a b s t r a c t

Background: Emergency Department (ED) overcrowding is a worldwide problem, and it might be caused by prolonged patient stay in the ED. This study tried to analyze if different practice models influence patient flow in the ED.

Materials and methods: A retrospective, 1-year cohort study was conducted across two EDs in the largest healthcare system in Taiwan. A total of 37,580 adult non-trauma patients were involved in the study. The clinical practice between two ED practice models was compared. In one model, urgent and non-urgent patients were treated by different emergency physicians (EPs) separately (separated model). In the other, EPs treated all pa- tients assigned randomly (merged model). The ED length of stay , diagnostic tool use (including laboratory examinations and computed tomography scans), and patient dispositions (including discharge, general ward ad- mission, intensive care unit admissions, and ED mortality) were selected as outcome indicators.

Result: Patients discharged from ED had 0.4 h shorter ED LOS in the separated model than in merged model. After adjusting for the potential confounding factors through regression model, there was no difference of patient dis- position of the two practice models. However, the separated model showed a slight decrease in laboratory exam- ination use (adjusted odds ratio, 0.9; 95% confidence interval, 0.83-0.96) compared with the merged model.

Conclusion: The separated model had better patient flow than the merged model did. It decreased the ED LOS in ED discharge patients and laboratory examination use.

(C) 2017

Introduction

Emergency Department (ED) overcrowding is not a new topic; it has been identified as a problem for over 20 years [1]. Many researchers had proven its negative effect on quality of patient care outcomes and even psychological outcomes, such as increased ED mortality and revisit rates, worsening perceived clinician-patient communication, prolonged hospital stays, and increased costs for admitted patients [2-5]. Numer- ous strategies had been implemented to address the issue, yet the prob- lem remains unsolved and threatens to become worse. Causes of ED overcrowding are complicated. Robert W. Derlet et al. listed some of the major causes including increased complexity and Acuity of patients presenting to the ED, overall increase in patient volume, lack of beds for

* Corresponding author at: No. 123, Dapi Rd., Niaosong Dist., Kaohsiung City 833, Taiwan.

E-mail address: [email protected] (C.-J. Li).

¹ Fei-Fei Yau and Tsung-Cheng Tsai contributed equally to this work.

patients admitted to the hospital, and shortage of ED staff [6]. Some of these issues require more medical resources despite budget limitations. We studied different models of ED practice that might influence patient flow during an ED visit.

Chang Gung Medical Foundation consists of several hospitals located throughout Taiwan. One of the largest hospitals in northern Taiwan, the Chang Gung Memorial Hospital (CGMH) based in Linkou County han- dles N 180,000 ED visits annually. Similarly, CGMH based in Kaohsiung City manages a large portion of ED patients in southern Taiwan. Both hospitals need an efficient, high-functioning ED. Although they shared the same system, the two ED adopted different approaches for manag- ing patients. At both EDs, patients are triaged upon entry using the stan- dard Taiwan Triage and Acuity Scale developed by the Ministry of Health and Welfare [7]. At Linkou CGMH, patients are assigned to dif- ferent examination areas according to the severity of their complaints. An ED physician specifically responsible for critical area will treat pa- tients categorized as level 1 (resuscitation) and level 2 (emergency). Pa- tients ranging from level 3 to level 5 (urgent to non-urgent) will be

https://doi.org/10.1016/j.ajem.2017.11.010

0735-6757/(C) 2017

assigned to a non-critical area where they will wait for treatment by other ED physicians. As for CGMH based in Kaohsiung City, after triaged, all patients are randomly assigned to an ED physician. That physician will be solely responsible for the patients despite their severity level.

The purpose of this study is to compare the two models in terms of clinical efficiency, diagnostic tool use, and patient dispositions. The aim is to determine whether one approach is superior and whether it helps to facilitate patient flow in the ED.

Materials and methods

Study design

This was a retrospective, 1-year cohort study approved by the insti- tutional review board of the Chang Gung Medical Foundation. All pa- tients’ and physicians’ records and information were anonymized and de-identified before analysis.

Study setting and participants

This study was conducted in two ED of the tertiary referral medical centers located in northern and southern Taiwan separately. Beds of these two EDs were 80 and 60, and total beds in observation rooms were 160 and 148, respectively. Observation rooms are designed for short stays to follow clinical changes or waiting for hospital admission. There are over 3500 and 2500 inpatient beds in the two hospitals.

From 1 July 2011 to 30 June 2012, all adult non-trauma patients who presented to the two EDs from 10:00 to 14:00 were included in the analysis. Overall, 76 full-time attending physicians were involved in this study; 59 worked in the northern ED, and 17 worked in the south- ern one. All 76 attending physicians were qualified emergency physi- cians and received the same residency training program developed by the Taiwan Society of Emergency Medicine. On the other hand, the nursing levels are the same for both EDs. All nurses in the two EDs have a college degree in nursing, and through the national examination. The ED visits were classified into different disease acuities based on the five-level Taiwan Triage and Acuity Scale (TTAS), a commonly used triage system formulated by the Ministry of Health and Welfare in Tai- wan [7]. Based on TTAS, patient acuity was determined according to presenting vital signs (heart rate, blood pressure, respiration rate, oxy- gen saturation) and the main problem. For example, a patient present- ing with dyspnea and unstable vital signs would be determined as triage II, or even triage I if immediate resuscitation is needed. According to these criteria, patients identified as triage levels 1 and 2 should be seen immediately or within 10 min, respectively, and are defined as ur- gent. Patients with triage levels 3, 4, and 5 should be assessed within 30,

60, or 120 min, respectively, and are classified as non-urgent.

In the northern ED, where three attending physicians worked at the same time, patients were assigned to different attending physicians ac- cording to their disease acuity according to the TTAS. One physician treated level 1 and 2 triage patients (urgent ones); the other two EPs treated level 3, 4, and 5 patients (non-urgent ones). In the southern ED, three attending physicians worked at the same time, and a comput- er assigned presenting patients in rotation to attending physicians. As all study sites were teaching medical units, residents assisted in the treatment of ED patients under an attending physician’s supervision. In the northern ED, there were 7 nurses working together, and in the southern one were 6. The northern ED practice was defined as a separat- ed model, and the southern one was defined as a merged model.

Measures

Patients’ diagnoses were categorized into 7 groups according to the In- ternational Classification of Diseases-9 (ICD-9) including nervous sys- tem disease (ICD-9-CM: 320-389 and 430-438), gastrointestinal disease (ICD-9-CM: 520-579), genitourinary disease (ICD-9-CM: 580- 629), pulmonary disease (ICD-9-CM: 460-519), cardiovascular disease

(ICD-9-CM: 390-429 and 439-459), neoplasms (ICD-9-CM: 140-239),

and others. Because the two study hospitals were teaching medical units, residents assisted in the treatment of ED patients under an at- tending physician’s supervision, so the visits were divided into super- vised visits and attending-alone visits. All supervised visits were initially evaluated and treated by residents; attending physician con- sults were always required. Considering the difference of visit numbers in the two EDs, the ED occupancy status, determined by the number of patients staying during their time of visit, was used to control the influ- ence of the ED crowding. The occupancy status was grouped into four levels according to the number of patients staying in ED, divided into quartiles [8,9]. The ED LOS, diagnostic tool use, and patient disposition were treated as outcome variables. The ED LOS was defined from the initial time that the patient presented to the ED as documented by the triage nurse to the final time that the patient left the ED. ED LOS were calculated using the following four points: EP completing initial Patient evaluation (the timing of first order or prescription), discharge from ED, admission to general ward, and admission to ICU. As diagnostic-tool use outcomes, we included computed tomography (CT) and any laboratory examinations (e.g., complete blood count, blood chemistry, urine analy- sis, stool analysis, or influenza screen test). The patient dispositions were classified into discharge, hospital admission (including general ward and ICU), and ED mortality [10].

Data analysis

For continuous variable (age), the data were summarized as means and standard deviations (SD). Because the distributions of ED LOS were not normal, we used medians with interquartile ranges (IQRs). The distributions of categorical variables including sex, triage, diagnosis, visit type, crowding status, patient disposition, and diagnostic tool use were presented with numbers and percentages. Student’s t-test, Mann-Whitney U test, and chi-square tests were used to evaluate the associations among these variables and the two ED practice models. To analyze the associations of the outcome variables with the two prac- tice models adjusting for the potential confounding factors, multinomial logistic regression was selected for patient disposition, and binomial lo- gistic regression for diagnostic tool use. Effects were estimated in terms of odds ratios (ORs) and corresponding 95% confidence intervals (CIs).

Table 1

Demographics and diagnosis of the patients and occupancy status in the emergency departments.

	Separated model N = 21,630	Merged model N = 15,950	p-Value
Age	56.3 +- 19.36	57.9 +- 18.66	b0.001
Male	11,264 (52.1%)	7908 (49.6%)	b0.001
Urgent	5776 (26.7%)	3296 (20.7%)	b0.001
diagnostic category Nervous	2992 (13.8%)	2437 (15.3%)	b0.001
Gastrointestinal	4398 (20.3%)	2971 (18.6%)
Genitourinary	2256 (10.4%)	1363 (8.5%)
Pulmonary	3209 (14.8%)	2279 (14.3%)
Cardiovascular	1621 (7.5%)	1274 (5.0%)
Neoplasm	1551 (7.2%)	1169 (7.3%)
Other	5603 (25.9%)	4457 (27.9%)
Supervised visits ED occupancy status	4298 (19.9%)	7079 (44.4%)	b0.001 b0.001
(stand-by patient number)

Variables were extracted from the ED administrative database and	1st quartile (b29)	994 (4.6%)	8655 (54.3%)
included patients’ age, sex, triage, diagnosis, visit type (seen by attend-	2nd quartile (29-43)	3554 (16.4%)	5641 (35.4%)

ing alone or resident under attending supervision), occupancy status upon ED arrival, patient disposition, diagnostic tool use, and ED LOS.

3rd quartile (43-67)	8177 (37.8%)	1636 (10.3%)
4th quartile (N 67)	8905 (41.2%)	18 (0.1%)

Significance testing was two-sided, and the significance threshold was set at pb 0.05. SPSS version 12.0 (SPSS, Chicago, IL) was used for all sta- tistical analyses.

Results

Patient characteristics

During the study period, there were 37,580 ED visits. Among them, 21,630 were treated as per separated model, and 15,950 were treated

as per merge model. The number of patients seen per hour by a physi- cian in Linkou was 4.9 and in Kaohsiung was 3.6 (p b 0.001). The mean patient number which a nurse need to care per hour was 2.1 in Linkou and in Kaohsiung was 1.8 (p b 0.001). The patients’ characteris- tics of visits were shown in Table 1. The mean age of patients treated by the separated model was 1.6 years younger than those treated by the merge model (p b 0.001). Male patients and urgent ones in the separat- ed model were 2.5% and 6.0% more than those in the merged model (both p b 0.001). The distribution of the Diagnostic categories of the ED visits was statistically different (0 b 0.001). The supervised visits

Fig. 1. A The distribution of patient dispositions in the separated model and merged model (p b 0.001). B The median of ED length of stay in the separated model and merged model of the different dispositions. C The use rate of laboratory examination and computed tomography in the separated model and merged model.

Table 2

The influence of separated model on patient disposition, adjust for patient’s age, sex, dis- ease acuity, diagnosis, visit type, and ED occupancy status by multinomial logistic regres- sion with discharge as reference category.

routines where patients are divided into different streams according to certain criteria set forth by the ED administration. Fast track is a form of streaming used to handle patients with lower level of emergen-

cy. Team triage is done by handling triage using a team that includes a

Separated model Merged model

	aOR	95% C.I.		Ref
Admission to General ward	1.0	0.90-1.04		1
Admission to ICU	1.0	0.80-1.14		1
ED mortality	0.8	0.55-1.13		1

aOR (Adjusted odds ratio), CI (confidence interval), ED (Emergency department).

were 24.5% fewer in the separated model than in the merged model (p b 0.001). In the separated model, the ED occupancy status was obviously higher than the merged model.

Clinical practice

The patient disposition in the two models was different. In patients treated by the separated model, there were 1.8% fewer patients discharged from the ED, 1.2% and 0.5% more patients admitted to gener- al ward and ICU, and 0.1% more ED mortality compared to those treated by the merged model (p = 0.001; Fig. 1A). The median of ED LOS in the separated model and merged model was different in patients discharged from ED (p b 0.001) and admitted to ICU (p = 0.002) (Fig. 1B). In the separated model, 2.1% fewer patients received laboratory ex- amination, and there was no difference in the use of CT (Fig. 1C). After adjusting for the potential confounding factors including patient’s age, sex, disease acuity, diagnosis, visit type, and ED occupancy status by multinomial logistic regression with ED discharge as reference, there was no difference among patient disposition and ED mortality rate (Table 2); by binomial logistic regression, patients treated by the sepa- rated model received fewer laboratory examinations (aOR: 0.9, CI: 0.83-0.96), but there was still no difference in the use of CT (Table 3).

Discussion

ED overcrowding is complicated and multifactorial. The input- throughput-output conceptual model proposed by Asplin et al. in 2003 had become an accepted approach toward understanding causes of overcrowding [11]. As suggested by the model, causes may be catego- rized into the three domains, and solutions to reduce overcrowding should be directed toward input, throughput, or output. The input com- ponent includes any condition, event, or system characteristic that con- tributes to demand for ED services. The throughput component of the model, which is the focus of our study, identifies patient length of stay in the ED, and the output component discusses inefficient disposition of ED patients. Asplin et al. further divided the throughput component into two phases. The first phase includes triage, room placement, and initial provider evaluation. The second phase of the component includes diagnostic testing and ED treatment. Over the years, many hypotheses and strategies had been developed to reduce throughput time. Oredsson et al. reviewed triage-related interventions to improve patient flow; interventions included streaming, fast track, team triage, point-of- care testing, and nurse-requested x-ray [12]. Streaming refers to

Table 3 The influence of separated model on the use rate of laboratory examination and computed tomography, adjust for patient’s age, sex, disease acuity, diagnosis, visit type, and ED occu- pancy status by binomial logistic regression.

Separated model Merged model

	aOR	95% C.I.		Ref
Laboratory examination	0.9	0.83-0.96		1
Computed tomography	1.0	0.92-1.10		1

aOR (Adjusted odds ratio), CI (confidence interval), ED (Emergency department), Inten- sive care unit (ICU).

physician to increase accuracy in initial evaluation of patients, and Point-of-care testing (POCT) is done by setting up an independent lab- oratory in the ED specific for examinations and related analyses. Finally, nurse-requested x-ray is a method used to shorten waiting time when nurses can order x-rays for patients before they are seen by a doctor. All these interventions were found to be useful for improving patient flow, with fast track being the best practice. Other studies had also discussed the application of lean principle from the Toyota production system to reduce wait times in the ED [13]. Our study compared two dif- ferent models; the separated model is somewhat a combination of streaming and team assignment, a system mentioned by Patel et al. [14]. In their 2005 study, they implemented a team assignment system in which patients were assigned to a specific care team in rotation upon arrival. Results showed that by using this system, time to physician as- sessment was reduced, fewer patients left the ED without being seen by a doctor, and overall patient satisfaction improved.

Results of our study have shown that by adopting the separated model we can decrease the use of laboratory examinations, hence de- creasing waiting time in the ED and therefore expediting patient flow. We chose the use of CT and laboratory examinations as a study focus be- cause these are the most time-consuming examinations in the ED. We suspected that when ED physicians had to treat patients with all levels of disease severity, it is easier for them to over-estimate the degree of severity. Another explanation for using more examinations in the merged model may be that more time is needed for physicians to eval- uate urgent patients, which results in having less time for less urgent patients. To maintain a certain level of patient satisfaction, the use of laboratory examinations can provide these patients with a feeling of being attended to, instead of being ignored because they are less sick. After receiving these examinations, ED physicians need to further ex- plain the results to patients. Boudreaux et al. reviewed factors contribut- ing to patient satisfaction in the ED [15]. Most studies concluded that the key to patient satisfaction was interpersonal interactions and the delivery of information. With objective information such as examina- tion results, it provides a more understandable proof to the patient about what is happening to them and in turn compensates for the fact that they spend less time with an actual physician. However, more stud- ies will be needed to confirm these conjectures.

This study has several limitations that should be considered when interpreting the results. First, the two study EDs belonged to the same healthcare system, potentially limiting the implications of the conclu- sions. For example, in the EDs where the study conducted, emergency specialists were only responsible for adult non-trauma patient. The sur- geons treated trauma patients in the EDs. However, the surgeons had different training course, and they did not treat non-trauma patients, so we excluded trauma patients from the study. Second, although the two EDs involved in this study belonged to the same healthcare founda- tion, EPs working in these two EDs belong to two different teams. This may affect our study results. However, the EPs working in both EDs had received the same residency training and used the same medical system, which we believed can minimize its effect on our outcome. Third, also given the retrospective design of this study, some confound- ing factors may have not been well controlled. Lastly, practice patterns in Taiwan are remarkably different in some ways as compared with those in the United States and other Western countries. We believe that these limitations may influence the interpretation of our results by other medical systems.

Acknowledgments

The authors gratefully acknowledge the support by Research Grants from the Kaohsiung Chang Gung Memorial Hospital (CMRP-G8C0361).

Chao-Jui Li would like to thank his daughter, Xuan-Le Li, for the inspira- tion of this research.

Data sharing

No additional data available.

Conflicts of interest statement

No conflict of interest for all authors.

Funding Sources/Disclosures

This study was supported in part by research grants from the Kaohsiung Chang Gung Memorial Hospital (CMRP-G8C0361).

References

1. Dickinson G. Emergency department overcrowding. CMAJ 1989;140(3):270-1.
2. Miro O, et al. Decreased health care quality associated with emergency department overcrowding. Eur J Emerg Med 1999;6(2):105-7.
3. Richardson DB. Increase in patient mortality at 10 days associated with emergency department overcrowding. Med J Aust 2006;184(5):213-6.
4. Chang BP, et al. Patient treatment in ED hallways and patient perception of clinician- patient communication. Am J Emerg Med 2016;34(6):1163-4.
5. Sun BC, et al. Effect of emergency department crowding on outcomes of admitted patients. Ann Emerg Med 2013;61(6):605-11 e6.
6. Derlet RW, Richards JR. Overcrowding in the nation’s emergency departments: com- plex causes and disturbing effects. Ann Emerg Med 2000;35(1):63-8.
7. Ng CJ, et al. Validation of the Taiwan triage and acuity scale: a new computerised

   five-level triage system. Emerg Med J 2011;28(12):1026-31.

   McCusker J, et al. Increases in emergency department occupancy are associated with adverse 30-day outcomes. Acad Emerg Med 2014;21(10):1092-100.

8. Verelst S, et al. Emergency department crowding in relation to in-hospital adverse medical events: a large prospective observational cohort study. J Emerg Med 2015;49(6):949-61.
9. Van Leer PE, et al. Resident supervision and patient safety: do different levels of res- ident supervision affect the rate of morbidity and mortality cases? J Emerg Med 2015;49(6):944-8.
10. Asplin BR, et al. A conceptual model of emergency department crowding. Ann Emerg

    Med 2003;42(2):173-80.

    Oredsson S, et al. A systematic review of triage-related interventions to improve patient flow in emergency departments. Scand J Trauma Resusc Emerg Med 2011; 19:43. https://doi.org/10.1186/1757-7241-19-43.

11. Ng D, et al. Applying the lean principles of the Toyota production system to reduce wait times in the emergency department. CJEM 2010;12(1):50-7.
12. Patel PB, Vinson DR. Team assignment system: expediting emergency department care. Ann Emerg Med 2005;46(6):499-506.
13. Boudreaux ED, O’Hea EL. Patient satisfaction in the emergency department: a review of the literature and implications for practice. J Emerg Med 2004;26(1):13-26.