all important variables that affect LWBS because they were not measured by NHAMCS, such as the extent of ED crowding. Specifically, variables of interest include waiting room time, “boarding” time, ambulance diversion hours, etc. ED crowding causes patients to LWBS so it may be that we have overestimated the risks associated with LWBS for factors that may be confounded by crowding, such as metropolitan area.