ACADEMIC EMERGENCY MEDICINE 2010; 17:834–839 © 2010 by the Society for Academic Emergency Medicine
Objectives: This study sought to determine if emergency department (ED) crowding was associated with longer ED length of stay (LOS) and time to ordering medications (nebulizers and steroids) in patients treated and discharged with acute asthma and to study how delays in ordering may affect the relationship between ED crowding and ED LOS.
Methods: A retrospective cohort study was performed in adult ED patients aged 18 years and older with a primary International Classification of Diseases, 9th Revision (ICD-9), diagnosis of asthma who were treated and discharged from two EDs from January 1, 2007, to January 1, 2009. Four validated measures of ED crowding (ED occupancy, waiting patients, admitted patients, and patient-hours) were assigned at the time of triage. The associations between the level of ED crowding and overall LOS and time to treatment orders were tested by analyzing trends across crowding quartiles, testing differences between the highest and lowest quartiles using Hodges-Lehmann distances, and using relative risk (RR) regression for multivariable analysis.
Results: A total of 1,716 patients were discharged with asthma over the study period (932 at the academic site and 734 at the community site). LOS was longer at the academic site than the community site for asthma patients by 90 minutes (95% confidence interval [CI] = 79 to 101 minutes). All four measures of ED crowding were associated with longer LOS and time to treatment order at both sites (p < 0.001). At the highest level of ED occupancy, patients spent 75 minutes (95% CI = 58 to 93 minutes) longer in the ED compared to the lowest quartile of ED occupancy. In addition, comparing the highest and lowest quartiles of ED occupancy, time to nebulizer order was 6 minutes longer (95% CI = 1 to 13 minutes), and time to steroid order was 16 minutes longer (95% CI = 0 to 38 minutes). In the multivariable analysis, the association between ED crowding and LOS remained significant. Delays in nebulizer and steroid orders explained some, but not all, of the relationship between ED crowding and ED LOS.
Conclusions: Emergency department crowding is associated with longer ED LOS (by more than 1 hour) in patients who ultimately get discharged with asthma flares. Some but not all of longer LOS during crowded times is explained by delays in ordering asthma medications.
In 2005, there were 1.8 million emergency department (ED) visits for acute asthma, making up a little more than 1% of all ED visits.1 Of those, only a minority (13%) were admitted to the hospital, making emergency care for asthma primarily an ambulatory condition. Asthma care in the ED typically consists of evaluation, assessment of the need for bronchodilator and steroid therapy, and other adjunctive treatments.2 Management decisions regarding therapy are relatively straightforward and many patients can be managed with protocols.3 Because most cases of asthma are treated and ultimately discharged from the ED, expediting treatment is important to quality of care for this population and overall ED flow.
Several recent studies have documented the negative effect of ED crowding on quality of care for conditions such as pneumonia, chest pain syndrome, and other pain syndromes.4–10 However, for asthma care, the ultimate goal in most cases is an improvement in clinical signs and symptoms. This makes time to time to symptom control and ultimate discharge an important outcome. One published article studied ED throughput times for admitted patients with asthma and found that the implementation of a short-stay unit was associated with shorter boarding times; however, no studies have studied the effect of a patient-level exposure to ED crowding on overall length of stay (LOS) in patients with asthma.11
We examined whether higher levels of ED crowding were associated with longer ED LOS in patients who are ultimately treated and discharged with a primary diagnosis of acute asthma. Secondarily, we examined whether ED crowding was associated with longer times to ED medication orders, including nebulizer and steroid therapy, and explored how delays in ordering medical therapy may affect the relationship between ED crowding and ED LOS.
This was a retrospective study of adult ED patients, age 18 and older, with a primary International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) diagnosis of asthma (493.xx) from January 1, 2007, to January 1, 2009 in two EDs, who were treated and discharged from the ED. The study was approved by the institutional review board.
Study Setting and Population
The study was conducted at two urban, inner-city EDs. One ED is an academic, tertiary care center with approximately 57,000 annual adult visits, 25 treatment rooms, 15 hallway spaces, a separate eight-bed fast track, and a three-bed trauma bay. This ED hosts a 4-year emergency medicine residency training program. Patients are seen by attending physicians, resident physicians, and medical students. The second ED is a community ED that sees both adult and pediatric patients in the same health system as the academic site, with approximately 35,000 annual visits with 20 treatment rooms, six hallway spaces, and three fast-track spaces. Primarily attending physicians staff the community ED, although residents do rotate through the hospital. There was no change in size in either ED during the study period.
Both EDs have standing orders by which nebulizer treatments may be given at ED triage as deemed appropriate by the ED triage nurse, and there are no strict guidelines on which patients will get treatment at triage. The academic site has an asthma care pathway, whereby eligible patients enter a protocol that involves a respiratory therapist providing a series of nebulizer treatments with frequent reevaluation, and there is an automatic order for oral steroid therapy. Deployment of the protocol is initiated by the treating physician or nurse. At both sites nebulizer treatments may be given by nurses, respiratory therapists, and occasionally physicians.
Patients were excluded if they were admitted to the hospital, left against medical advice, left without completing treatment, left without being seen, were transferred, or died in the ED. These populations were excluded because we wanted to identify a homogeneous population to isolate the effect of ED crowding on care times for a patient population where the ultimate goal is expeditious care. ED LOS for other populations may be more likely a function of factors extrinsic to the ED (such as inpatient bed availability) than of ED-based resources.
The computerized medical record and order entry system EMTrac (University of Pennsylvania, Philadelphia, PA) was used to identify patients in both EDs by extracting data by the ICD-9-CM code. Only cases where asthma was the primary diagnosis were included. Both hospitals use EMTrac as their ED information system. The tracking, order entry, and charting systems are run identically at each site.
Triage time, room time, medication order time, and discharge time are automatically time-stamped, while medication done time is entered by nurses after medications are given. For study purposes, we used the medication order time because medication done times are less reliable and can be entered long after a medication is given. To be counted as having been given a nebulizer treatment, we included the administration of either albuterol sulfate or ipratropium bromide by nebulization. Other nebulizer treatments are not used at the study sites. For steroid therapy, we included prednisone, methylprednisolone, or dexamethasone. Discharge medications, such as an albuterol metered dose inhaler, were not counted as ED medications for study purposes.
EMTrac was used to assign the following ED crowding measures at triage to each patient encounter: ED occupancy rate (proportion of ED beds filled including hallway spaces), total patient-care hours (defined as the arithmetic sum of the hours of all patients in the ED, excluding trauma and fast-track patients), number of patients in the waiting room, and inpatient number boarding in the ED (for whom a bed request had been entered). The calculation of the crowding variables was performed through Microsoft Access (Microsoft Corp., Redmond, WA) using queries that reconstructed the state of the ED at the time of triage for each patient. These were chosen because they have been associated with other studies testing the association between crowding and quality of care in these two EDs.4–6,10,12,13
We also extracted the following variables to use as covariates in the adjusted model: patient age, race, sex, triage level (a four-level system is used, in which triage Level 1 is the most severe and triage Level 4 is the least severe), and time of day of ED arrival (7:00AM–2:59PM, 3:00PM–10:59PM, 11:00PM–6:59AM), stratified according to clinical shift. The main study outcomes were overall ED LOS, time to the order of the first nebulizer treatment, and time to the order of steroid treatment.
The data are presented as means ± standard deviation (SD) or frequencies with percentages. All times are presented as medians with interquartile ranges (IQR). The crowding measures were divided into quartiles within each individual hospital. Quartiles were then combined for a pooled analysis. We used Fisher’s exact test and the Mann-Whitney U-test to assess differences in overall treatment times and time to treatment between the two hospitals. To generate differences in overall ED LOS with a 95% confidence interval (CI), we used Hodges-Lehmann distances. We tested if there were significant trends across crowding quartiles in overall ED LOS, time to the order of the first nebulizer treatment, and steroid treatment using the Cuzick test for trend across ordered quartiles. We then used Hodges-Lehmann distances to determine differences in time intervals for the highest and lowest occupancy quartiles.
Next, we used multivariable analysis to assess for the adjusted effects of ED crowding on ED LOS. For this analysis, we signified a delay as being greater than the 50th percentile of LOS. In the initial model, we tested the effect of ED crowding as continuous variables on prolonged LOS and adjusted for the following covariates: age, sex, triage class, time of day, and hospital. ED crowding variables were continuous because of the relatively linear relationship between ED crowding and ED LOS observed in the quartile analysis. Subsequent models were then performed in the cohort who received any nebulizer treatment, where we first tested the effect of ED crowding on adjusted LOS and then added time to nebulizer treatment to the model to assess for a change in relative risk (RR). The latter analysis was performed to test how time to nebulizer treatment modified the relationship between ED crowding and ED LOS; specifically, whether delays in nebulizer treatment accounted for the association between higher levels of ED crowding and ED LOS. A similar stepped approach was used in the cohort of patients who received both nebulizer treatments and steroids. For all multivariable analyses, we used a generalized linear model with a log link, Gaussian error, and robust estimates of the standard errors of the model coefficients to calculate a RR. All analyses were powered such that there were more than 10 outcomes for each variable in the adjusted models. Since there were little missing data, no imputation was performed. We tested model fit using the Akaike’s information criterion. All analyses were performed using Stata statistical software (Version 10, StataCorp, College Station, TX). To adjust for multiple statistical tests performed on the same data, the Bonferroni correction was used. An n of 4 was selected for the correction because there were four separate crowding measures tested. Therefore, a probability of <0.0125 was considered statistically significant.
A total of 1,716 patients were included who were treated and discharged with a primary diagnosis of asthma during the study period at the two hospitals; of these, 932 patients were seen at the academic site and 784 at the community site. The patients were primarily female (54%), African American or black (63%), and young with a mean (±SD) age of 37 ± 14 years. The most common ED triage level was 3 of 4 (69%), and patients were seen most commonly during the hours of 3PM to 11PM (44%). Combining data from the two hospitals, the median LOS was 189 minutes (IQR = 115–289 minutes), and the median waiting room time was 13 minutes (IQR = 1–64 minutes). The majority of patients (88%) received nebulizer treatments while in the ED. Time to first nebulizer order on average was 17 minutes (IQR = 1–67 minutes). A total of 29% received steroid treatment while in the ED. The median time to steroid order was 25 minutes (IQR = 1–79 minutes; Table 1).
|Age (yr), mean (SD)||37 (±14)|
|Female sex, n (%)||909 (53)|
|African American or black race, n (%)||1,081 (63)|
|Triage level, n (%)|
|1 (most severe)||14 (1)|
|4 (least severe)||105 (6)|
|Triage time of day|
|Academic hospital, n (%)||932 (54)|
|Overall ED LOS, minutes (IQR)||189 (115–289)|
|Waiting room time, minutes (IQR)||13 (1–64)|
|Received ED nebulizer treatments, n = 1,510 (88%)|
|Time to nebulizer treatment, minutes (IQR)||17 (1–67)|
|Received ED steroid treatment, n (%)||498 (29)|
|Time to steroid treatment, minutes (IQR)||25 (10–79)|
Crowding levels were higher at the academic hospital than the community site by all measures. In addition, waiting time, time to first nebulizer order, time to steroid order, and overall ED LOS were longer at the academic site than the community site (all p < 0.001; Table 2). ED LOS was longer at the academic site than the community site by 90 minutes (95% CI = 79 to 101 minutes).
|Variable||Academic Hospital (n = 932)||Community Teaching Hospital (n = 784)|
|Admitted patient number, median (IQR)||8 (6–12)||4 (2–7)|
|Occupancy percent, median (IQR)||75 (58–83)||53 (39–73)|
|Patient-hours, median (IQR)||116 (8–152)||42 (24–63)|
|Waiting room number, median (IQR)||17 patients (10–23)||3 patients (1–7)|
|Overall ED LOS in minutes, median (IQR)||236 (160–332)||140 (86–221)|
|Waiting room time, median (IQR)||40 (13–110)||1 (0–10)|
|Received ED nebulizer treatments, n (%)||855 (92)||655 (84)|
|Time to nebulizer treatment, median (IQR)||24 (1–91)||12 (2–48)|
|Received ED steroid treatment, n (%)||239 (26)||259 (33)|
|Time to steroid treatment, median (IQR)||43 (1–124)||19 (1–53)|
|Left-without-being seen rate (%)||5||4|
In the trend analysis, which combined data from both hospitals, higher levels of ED crowding by all measures were associated with longer ED LOS and longer time to treatment orders (p < 0.001 for all trends). For example, at the highest levels of ED occupancy, the total ED LOS was 229 minutes (IQR = 144–347 minutes), compared to 157 minutes (IQR = 95–238 minutes) at the lowest level, with a Hodges-Lehmann distance of 75 minutes (95% CI = 58 to 93 minutes). Similarly, time to first nebulizer order was 25 minutes (IQR = 1–126 minutes) at the highest quartile of ED occupancy and 17 minutes (IQR = 2–40 minutes) at the lowest quartile, with a Hodges-Lehmann distance between the two distributions of 6 minutes (95% CI = 1 to 13 minutes). Time to first steroid was 41 minutes at the highest quartile of ED occupancy (IQR = 0–155 minutes), compared to 21 minutes (IQR = 2–40 minutes) at the lowest quartile, with a Hodges-Lehmann distance of 16 minutes (95% CI = 0 to 38 minutes; Table 3).
|Overall ED LOS in Minutes (IQR)|
|Admitted patients*||157 (95–238)||174 (109–275)||214 (128–302)||229 (144–347)|
|ED occupancy*||144 (88–230)||191 (118–281)||211 (132–309)||228 (144–342)|
|Patient-hours*||149 (91–239)||176 (110–274)||211 (125–309)||230 (144–332)|
|Waiting room no.*||151 (88–234)||185 (118–270)||207 (129–297)||243 (140–376)|
|Time to first nebulizer|
|Admitted patients*||17 (2–17)||22 (2–49)||21 (1–97)||26 (2–143)|
|ED occupancy*||17 (2–40)||17 (1–59)||15 (1–103)||25 (1–126)|
|Patient-hours*||17 (2–44)||15 (1–59)||19 (1–87)||21 (1–116)|
|Waiting room no.*||15 (1–42)||18 (2–60)||21 (1–80)||17 (1–155)|
|Time to steroid|
|Admitted patients*||23 (2–47)||24 (0–54)||21 (1,63)||58 (19–142)|
|ED occupancy*||21 (2–40)||25 (1–25)||43 (2–118)||41 (0–155)|
|Patient-hours*||19 (0–45)||18 (1–57)||30 (1–114)||54 (7–155)|
|Waiting room no.*||21 (2–44)||27 (1–94)||22 (0–80)||47 (3–175)|
After adjusting for potential confounders, all ED crowding variables were broadly predictive of longer ED LOS than the median for all cohorts of patients (entire cohort, those who received nebulizers only, and those who received nebulizers and steroids). The number of admitted patients had the largest effect with an RR of 1.06 (95% CI = 1.05 to 1.07) per patient in the waiting room. Time to nebulizer order and time to steroid order explained some of the relationship between LOS and three crowding measures: the number of admitted patients, ED occupancy, and waiting room number. For example, among those treated with nebulizers, the RR for the number of admitted patients decreased from 1.06 (95% CI = 1.05 to 1.07) per patient to 1.04 (95% CI = 1.03 to 1.05) per patient when time to nebulizer order was included in the model. Treatment timing did not affect the relationship between patient-hours and LOS, with the RR staying constant at 1.02 (95% CI = 1.01 to 1.02). None of the adjusted models were rejected because of lack of fit (Table 4).
|Overall ED LOS||Risk of Delay||+Nebs||+Nebs and Steroids|
|Overall cohort||1.06 (1.05–1.07)†||—||—|
|Those receiving nebs only||1.06 (1.05–1.07)†||1.04 (1.03–1.05)†||—|
|Those receiving nebs and steroids||1.07 (1.03–1.10)†||1.06 (1.03–1.09)†||1.04 (1.01–1.07)†|
|Overall cohort||1.02 (1.01–1.03)†||—||—|
|Those receiving nebs only||1.02 (1.01–1.03)†||1.01 (1.00–1.03)||—|
|Those receiving nebs and steroids||1.02 (0.99–1.05)||1.00 (0.97–1.03)||1.00 (0.97–1.03)|
|Overall cohort||1.02 (1.02–1.02)†||—||—|
|Those receiving nebs only||1.02 (1.02–1.02)†||1.02 (1.01–1.02)†||—|
|Those receiving nebs and steroids||1.02 (1.01–1.03)†||1.02 (1.01–1.02)†||1.02 (1.01–1.02)†|
|Waiting room no.*|
|Overall cohort||1.03 (1.02–1.03)†||—||—|
|Those receiving nebs only||1.03 (1.02–1.03)†||1.02 (1.01–1.02)†||—|
|Those receiving nebs and steroids||1.03 (1.02–1.04)†||1.02 (1.01–1.03)†||1.01 (1.01–1.02)†|
We found that higher levels of crowding are independently associated with increased LOS and delayed time to order of treatment for asthma patients who are treated and ultimately discharged from the ED. This is a demonstration of Little’s Law (L = λ × W) in action in two busy EDs. Little’s Law states that in a stable system, the mean number of patients within the system (L, the level of ED crowding) is equal to the average arrival rate (λ) multiplied by the long-term mean time a customer spends in the system (W, ED LOS). The underlying assumption is that arrival rate (λ) stays the same regardless of the number of patients within the system, which is a reasonable assumption. In real terms, what this means is that the more crowded the ED, the longer patients stay within the ED—crowding ultimately breeds more crowding because newly arriving patients continue to arrive and enter a gridlocked system and then stay longer within that system. The result is that patients with asthma spend more time in the ED (more than 1 hour) when presenting at crowded versus less crowded times.
Several other recent studies have demonstrated a similar phenomenon, that when the ED is crowded, patients tend to spend a longer overall amount of time in the ED.14–16 Across four EDs, ED crowding was found to have a substantial effect on waiting room time and boarding time, but not treatment time, indicating that certain elements of ED LOS may be more affected than others by having more patients within the system.14 A similar phenomenon is seen as crowding and has been shown to delay time to critical laboratory results such as troponin I and preliminary computed tomography readings in patients with abdominal pain.13,17 For patients, this means the longer the line for a test, the longer the wait for the result, assuming insufficient demand-capacity matching.18 However, another study found that low-complexity patients did not appreciably extend LOS for other patients waiting.19 This may indicate that patient-level resource consumption may guide how much the system slows down when it reaches higher crowding levels. Therefore, models that attempt to use Little’s Law to study ED care (i.e., the ED crowding–LOS connection) should account for resource utilization by individual patients.
The multivariable analysis yielded an interesting observation that delays in nebulizer treatment and steroid treatment were not sufficient explanations delays in overall LOS. The purpose of this analysis was to decompose the effect of value-added activities for patients that could directly improve their condition (nebulizers and steroids) from the non–value-added activities such as waiting to be seen, waiting for evaluation, and finally waiting for discharge paperwork. The delays associated with ED crowding appeared to affect both delays in value-added and non–value-added processes. A reason for this is that the value-added process (i.e., the care in the asthma pathway) may be less sensitive to the administrative state of the ED than other processes that must occur prior to patients being discharged home. This suggests that one solution may be to not just initiate protocols to the front-end care (i.e., treatment for asthma), but also to implement them later in the care process to reduce the treatment delays associated with higher ED volumes.
This is only a two-hospital study and may not be generalized to other hospitals. In addition, medication order times may not reflect the time the medication was actually given. This bias would likely be toward the null, because a crowded ED can be expected to have a longer delay in time from order until medication administration. Another limitation is that we could not properly adjust for severity of illness. We used triage levels, which are a proxy for severity, but did not use any physiologic measurements (such as peak flow). This may bias our study toward finding a difference because it is possible that patients with more severe asthma might present during more crowded times, because patients with less severe disease may have chosen to leave without being seen. Another potential unmeasured confounder was that there are no strict, explicit criteria for nurses using the standing orders. However, while it is possible that certain nurses work at more or less crowded times, we included time of day in the adjusted model to control for this, and the adjusted model remained significant.
We found that ED crowding is associated with longer ED lengths of stay for patients who are treated and discharged from two hospital EDs with acute asthma. These associations were robust in the multivariable analysis, where we found that delays in time to nebulizer and steroids explain some but not all of the effect of ED crowding on longer length of stay in ambulatory patients with asthma.
The authors thank Christian Boedec for his help with extracting the crowding data for this study.