patient length of stay in “>American Journal of Emergency Medicine (2012) 30, 1860-1864

Original Contribution

Predictors of patient length of stay in 9 emergency departments

Jennifer L. Wiler MD, MBA ^a,b,?, Daniel A. Handel MD, MPH ^c, Adit A. Ginde MD, MPH ^a, Dominik Aronsky MD, PhD ^d, Nicholas G. Genes MD, PhD ^e, Jeffrey L. Hackman MD ^f, Joshua A. Hilton MD ^g, Ula Hwang MD, MPH ^e,h, Michael Kamali MD ⁱ,

Jesse M. Pines MD ^j, Emilie Powell MD ^k,

Medhi Sattarian MD ^l, Rongwei Fu PhD ^m

^aDepartment of Emergency Medicine, University of Colorado Denver School of Medicine, Aurora, CO 80045, USA ^bDivision of Emergency Medicine; Washington University in St Louis School of Medicine, St Louis, MO 63110, USA ^cCenter for Policy and Research in Emergency Medicine, Department of Emergency Medicine,

Oregon Health and Science University, Portland, OR 97239, USA

^dDepartment of BioMedical Informatics and Emergency Medicine, Vanderbilt University, Nashville, TN 37232, USA

^eDepartment of Emergency Medicine, Mount Sinai School of Medicine, New York, NY 10029, USA

^fDepartment of Emergency Medicine, University of Missouri-Kansas City, Truman Medical Center,

Kansas City, MO 64139, USA

^gDepartment of Emergency Medicine, University of Pennsylvania, Philadelphia, PA 19104, USA ^hGeriatric Research, Education and Clinical Center, James J. Peters VAMC, Bronx, NY 10468, USA ⁱDepartment of Emergency Medicine, University of Rochester, Rochester, NY 14642, USA

^jDepartments of Emergency Medicine, George Washington University School of Medicine and Department of Health Policy,

George Washington School of Public Health, Washington, DC 20037, USA

^kDepartment of Emergency Medicine, Northwestern University Feinberg School of Medicine, Chicago, IL 60611, USA

^lDepartment of Emergency Medicine, George Washington University, Washington, DC 20037, USA

^mDepartment of Emergency Medicine and Public Health and Preventive Medicine, Oregon Health and Science University, Portland, OR 97239, USA

Received 23 July 2011; revised 20 March 2012; accepted 29 March 2012

Abstract

Objectives: Prolonged emergency department (ED) length of stay (LOS) is linked to adverse outcomes, decreased patient satisfaction, and ED crowding. This multicenter study identified factors associated with increased LOS.

Methods: This retrospective study included 9 EDs from across the United States. Emergency department daily operational metrics were collected from calendar year 2009. A multivariable linear population average model was used with log-transformed LOS as the dependent variable to identify which ED operational variables are predictors of LOS for ED discharged, admitted, and overall ED patient categories.

* Corresponding author. Department of Emergency Medicine, University of Colorado Denver SOM, Aurora, CO 80045, USA. Tel.: +1 720 848 5569 (office), +1 716 390 1288 (cell).

E-mail address: [email protected] (J.L. Wiler).

0735-6757/$ - see front matter (C) 2012 http://dx.doi.org/10.1016/j.ajem.2012.03.028

Results: Annual ED census ranged from 43 000 to 101 000 patients. The number of ED treatment beds ranged from 27 to 95. Median overall LOS for all sites was 5.4 hours. Daily percentage of admitted patients was found to be a significant predictor of discharged and admitted patient LOS. Higher daily percentage of discharged and eloped patients, more hours on Ambulance diversion, and weekday (vs weekend) of patient presentation were significantly associated with Prolonged LOS for discharged and admitted patients (P b .05). For each percentage of increase in discharged patients, there was a 1% associated decrease in overall LOS, whereas each percentage of increase in eloped patients was associated with a 1.2% increase in LOS.

Conclusions: Length of stay was increased on days with higher percentage daily admissions, higher elopements, higher periods of ambulance diversion, and during weekdays, whereas LOS was decreased on days with higher numbers of discharges and weekends. This is the first study to demonstrate this association across a broad group of hospitals.

(C) 2012

Introduction

Emergency department (ED) crowding has received considerable national attention in recent years [1-3]. Emergency department patients have had increasingly longer wait times and lengths of stay (LOSs) over the last decade, indicating that the problem of ED crowding is worsening across the United States [4,5]. Increased ED LOS has been associated with Negative outcomes such as increased mortality and complication rates [6-9], decreased satisfaction [10-12], and higher levels of ED crowding [1,9]. Emergency department patient ftow and treatment times, including LOS, are affected by numerous complex factors such as hospital occupancy rates [1,13,14], patient acuity [15], triage in- terventions [16-19], and staffing [20]. However, many studies evaluating factors that inftuence ED patient LOS have been conducted, thereby limiting generalizability [14,16-20].

The ability to predict and, subsequently, mitigate factors prolonging ED LOS is important to improve ED care and efficiency. This is the first multicenter study that evaluates predictors of ED LOS across 9 geographically disparate institutions.

Methods

Study design

This was a retrospective, multicenter study of 9 EDs from across the United States (1 community hospital and 8 academic medical centers). Daily ED operational metric data were collected for a 12-month period (calendar year 2009). The institutional review board of each participating institution either approved the study protocol or exempted it from review.

Participating study sites were selected based on their geographic diversity, the interest of local investigators, and range of annual censuses during the study period. Study sites needed to be able to provide daily ED level data to be included.

Study protocol and data collection

A standardized template was developed to collect the data and provided to each site. Each site was responsible for compiling its own data and returning these to the designated data coordinating center. The data coordinating center reviewed the submitted data to assure consistency in terms of measurement units before being compiled. Daily variables included the total number of ED patients, total number of ED patients admitted and discharged, average ED overall LOS, average LOS for admitted and discharged patients, total number of ED elopements, number of ED beds, and the number of hours an ED was on ambulance diversion each day. The total number of ED patients included both adult and pediatric patients seen each day; the calculation of average overall LOS also included both adult and pediatric patients. The primary study outcome was daily mean LOS, analyzed by total, admitted, and discharged patients. Length of stay was defined as the time from first arrival in the ED to when the patient left the ED for their final destination.

Emergency department operational variables were

assessed as predictors of LOS. To control for ED and hospital size, we calculated several standardized predictor variables including standardized “total patient volume” (no. of total ED patients/no. of ED beds), standardized “admittED patient volume” (no. of admitted patients/no. of Hospital beds), and percentage of discharged and left-without-being- seen patients each day. All data were compiled in Microsoft Excel (Microsoft Corporation, Redmond, WA).

Data analysis

Institutional characteristics were summarized by descrip- tive statistics. A multivariable linear population average model was used with log-transformed mean LOS as the dependent variable to identify predictors of LOS while controlling for clustering within each site. This analysis was performed for overall LOS and subgroup LOS for patients admitted to the hospital and for those discharged from the ED. All models also controlled for month as a way to adjust for seasonality and whether the day was a weekend vs

weekend. Log-transformed LOS was used to satisfy the assumption of normality. Therefore, the exponentiated coefficients were interpreted as the ratio of LOS between the 2 levels in comparison. For example, for a categorical independent variable such as weekend, the exponentiated coefficient represented the ratio of mean LOS of a weekend day to mean LOS of a weekday. A ratio less than 1 indicated a decrease in LOS, and a ratio greater than 1, an increase of LOS, as compared with the reference group. Equivalently, we could also interpret the ratio as the percentage decrease (ratio b1) or increase (ratio N1). For example, for a categorical variable, a ratio of 1.10 means a 10% increase in mean LOS compared with the reference group. For a continuous variable, it is the ratio of LOS for each unit increase of that variable. We reported this ratio (ratio of means [ROM]) in tables and interpreted the results as percentage increase or decrease in the results section. Associations between LOS and each independent variable were investigated first in a univariate analysis; variables with P <= .20 were then considered in the multivariate model. P b .05 was considered significant in the final model. In addition, linearity between log of LOS and continuous variables was assessed using locally weighted scatterplot smooth curve and testing the categorized version of the continuous variables in the model [21]. If the relationship between log of LOS and a continuous variable is not linear, the continuous variable was categorized and entered into the model as a categorical variable. In particular, the variable of percentage of admitted patients was categorized based on quartiles, and hours on ambulance diversion was catego- rized as a 5-level variable with no diversion as 1 group and the rest of the data roughly equally divided into 4 groups as 0 b hours b 2, 2 <= hours b 5, 5 <= hours b 10, and hours >= 10. All analyses were performed using SAS version 9.1.3 (SAS Institute, Cary, NC).

Results

Demographic data for the 9 participating EDs sites are listed in Table 1. All sites except 1 were in an urban

Table 1 Study site demographics

Variables ROM 95% CI P

Percentage of discharged patients 0.990 0.989-0.991 b.001 Percentage of elopements 1.012 1.009-1.014 b.001 Time on diversion (h/d) b.001 No diversion Reference 0 b h b 2 1.012 0.991-1.033 .256 2 <= h b 5 1.030 1.010-1.052 .004 5 <= h b 10 1.058 1.036-1.080 b.001 h >= 10 1.079 1.054-1.105 b.001 Weekend day (Y/N) 0.966 0.955-0.977? b.001

Y indicates yes; N, no.

academic setting. The annual ED census ranged from 43 000 to 101 000 patients. The number of ED treatment beds ranged from 27 to 95. Only 1 site reported a change in the number of ED beds available during the study period. The median overall LOS for the 9 sites was 5.4 hours (range, 2.6- 33.7 hours), with a median LOS of 8.0 hours (range, 3.4-98.3 hours) for admitted and 4.5 hours (range, 2.2-40.3 hours) for discharged patients. The percentage of days that an ED was on diversion at least once during that 24-hour period varied from 4.9% to 86.6%. Daily admission rates at the 9 sites ranged from 4.3% to 15.5%.

Table 2 Emergency department overall LOS

In the final multivariable model, daily percentage of discharged and eloped patients, hours on ambulance diversion, and weekend (vs weekend) day of patient presentation were significantly associated with overall daily mean LOS (P b .05; Table 2). Specifically, for each percentage increase in discharged patients, there was a 1% associated decrease in overall daily mean LOS (ROM, 0.990; 95% confidence interval [CI], 0.989-0.991), whereas each percentage of increase in eloped patients was associated with a 1.2% increase in LOS (ROM, 1.012; 95% CI, 1.009-

1.014). Compared with days without diversion, days that had higher ambulance diversion hours were associated with a longer LOS, although the association was not significant for days with only 0 to 2 hours of diversion. Lastly, weekend days were associated with shorter LOS compared with weekdays (Table 2). Other variables were not associated with overall daily mean LOS.

Site	Annual census (2009)	Average overall LOS (h)	Average admitted LOS (h)	Average discharged LOS (h)	Days with ambulance diversion (%)	ED beds	Type of practice
1	43 028	4.1	5.5	3.7	14.0	27-33	Academic
2	50 259	5.0	8.1	4.1	13.4	33	Community
3	56 756	5.2	6.7	4.6	43.8	48	Academic
4	57 911	7.6	12.5	5.7	86.6	45	Academic
5	59 453	4.7	8.2	3.9	5.5	48	Academic
6	62 186	7.2	13.6	5.4	75.1	48	Academic
7	85 066	4.9	7.8	4.0	4.9	56	Academic
8	98 214	21.6	8.8	24.8	71.0	95	Academic
9	101 236	5.3	8.6	4.3	16.7	44	Academic

Variables	ROM 95% CI	P
Percentage of discharged patients	0.998 0.996-0.999	.008
Percentage of elopements	1.017 1.015-1.019	b.001
Time on diversion (h/d)		b.001
No diversion	Reference
0 b h b 2	1.024 1.006-1.043	.010
2 <= h b 5	1.027 1.009-1.046	.003
5 <= h b 10	1.045 1.026-1.064	b.001
h >= 10	1.036 1.015-1.058	.001
Admitted patients		.0065
(quartiles percentage)
1.5%-6.1%	Reference
6.2%-8.3%	0.993 0.971-1.015	.519
8.4%-10.6%	1.017 0.991-1.043	.197
10.7%-43.0%	1.022 0.990-1.054	.181
Weekend day (Y/N)	0.988 0.979-0.998	.020

Association between daily mean LOS and ED operational variables is reported in Table 3 for patients discharged from the ED. Similar to the overall daily mean LOS, increased daily percentage of discharged patients and weekend days (vs weekdays) were significantly associated with shorter LOS, and increased percentage of eloped patients and increased hours on ambulance diversion were associated with longer daily mean LOS. In addition, daily percentage of admitted patients was found to be a significant predictor of daily mean LOS (P = .0065). Compared with days with fewer patient admissions (1.5%-6.1%), there was no significant increase in LOS associated with a higher percentage of admitted patients (N6.1%) (Table 3). However, compared with days when 6.2% to 8.3% of patients were admitted, there was a 2.4% (ROM, 1.024; 95% CI, 1.010-1.039) and 2.9% (ROM, 1.029; 95%

Table 3 Emergency department LOS for discharged patients

CI, 1.007-1.051) significant increase in LOS, respectively, on days when 8.4% to 10.6% of patients and greater than or equal to 10.7% of patients were admitted.

Table 4 reports the association between daily mean LOS and significant ED operational variables for patients admitted to the hospital. Percentage of eloped and admitted patients and hours of ambulance diversion were all positively associated with LOS, and weekends (vs weekdays) were negatively associated with LOS, similar to LOS for discharged patients. In addition, the patient volume (total no. of patients/no. of ED beds) was found to be negatively associated with LOS. After adjusting for the percentage of admitted patients and other variables in the model, each additional patient per ED bed was associated with 3.8% (ROM, 0.962; 95% CI, 0.946-0.978)

decrease in LOS for discharged patients.

Discussion

Prolonged ED LOS has a known negative association on patient satisfaction [10-12] and quality outcomes [6-9]. Emergency department LOS is determined by several factors

including hospital-based census issues that have a down- stream effect on ED crowding and patient LOS [5,9,14,22]. Many have reported factors that inftuence ED patient LOS from a single institution [14,16-20]. However, to our knowledge, this is the first multicenter validation study performed to define ED predictors of patient LOS.

We found that the daily percentage of admitted patients was a significant predictor of daily mean LOS for discharged ED patients, as was any amount of ambulance diversion. This is consistent with other reports that found that ED boarding and ambulance diversion are a surrogate marker for impaired ED efficiency and ftow [9,14,23,24]. A shorter overall ED LOS was associated with weekend patient presentation and an increased percentage of discharged patients. One previous study reported that more low-acuity patients used the ED on the weekend than during the week in 1 state (Nebraska) [25]. It is unknown if this finding mirrors national trends, but our finding that LOS is shorter for discharged patients on weekends appears to be consistent with this report. For each percentage increase in discharged patients, there was 1% associated decrease in overall daily mean LOS for all ED patients, and the daily percentage of admitted patients was found to be a significant predictor of daily mean LOS. This is consistent with previous reports that found that, as the demand for ED services grows, particularly from higher acuity patients, operational ftow metrics are impaired and ED crowding occurs [2,3]. What is novel is the quantification of the impact of discharges on LOS for all patients. Operational improvement strategies that focus on improving the through- put of discharged patient cohorts, including improving outpatient Care coordination to ensure timely outpatient follow-up, may improve ED LOS for all patient cohorts but may be limited in settings that have high patient acuity.

For patients who are admitted to the hospital from the ED, higher total ED patient acuity (ie, percentage of admitted patients), a higher percentage of elopements, and ambulance diversion greater than 2 hours per day were associated with

Table 4 Emergency department LOS for admitted patients

Variables	ROM 95% CI	P
Total no. of patients/ED beds	0.962 0.946-0.978	b.001
Percentage of elopements	1.021 1.018-1.024	b.001
Time on diversion (h/d)
No diversion	Reference
0 b h b 2	1.021 0.990-1.058	.168
2 <= h b 5	1.034 1.001-1.069	.045
5 <= h b 10	1.075 1.040-1.112	b.001
h >= 10	1.142 1.099-1.186	b.001
Admitted patients		b.001
(quartiles percentage)
1.5%-6.1%	Reference
6.2%-8.3%	1.075 1.033-1.118	b.001
8.4%-10.6%	1.126 1.079-1.175	b.001
10.7%-43.0%	1.134 1.080-1.192	b.001
Weekend day (Y/N)	0.905 0.888-0.922	b.001

an increased admitted ED patient LOS. This is not surprising as the number of patients who require inpatient admission rises, demand outpaces resources and ED boarding, and crowding develops, which has a known negative impact on ED operations [1-3].

Identifying factors that positively or negatively impact LOS for all ED patients as well as admitted or discharged subgroups is an important element to understanding and improving ED patient care delivery and operations. As such, these findings can more appropriately be generalized to other institutions to help inform targeted ED operational improve- ment strategies.

Limitations

A convenience sample of sites was selected for this study and depended on the local investigator’s interest in ED Crowding research and access to electronic data. Although 1 community hospital was included and the sites were geographically diverse, most sites were large academic medical centers, which may reduce the generalizability of the findings to other settings.

Given the scope of this study, only daily operational metrics data were able to be collected, which excluded important factors such as individual patient acuity, duration of diversion episodes, and staffing levels. Of the 32 variables originally considered, data for 12 were not available at all participating sites (eg, daily staffing level variations). Some variables were not reliably collected and were also not included (eg, elopements and left-without-being-seen patients). The associ- ation between variables is complex; the possible interactions between all factors and those relationships were not studied.

Conclusions

Emergency department operational variables are associ- ated with LOS across these 9 hospitals. Length of stay was increased on days with higher elopements and higher periods of ambulance diversion, whereas LOS was decreased on days with higher numbers of discharges and during weekends. The daily percentage of admitted patients was found to be a significant predictor of daily mean LOS. This is the first study to demonstrate this in a multicenter fashion.

References

1. Moskop JC, Sklar DP, Geiderman JM, et al. Emergency department crowding, part 1: concept, causes, and moral consequences. Ann Emerg Med 2009;53:605-11.
2. Hoot NR, Aronsky D. Systematic review of emergency department crowding: causes, effects, and solutions. Ann Emerg Med 2008;52: 126-36.
3. Derlet RW. Overcrowding in emergency departments: increased demand and decreasED capacity. Ann Emerg Med 2002;39:430-2.
4. Wilper AP, Woolhandler S, Lasser KE, et al. Waits to see an emergency department physician: U.S. trends and predictors, 1997- 2004. Health Aff (Millwood) 2008;27(2):w84-95.
5. Horwitz LI, Green J, Bradley EH. US emergency department performance on wait time and length of visit. Ann Emerg Med 2010;55(2):133-41.
6. Liew D, Liew D, Kennedy MP. Emergency department length of stay independently predicts excess inpatient length of stay. Med J Aust 2003;179(10):524-6.
7. Trzeciak S, Rivers EP. Emergency department overcrowding in the United States: an emerging threat to patient safety and public health. Emerg Med J 2003;20:402-5.
8. Bernstein SL, Aronsky D, Duseja R, et al. The effect of emergency department crowding on clinically oriented outcomes. Acad Emerg Med 2009;16:1-10.
9. McCarthy ML, Zeger SL, Ding R, et al. Crowding delays treatment and lengthens emergency department length of stay, even among high- acuity patients. Ann Emerg Med 2009;54(4):492-503 e4.
10. Pines JM, Hollander JE. Emergency department crowding is associated with poor care for patients with severe pain. Ann Emerg Med 2008;51:1-5.
11. Rodi SW, Grau MV, Orsini CM. Evaluation of a fast track unit: alignment of resources and demand results in improved satisfaction and decreased length of stay for emergency department patients. Qual Manag Health Care 2006;15(3):163-70.
12. Taylor C, Benger JR. Patient satisfaction in emergency medicine. Emerg Med J 2004;21(5):528-32.
13. Lucas R, Farley H, Twanmoh J, et al. Emergency department patient flow: the influence of hospital census variables on emergency department length of stay. Acad Emerg Med 2009; 16(7):597-602.
14. Rathlev NK, Chessare J, Olshaker J, et al. Time series analysis of variables associated with daily mean emergency department length of stay. Ann Emerg Med 2007;49(3):265-71.
15. Gardner RL, Sarkar U, Maselli JH, et al. Factors associated with longer ED lengths of stay. Am J Emerg Med 2007;25(6):643-50.
16. Partovi SN, Nelson BK, Bryan ED, et al. Faculty triage shortens emergency department length of stay. Acad Emerg Med 2001;8(10): 990-5.
17. Chan TC, Killeen JP, Kelly D, et al. Impact of rapid entry and accelerated care at triage on reducing emergency department patient wait times, lengths of stay, and rate of Left without being seen. Ann Emerg Med 2005;46(6):491-7.
18. Russ S, Jones I, Aronsky D, Dittus RS, et al. Placing physician orders at triage: the effect on length of stay. Ann Emerg Med 2010;56(1):27-33.
19. Wiler JL, Gentle C, Halfpenny JM, et al. Optimizing emergency department front-end operations. Ann Emerg Med 2010;55(2): 142-60.e1.
20. McGarry J, Krall SP, McLaughlin T. Impact of resident physicians on emergency department throughput. West J Emerg Med 2010;11(4): 333-5.
21. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: John Wiley & Sons, Inc.; 2000. p. 92-115.
22. Forster AJ, Stiell I, Wells G, et al. The effect of hospital occupancy on emergency department length of stay and patient disposition. Acad Emerg Med 2003;10(2):127-33.
23. Handel DA, Pines J, Aronsky D, et al. Variations in crowding and ambulance diversion in nine emergency departments. Acad Emerg Med 2011;18(9):941-6.
24. Schull MJ, Lazier K, Vermeulen M, et al. Emergency department contributors to ambulance diversion: a quantitative analysis. Ann Emerg Med 2003;41(4):467-76.
25. Schoenfeld EM, McKay MP. Weekend emergency department visits in Nebraska: higher utilization, lower acuity. J Emerg Med 2010;38(4): 542-5.