a b s t r a c t

Study objective: Few investigations have been performed that address why emergency department (ED) crowding is associated with an increase in hospital mortality for emergency patients. The purpose of this study was to evaluate whether ED crowding is associated with delayed resuscitation efforts (DREs) that resulted in hospital mortality.

Methods: This is a retrospective observational study performed at a single urban Tertiary ED. All adult patients who entered the resuscitation room and underwent resuscitative procedures from October 2008 to May 2010 were enrolled in the study. Demographic data were collected from a designed resuscitation room registry. The ED electronic log data were used for calculating the crowding status. A crowded day was defined as a daily number of patients greater than 93, which was a cut-off derived from a sensitivity analysis. The primary outcome was a DRE, which occurred when a patient was located in the hallway or waiting room, then entered the resuscitation room, and received resuscitative procedures after the patient had clinically deteriorated. A secondary outcome was hospital mortality. Matched samples were selected using propensity scores to consider the clinical parameters and emergency severity index when the patients received triage immediately after registration. A logistic regression analysis was modeled to estimate the odds ratios (ORs) with 95% confidence intervals (CIs) on the DRE.

Results: A total of 1296 patients underwent resuscitative procedures in the resuscitation room. Of these, 226 (17.4%) were classified as the DRE group. A final 396 cases (30.6%) were matched and analyzed between DRE and non-DRE using the propensity score. The incidence of DRE was significantly higher on crowded days (OR, 2.00; 95% CI, 1.28-3.15). Mortality during the ED stay or during the total hospital stay was significantly higher in the DRE group (OR, 3.39; 95% CI, 1.22-9.45 and OR, 3.96; 95% CI, 2.28-6.88, respectively) compared with the non-DRE group.

Conclusion: Delays in resuscitation efforts occurred more frequently on crowded days and were associated with higher in-hospital mortality.

(C) 2013

Introduction

Background

Emergency department (ED) crowding has been a growing crisis in emergency care and is an impediment to the quality improvement of ED performance [1,2]. In previous studies, the adverse effect of ED crowding on the safety and efficacy of medical care delivery has been well established [3-5]. For example, ED crowding has been associated with delays in the timely administration of antibiotic treatment to patients with pneumonia [6]. Emergency department crowding also

? Funding acknowledgment: No author has conflict of interest in this study.

* Corresponding author.

E-mail addresses: [email protected] (K.J. Hong), [email protected] (S.D. Shin), [email protected] (K.J. Song), [email protected] (W.C. Cha), [email protected] (J.S. Cho).

delayed the administration of analgesic agents to patients in pain and was associated with overall increased hospital mortality [3,4,7]. These studies linked ED crowding with basic and relatively nonurgent ED treatment for stable patients. In addition, these studies have explained the effect of ED crowding on increased overall mortality for ED patients.

Resuscitation efforts are a critical and time-sensitive medical treatment that is frequently offered in the ED. When a critically ill patient arrives at the ED, accurate triage and immediate delivery of resuscitative procedures are important. Resuscitative procedures usually consist of hemodynamic monitoring, advanced airway management, mechanical ventilation, defibrillation, cardiopulmonary resuscitation, and other treatment modalities. For comprehensive and high-quality resuscitation, many institutions have separate opera- tional spaces termed resuscitation rooms or shock-trauma rooms that are equipped with medical devices for resuscitation [8-10]. In these rooms, experienced medical personnel perform comprehensive and

0735-6757/$ - see front matter (C) 2013 http://dx.doi.org/10.1016/j.ajem.2012.09.029

multidisciplinary medical care for critically ill patients. Although the Initial resuscitation effort is important and time sensitive, there have been a few reports of the effect of ED crowding on delayed resuscitation efforts (DREs) for patients in Serious condition during the initial ED approach. If ED crowding affects the delivery of resuscitation in critical cases, DRE could lead to worse clinical outcomes, including mortality.

To assess the effect of overcrowding on the timing of delivery of resuscitation on clinical outcomes, it is important to match the condition severity of patients who underwent resuscitation. Patients with more severe conditions could undergo resuscitative care immediately yet still experience higher mortality than stable patients who did not require immediate resuscitation. For example, most out- of-hospital cardiac arrest victims used the resuscitation room immediately. This situation could distort the results, with a potential skewing of the data to show that DRE is actually beneficial. To overcome this bias, propensity score matching could be a useful method of matching patients’ condition severity to assess a robust effect of ED crowding on DRE.

The goal of this investigation

The goal of this study is to assess the effect of ED crowding on delayed resuscitation efforts and its association with mortality.

Materials and methods

Study design

This investigation was a retrospective observational study using a registry for patients who used the resuscitation room in a single urban Tertiary teaching hospital in Korea. This study was approved by the institutional review board of the study institution.

Study setting

This investigation was conducted in an ED of urban tertiary teaching hospital with 57000 annual ED visits. This ED operated 35 available beds including a resuscitation room and any other Observation Units. This ED operated 1 separate “resuscitation room” for the timely and comprehensive delivery of resuscitative procedures for patients with Critical illnesses, such as cardiac arrest, acute coronary syndrome, stroke, unstable arrhythmia, multiple trauma, and respiratory failure, among others. This room was equipped with equipment for hemodynamic monitoring, central venous pressure monitoring, Arterial line monitoring, echocardiography, and ultra- sounds as well as a mechanical cardiac compression device, a defibrillator, an advanced airway including cricothyroidotomy set, mechanical ventilation, a rapid blood infusion device, an automatic medication infusion pump, and Chest tubes. Comprehensive and timely resuscitation was performed in this room in a Multidisciplinary team approach that included experienced emergency-boarded phy- sicians, high-grade emergency residents, emergency specialty nurses, and, if necessary, specialists.

When critically ill patients initially arrived at the ED, trained Triage nurses assessed the severity using the Emergency Severity Index

-version 4. The triage nurse or attending physician decided whether the patient should entered the resuscitation room for resuscitation efforts. If a patient was decided not to be entered immediately into resuscitation room by the triage nurse, the patient was assigned to the hallway or any other observation unit including available ED beds. If the deterioration of a patient was identified in a hallway or an observation unit other than resuscitation room after initial triage, the attending physician also ordered the transport of the patient into resuscitation room for immediate resuscitation efforts. The attending physician in the resuscitation room was also responsible for the other

ED patients at the same time. After entering the resuscitation room, the resuscitation team performed resuscitative procedures appropri- ate to the clinical condition. After finishing initial resuscitation efforts and stabilizing the patient in the resuscitation room, the patients were transported to the intensive care unit or the operation room, if necessary. Common condition categories that required the resuscita- tion room were cardiac arrest, shock, trauma, respiratory failure, mental changes, acute coronary syndrome, unstable arrhythmia, stroke, and seizure, among others.

This institution operated a standardized resuscitation room registry. This registry enrolled all patients who used the resuscitation room. The registry was recorded by resuscitation room nurse and daily reviewed by Attending emergency physicians.

Study population

Patients were enrolled who entered the resuscitation room during their ED stay between October 2008 and May 2010. Pediatric or Adolescent patients younger than 18 years were excluded from the study because an isolated pediatric ED was operated independently in this hospital. Any patients who used the resuscitation room with a do- not-attempt-resuscitation order were also excluded. Finally, if the vital signs or emergency severity index were not recorded in the resuscitation room registry, we excluded the case from analysis.

Data collection and processing

We collected the clinical data of patients from the resuscitation room registry database. In this registry, we acquired data including age, sex, injury, ambulance use, triage by ESI, mental status by the AVPU (Alert, Verbal response, Pain response, Unresponsiveness) method, vital signs, ED visit date and time, mortality during the ED stay and in- hospital mortality, disease entity causing entrance into resuscitation room, contents of resuscitative procedures, and resuscitation room entrance and departure times. The registry also recorded whether the patients entered the room resuscitation room directly after ED arrival or stayed in a hallway or waiting room before resuscitation.

We defined DREs as when patients did not enter the resuscitation room directly after ED arrival but entered the room after initially staying any other place. If patients entered the resuscitation room immediately after ED arrival, we defined them as non-DRE. Because immediate and comprehensive delivery was performed under the supervision of an attending physician upon patient entry of the resuscitation room, the room entry time could be used as a surrogate of the delivery time of resuscitation efforts. If resuscitation room was already in use, although the patients were qualified for immediate triage, they could not enter the resuscitation room and were assigned to other available ED beds.

To assess overcrowding, we collected the number of daily patients (Daily census) during study period based on electronic medical record logs. We then classified the day as crowded or noncrowded according to the number of daily patients by a cut-off derived from the receiver operating characteristic curve (ROC) analysis. The number of daily patients at the point with the highest sensitivity and the highest specificity for discriminating DRE vs non-DRE group was determined as the cut-off value.

Propensity score matching

To match the condition severity and other clinical characteristics of the DRE and non-DRE groups, we performed propensity score matching using the Greedy 5 to 1 Digit Match algorithm. We used the software packages SAS 9.2 (SAS Inc, Cary, NC) to conduct propensity score matching. To compute the propensity score, we selected indicators that could affect decision making about whether to send a patient into the resuscitation room. The variables used in the

propensity score matching were age, sex, injury, ambulance use, ESI level, mental status by the AVPU method, systolic blood pressure (millimeters of mercury), heart rate (per minute), respiratory rate (per minute), weekday/weekend visit, and daytime/nighttime visit. After propensity score matching, we compared the matching process variables between the original population and the propensity score- matched population. To compare systolic blood pressure and respiratory status used in the propensity score matching, shock was defined when initial systolic blood pressured was less than 90 mm Hg. In addition, respiratory distress was defined when initially measured respiratory rate was less than 10 or more than 30 per minute.

Outcome measures

The primary outcome was the DRE. We compared the percentage of patients with DRE among overcrowded and nonovercrowded group. The secondary outcome was overall in-hospital mortality, including the ED as well as the ward in general.

Statistical analysis

We conducted descriptive analysis and compared the demograph- ic and clinical variables and the outcomes between crowded days and noncrowded days. We compared the variables used for the propensity score matching for the original population and the propensity score- matched population by ?² tests and t tests. To assess the effect of crowding on DRE, we compared the proportion of the DRE group between crowded and noncrowded days for the original population and for the propensity score-matched population. We also assessed overall in-hospital and ED-specific mortality between the DRE and non-DRE groups by ?² tests and odds ratio (OR) calculations. We performed statistical analysis using the software packages SAS 9.2 (SAS, Inc) and STATA 11 (StataCorp LP, College Station, TX).

Results

Study population and propensity score matching process

During the study period, 59938 adult patients visited the ED of the institution. Among them, 1433 patients (2.4%) used the resuscitation

room (Figure). We excluded 72 patients (5.0%) due to their use of the resuscitation room for a death certificate without medical care and 65 patients (4.5%) due to input errors in their registry entries. Finally, 1296 patients were included for the analysis. The propensity score matching was performed for the 1296 patients, and 396 cases (30.6%) were selected as the propensity score-matched population.

The original population and the propensity score-matched population analysis

In the original population, the non-DRE group had higher rates of more Serious injuries, ambulance use, a higher ESI grade, poorer mental status, shock, and respiratory distress relative to the DRE group (Table 1). Therefore, critically ill patients with higher initial condition severity usually used the resuscitation room immediately.

After propensity score matching, there was no significant difference among the demographic and clinical variables, such as ESI or vital signs (Table 2). Therefore, no difference of initial case severity between the DRE and non-DRE groups in the original population was observed after propensity score matching.

Measuring ED crowding status: ROC analysis

During the study period, daily ED census was distributed from 57 to 140, and the mean value was 98.6. We divided overcrowded and nonovercrowded days by the cut-off value of 93 derived from the ROC analysis. If the number of Daily ED visits was equal to or greater than 93, we defined the day as crowded. Otherwise, the day was defined as noncrowded. Descriptive data pertaining to the ED visit volumes and the clinical characteristics of emergency patients who visited on crowded and noncrowed days are shown in Table 3.

Association between ED crowding and DRE

The proportion of patients in the DRE group was significantly higher on overcrowded days than on nonovercrowded days in the propensity-matched population (54.7% vs 37.6%, P b .01). Days with ED crowding had an OR for DRE of 2.00 (95% confidence interval [CI], 1.28-3.15) (Table 4).

Figure. Study population: original data set and propensity matched data set.

Table 1

Comparison of the baseline characteristics between the non-DRE group and the DRE group

Clinical characteristics		Resuscitation effort					P
		Nondelayed group			Delayed group
		N	%		N	%
Number (n = 1296)		1070	82.6		226	17.4
Age, years	Mean +- SD	61.8 +- 16.4			61.1 +- 16.1		.55
Sex	Male	661	61.8		134	59.3	.49
	Female	409	38.2		92	40.7
Injury	Disease	945	88.3		213	94.3	b.01
	Injury	125	11.7		13	5.80
Ambulance use	Prehospital	555	51.9		70	31.0	b.01
	Interhospital	231	21.6		52	23.0
	Nonuse	284	26.5		104	46.0
ESI level	ESI 1	612	47.2		46	20.4	b.01
	ESI 2	441	34.1		135	59.7
	ESI 3	16	1.2		42	18.6
	ESI 4	0	0		3	1.30
	ESI 5	1	0.10		0	0
Mental status	Alert	552	51.6		168	74.3	b.01
	Verbal	163	15.2		37	16.4
	Painful	166	15.5		15	6.64
	Unresponsive	189	17.7		6	2.65
Systolic blood pressure	b90 mm Hg	306	28.6		38	16.8	b.01
	>=90 mm Hg	764	71.4		188	83.2
Respiratory rate, per min.	b10 or >=30	275	25.7		21	9.29	b.01
	10 <= RR b 30	795	74.3		205	90.7
Heart rate, per min.	Bradycardia/tachycardia	671	62.7		127	56.2	.06
	Normal	399	37.3		99	43.8
ED visit day	Weekday	810	75.7		184	81.5	.06
	Weekend	260	24.3		42	18.5
ED visit time	8 AM to 6 PM	555	51.9		117	51.8	.98
	6 PM to 8 AM of the next day	515	48.1		109	48.2

Association of DRE and mortality

The DRE group experienced higher in-hospital mortality in the original and propensity score-matched populations (DRE vs non-DRE:

30.1% vs 20.2%, P b .01, and 30.8% vs 10.1%, P b .01, respectively) (Table 5). However, the OR of 3.96 (95% CI, 2.28-6.88) was much higher in the matched population than in the original population. There was no significant difference of ED mortality in the original population.

Table 2

Comparison of the baseline characteristics between the non-DRE group and the DRE group in the propensity score-matched population

Clinical characteristics		Resuscitation effort					P
		Nondelayed group			Delayed group
		N	%		N	%
Number (n = 396)		198	50.0		198	50.0
Age, years	Mean +- SD	61.6 +- 16.7			61.1 +- 16.0		.75
Sex	Male	125	63.1		120	60.6	.60
	Female	73	36.9		78	39.4
Injury	Disease	186	93.9		186	93.9	1.00
	Injury	12	6.06		12	6.06
Ambulance use	Prehospital	69	34.9		68	34.3	.99
	Interhospital	45	22.7		45	22.7
	Nonuse	84	42.4		85	42.9
ESI level	ESI 1	46	23.2		46	23.2	.74
	ESI 2	135	68.2		135	68.2
	ESI 3	16	8.1		16	8.1
	ESI 4	0	0		1	0.5
	ESI 5	1	0.5		0	0
Mental status	Alert	139	70.2		142	71.7	.85
	Verbal	33	16.7		35	17.7
	Painful	20	10.1		15	7.58
	Unresponsive	6	3.03		6	3.03
Systolic blood pressure	b90 mm Hg	38	19.2		36	18.2	.80
	>=90 mm Hg	160	80.8		162	81.8
Respiratory rate, per min.	b10 or >=30	30	15.2		20	10.1	.13
	10 <= RR b 30	168	84.9		178	89.9
Heart rate, per min.	Bradycardia/tachycardia	110	55.6		114	57.6	.69
	Normal	88	44.4		84	42.4
ED visit day	Weekday	150	75.8		159	80.3	.27
	Weekend	48	24.2		39	19.7
ED visit time	8 AM to 6 PM	109	55.1		99	50.0	.31
	6 PM to 8 AM	89	44.9		99	50.0

Table 3

Demographic findings of emergency patients who visited the ED on crowded and noncrowded days during the study period

Clinical characteristics		Status of ED crowding					P
		Crowded day?			Noncrowded day
		n	%		n	%
No. of days No. of ED visit		403	66.3%		205	33.7
Total ED visit		42857	71.5%		17081	28.5
Daily ED visit	Mean +- SD	106.3 +- 10.2			83.3 +- 6.9		b.01
Age	Mean +- SD	53.3+-18.3			52.4 +- 18.6		.02
Geriatrics	>=65 y	13774	32.1		5281	30.9	b.01
	b 65 y	29083	67.9		11800	69.1
Sex	Male	21551	50.3		8646	50.6	.46
	Female	21306	49.7		8435	49.4
Ambulance use	Prehospital	5760	13.4		2455	14.4	b.01
	Interhospital	2814	6.6		1181	6.9
	Non-use	34283	80.0		13445	78.7
Injury	Injury	6301	14.7		2854	16.7	b.01
	Disease	36556	85.3		14227	83.3
Admission	Yes	12345	28.8		4613	27.2	b.01
	No	30512	71.2		12468	29.0
	Median number	31			22
Admission to ICU	Yes	2,025	4.7		840	4.9	.32
	No	40832	95.3		16241	95.1
	Median number	5			4
ED mortality	Yes	307	0.7		143	0.8	.13
	No	42550	99.3		16938	99.2
	Median number	1			1
Resuscitation effort	Yes	875	2.0			2.5	b.01
	No	41982	98.0		16660	97.5
	Median number	2			2

Abbreviation: ICU, intensive care unit.

* If the number of daily ED visits was equal to or more than 93, we defined the day as a “crowded day” by the ROC analysis.

However, after adjusting for condition severity, a higher ED mortality was observed in the DRE group (DRE vs non-DRE: 8.1% vs 2.5%, P b .01). Therefore, DRE affected in-hospital and ED mortality adversely among patients with similar condition severity.

Distribution of disease entities, resuscitative procedures, and time factors of the DRE and non-DRE groups

We compared disease entities, resuscitative procedures, and time factors between the DRE and non-DRE groups, and the results are described in Table 6. In the non-DRE group, the most frequent disease entities necessitating resuscitation were arrhythmia, chest pain, and stroke. However, the DRE group matched by the propensity score showed significantly higher proportions of cardiac arrest, respiratory failure, and shock.

The delivery of critical procedures, such as advanced airway management, cardioversion, cardiopulmonary resuscitation, central venous monitoring, vasopressor administration, and mechanical ventilation, were also significantly increased in the DRE group. The time from ED arrival to the delivery to the resuscitation room was

3 minutes (median, interquartile range [IQR], 0-10) in the non-DRE group and 90 minutes (median, IQR, 37-310) in the DRE group. The time spent in the resuscitation room was 35 minutes (median, IQR, 19-62) and 46 minutes (median, IQR, 28-69) for the non-DRE and DRE groups, respectively.

Discussion

In this evaluation, we identified an association between ED crowding and the delayed delivery of resuscitation efforts. Previously, many researchers have demonstrated that overcrowding worsened the quality and safety of emergency care. In those studies, ED crowding delayed the delivery of medical care to emergency patients. For example, Pines et al [11] reported that the proportion of patients with pneumonia who received antibiotics within 4 hours was decreased when overcrowding occurred. Emergency department crowding was also associated with the delayed administration of analgesics to patients in pain and with the delayed delivery of thrombolytic therapy to patients with coronary disease [12-15]. However, the results focused on specific diseases could not explain

Table 4

Association of ED crowding and delayed resuscitation efforts

Mortality Original population Propensity score-matched population

	Noncrowded		Crowded		P		Noncrowded		Crowded		P
	n %		n	%			n %		n	%
n Resuscitation effort	421 32.5		875	67.5			109 27.5		287	72.5
Delayed	63 15.0		163	18.6	.12		41 37.6		157	54.7	b.01
Nondelayed	358 85.0		712	72.1			68 62.4		130	45.3
OR, 95% CI	Referent		1.30	0.95-1.79			Referent		2.00	1.28-3.15

Table 5

Association of delayed resuscitation effort and patient mortality

Mortality	Original population						Propensity score-matched	population
	Nondelayed		Delayed		P		Nondelayed	Delayed		P
	n %		n	%			n %	n	%
n In-hospital mortality	1070 82.6		226	17.4			198 50.0	198	50.0
Survival	854 79.8		158	69.9	b.01		178 89.9	137	69.2	b.01
Death	216 20.2		68	30.1			20 10.1	61	30.8
OR, 95% CI ED mortality Survival	Referent 1016 95.0		1.70 208	1.23-2.35 92.0	.11		Referent 193 97.5	3.96 182	2.28-6.88 91.9	.02
Death	54 5.0		18	8.0			5 2.5	16	8.1
OR, 95% CI	Referent		1.63	0.94-2.83			Referent	3.39	1.22-9.45

how overall mortality increased due to ED crowding. We identified that ED crowding delayed resuscitation efforts for critical patients with similar condition acuity and severity. This result could explain the mechanism of ED crowding on increased overall mortality.

Resuscitation entails time-sensitive procedures, and emergency physicians focus on the identification of the most serious patients and the timely delivery of life-saving resuscitation during clinical practice. Therefore, it is reasonable to ask whether ED crowding influences the performance of critical care, such as resuscitative procedures. For example, there is controversy regarding the effect of overcrowding on delayed primary coronary intervention in ST-segment elevation myocardial infarction cases. Schull et al [15] reported that the percentage of Regional hospitals on bypass was associated with delays to thrombolysis for Acute myocardial infarction patients. However, Harris et al [16] demonstrated the seemingly contradictory result that the time to percutaneous coronary intervention is not positively correlated with ED crowding. In our study, we identified that ED crowding, as measured by daily ED census, is associated with delays in the delivery of resuscitation to critically ill patients.

In assessing the effect of delayed resuscitation on clinically oriented outcomes such as mortality, bias reduction is critical.

Delays in the resuscitation of critically ill patients cause the deterioration of the clinical outcome. However, if we compare the DRE and non-DRE groups without adjustment, we would introduce a serious bias. Usually, more critically ill patients with high associated mortality, such as out-of-hospital cardiac arrest, would undergo resuscitation without delay. Therefore, to overcome this limitation and to identify robust effects, sample matching according to the initial severity of the enrolled population should be performed before further analysis. Our topic of the delay of critical care cannot be investigated by randomized controlled trial due to Ethical concerns. In retrospective observation studies like ours, propensity score matching could be a good solution to reduce study bias. Many previous investigators have used the propensity score matching method to adjust confounding factors in nonrandomized clinical studies [17-20]. We matched the initial condition severity of patients at arrival between delayed and immediate resuscitation groups by propensity score using the greedy matching technique. Before the matching process, there was no significant effect of ED crowding on delayed resuscitation efforts. However, after propen- sity matching to adjust for the initial condition severity of the 2 groups, ED crowding showed a significant association with DRE. The

Table 6

Comparison of condition entities, resuscitative procedures, and time factors between nondelayed and delayed resuscitation effort groups

Clinical characteristics	Resuscitation effort					P
	Nondelayed group			Delayed group
	n	%		n	%
Number (n = 396) Condition entity causing resuscitation Cardiac arrest	198 4	50.0 2.02		198 20	50.0 10.1	b.01
Respiratory failure	39	19.7		55	27.8	.06
Shock	24	12.1		42	21.2	.02
Arrhythmia	35	17.7		21	10.6	.04
Chest pain	28	14.1		1	0.51	b.01
Stroke	20	10.1		6	3.03	b.01
Seizure	8	4.04		10	5.05	.63
Trauma	10	5.05		6	3.03	.31
Nontraumatic hemorrhage	10	5.05		15	7.58	.30
Mental change	20	10.1		22	11.1	.74
Resuscitative procedures Advanced airway	12	6.06		75	19.0	b.01
Cardioversion	2	1.01		18	9.14	b0.01
Cardiopulmonary resuscitation	6	3.05		29	14.8	b.01
Central venous pressure monitoring	22	11.1		47	23.9	b.01
Vasopressor	13	6.57		36	18.4	b.01
Ventilator Time from ED arrival to resuscitation (min)	0	0		6	3.05	.02
Median, IQR Time staying at resuscitation room (min) Median, IQR	3 35	0-10 19-62		90 46	37-310 28-69

association between DRE and ED mortality was also more powerful in the propensity score-matched population.

We selected several clinical variables for calculating the propen- sity score. We selected factors that affect decision making regarding initial resuscitation that were available from the resuscitation room registry. Ambulance use was selected as a useful factor because critically ill patients frequently arrive at the ED by ambulance rather than by walking. We also included the ESI because the initial decision to offer resuscitation is determined by a triage nurse based on the ESI. We also included the day of the week of the ED visit and the ED visit time because there could be a difference of daily ED census and available medical resources during weekends or nighttime. By propensity score-matching using these variables, a bias for entering the resuscitation room and the associated clinical outcomes could be reduced.

In our study, ED crowding was associated with an increased proportion of delayed resuscitation. A higher patient volume than originally intended for the available medical resources of the ED would deteriorate ED performance. Our institution operated 1 resuscitation room equipped with prepared medical resources. Emergency department operation during nonovercrowded days showed low delays to resuscitative procedures. However, when the ED visit number exceeded the cut-off value proposed by the sensitivity analysis, the proportion of DRE was increased. In this investigation, a daily ED visit number higher than 93 resulted in the deteriorated performance of emergency care as indicated by the delay of resuscitation efforts. The ED had 35 beds, and mean number of daily ED census was 98.6. The institution was a tertiary academic hospital with high proportions of patients to be admitted on ward. The 35 beds were usually occupied by patients with exacerbated chronic disease or complicated oncologic patients and not available for new patients. Therefore, the 93 could be a threshold for being a crowded ED.

This investigation has several limitations. First, this study was performed at a single institution; therefore, it has a limitation of generalization. This hospital operates according to the critical care pathway protocol. Based on this policy, critically ill patients should be transported to the resuscitation room for resuscitative procedures to be conducted by a dedicated resuscitation team. In different practical settings or environments, the results of this study may not be applicable. Because the ED operated only 1 resuscitation room, the results may not apply to EDs with more than 1 resuscitation room. Second, we measured the time of delivery of resuscitative procedures by the time entering the resuscitation room. Usually, if critical patients entered the resuscitation room, we offered resuscitation immediately. However, there is a possibility of delay between entering the room and the performance of medical procedures. Furthermore, we defined delayed resuscitation efforts as cases that did not enter the resuscitation room immediately after ED arrival. However, there is no consensus about the time criteria that constitute delayed use of resuscitation rooms. Third, we selected a daily ED visit bulk cut-off as the indicator of ED crowding. Many different methods to assess crowding have been demonstrated. Adopting a different Crowding index might influence the results of this investigation. In addition, ED crowding measured by daily ED visits may not guarantee that the ED would be always busy during the minutes or hours while patients were in the resuscitation room. Fourth, this investigation was conducted by retrospective design. Finally, the qualities of the

resuscitative procedures other than the timeliness of delivery were not measured. In clinical practice, in addition to timeliness, the appropriate choice of procedures and their successful completion is important for favorable clinical outcomes.

Conclusion

We investigated the effect of overcrowding on delays to the delivery of resuscitation and clinically oriented outcome. After propensity score matching, ED crowding was associated with DRE, and DRE resulted in higher overall in-hospital or ED-specific mortality.

References

1. American College of Emergency Physicians. Policy statements: crowding. Ann Emerg Med 2006;47:585.
2. Derlet RW, Richards JR. Emergency department overcrowding in Florida, New York, and Texas. South Med J 2002;95(8):846-9.
3. Sprivulis PC, Da Silva JA, Jacobs IG, et al. The association between Hospital overcrowding and mortality among patients admitted via Western Australian emergency departments. Med J Aust 2006;184(5):208-12.
4. Richardson DB. Increase in patient mortality at 10 days associated with emergency department overcrowding. Med J Aust 2006;184(5):213-6.
5. Miro O, Antonio MT, Jimenez S, et al. Decreased health care quality associated with emergency department overcrowding. Eur J Emerg Med 1999;6(2):105-7.
6. Sikka R, Mehta S, Kaucky C, et al. ED crowding is associated with an increased time to pneumonia treatment. Am J Emerg Med 2010;28(7):809-12.
7. Pines JM, Shofer FS, Isserman JA, et al. The effect of emergency department crowding on analgesia in patients with back pain in two hospitals. Acad Emerg Med 2010;17(3):276-83.
8. Bernhard M, Becker TK, Nowe T, et al. Introduction of a treatment algorithm can improve the early management of emergency patients in the resuscitation room. Resuscitation 2007;73(3):362-73.
9. Claudet I, Bounes V, Federici S, et al. Epidemiology of admissions in a pediatric resuscitation room. Pediatr Emerg Care 2009;25(5):312-6.
10. Cattermole GN, Mak SK, Liow CH, et al. Derivation of a prognostic score for identifying critically ill patients in an emergency department resuscitation room. Resuscitation 2009;80(9):1000-5.
11. Pines JM, Hollander JE, Localio AR, et al. The association between emergency department crowding and hospital performance on antibiotic timing for pneumonia and percutaneous intervention for myocardial infarction. Acad Emerg Med 2006;13(8):873-8.
12. Pines JM, Hollander JE. Emergency department crowding is associated with poor pain care for patients with severe pain. Ann Emerg Med 2008;51:1-5.
13. Hwang U, Richardson LD, Sonuyi TO, et al. The effect of emergency department crowding on the management of pain in older adults with Hip fracture. J Am Geriatr Soc 2006;54:270-5.
14. Schull MJ, Morrison LJ, Vermeulen M, et al. Emergency department overcrowding and Ambulance transport delays for patients with chest pain. CMAJ 2003;168: 277-83.
15. Schull MJ, Vermeulen M, Slaughter G, et al. Emergency department crowding and thrombolysis delays in acute myocardial infarction. Ann Emerg Med 2004;44: 577-85.
16. Harris B, Bai JC, Kulstad EB. Crowding does not adversely affect time to percutaneous coronary intervention for acute myocardial infarction in a community emergency department. Ann Emerg Med 2012 Jan;59(1):13-7.
17. Austin PC. Propensity-score matching in the cardiovascular surgery literature from 2004 to 2006: a systematic review and suggestions for improvement. J Thorac Cardiovasc Surg 2007;134:1128-35.
18. Austin PC. Primer on statistical interpretation or methods report card on propensity-score matching in the cardiology literature from 2004 to 2006: a systematic review. Circ Cardiovasc Qual Outcomes 2008;1(1):62-7.
19. Kim SI, Kim CJ, Kim YJ, et al. Antiviral prophylaxis versus preemptive therapy to prevent cytomegalovirus infection and related death in liver transplantation: a retrospective study with propensity score matching. Transplant Proc 2012;44(3): 787-90.
20. Loucks EB, Buka SL, Rogers ML, et al. Education and coronary heart disease risk associations may be affected by early-life common prior causes: a propensity matching analysis. Ann Epidemiol 2012;22(4):221-32.