a b s t r a c t

Introduction: Previous work has suggested that Emergency Department rotational patient assignment (a system in which patients are algorithmically assigned to physicians) is associated with immediate (first-year) improve- ments in operational metrics. We sought to determine if these improvements persisted over a longer follow-up period.

Methods: Single-site, retrospective analysis focused on years 2-4 post-implementation (follow-up) of a rotational patient assignment system. We compared Operational data for these years with previously published data from the last year of physician self-assignment and the first year of rotational patient assignment. We report data for patient characteristics, departmental characteristics and facility characteristics, as well as outcomes of length of stay (LOS), arrival to provider time (APT), and rate of patients who left before being seen (LBBS).

Results: There were 140,673 patient visits during the five year period; 138,501 (98.7%) were eligible for analysis. LOS, APT, and LBBS during follow-up remained improved vs. physician self-assignment, with improvements sim- ilar to those noted in the first year of implementation. Compared with the last year of physician self-assignment, approximate yearly average improvements during follow-up were a decrease in median LOS of 18 min (8% im- provement), a decrease in median APT of 21 min (54% improvement), and a decrease in LBBS of 0.69% (72% im- provement).

Conclusion: In a single facility study, rotational patient assignment was associated with sustained operational im- provements several years after implementation. These findings provide further evidence that rotational patient assignment is a viable strategy in front-end process redesign.

(C) 2018

1. Introduction

The Emergency Department (ED) front-end (that portion of the visit between patient arrival and the time a provider assumes definitive care of a patient [1]) is an area in which ED administration and personnel generally have a great deal of operational control. As such, it is often an area for innovation and redesign to improve patient flow. Examples

? This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

* Corresponding author at: Department of Emergency Medicine, Mayo Clinic Arizona, 5777 E. Mayo Boulevard, Room 1-134A, Phoenix, AZ 85054, United States.

E-mail address: [email protected] (S.J. Traub).

of front-end interventions include the use of a fast-track [2], Physician in triage [3,4], telemedical triage [5,6], bedside registration [7], virtual patient streaming [8], and split-flow models [9].

Rotational patient assignment is another ED front-end process rede-

sign, in which patients are algorithmically assigned to physicians. This workflow inverts a typical approach of relying on physicians to acquire (or “pick up”) new patients at their discretion. Previous work has shown that rotational patient assignment is associated with short-term opera- tional gains [10-15], but the Long-term effects of rotational patient as- signment on operational metrics are unclear.

Improving clinical operations is an overarching goal of virtually every ED. Operational improvements have the potential to improve clinical care, increase patient satisfaction, and improve financial

https://doi.org/10.1016/j.ajem.2017.12.045

0735-6757/(C) 2018

performance. Such improvements can also help to decrease ED crowding, a condition associated with a number of negative patient out- comes [16]. Knowing if an intervention was likely to produce long-term gains might help ED administrators and personnel decide if such an in- tervention was a worthwhile investment at their facility.

We report long-term operational metrics (years 2-4) in a rotational patient assignment system, with the goal of understanding the durabil- ity of improvements over time.

1. Methods
   1. Institutional review board

This work was part of a quality improvement initiative, and was identified as exempt by our Institutional Review Board process, with a waiver of the requirement for informed consent.

Study design and setting

This was a retrospective, single-site analysis of routinely gathered operational data.

The Mayo Clinic Arizona ED is located in tertiary care teaching hospi- tal in Phoenix, Arizona. The average yearly census during the 5 year pe- riod was 28,135 (range: 25,651-32,224), and the average admission rate was 31%. The ED has 24 rooms, and 9 hallway bays. The ED is staffed 24 h per day with board-certified emergency physicians. There are no advanced practice providers (nurse practitioners or physician assis- tants), and there is no fast-track. There is no Emergency Medicine train- ing program, although residents from multiple services rotate through the department as learners and are involved in the evaluation of ap- proximately 5% of visits. Residents acquire patients ad libitum, without having patients assigned to them, and may see patients assigned to any attending physician.

Selection of participants

We electronically reviewed all patient visits identified by our electron- ic medical record (EMR; Cerner(R); Kansas City) within a five-year period (June 12, 2011-June 11, 2016) that included our last year of physician self-assignment and the first four years of rotational patient assignment. We have previously reported a comparison of operational metrics for our last year of physician self-assignment (June 12, 2011 to June 11, 2012) and our first year of rotational patient assignment (June 12, 2012 to June 11, 2013) [14]. We include data from our original publica- tion herein for reference and as a part of a regression analysis, but note

that these are not novel data.

Interventions

For the entirety of the five year period, virtually all ambulance pa- tients were roomed immediately and registered at bedside. Ambulatory patients underwent quick registration and nursing triage, in a dedicated triage area or at bedside.

In the last year of physician self-assignment, there was one queue (“rack”) into which the charge nurse placed all charts, in the order that she or he thought they should be evaluated. Physicians then picked up patients at their discretion.

After implementation of rotational patient assignment, a computer algorithm electronically assigned patients to physicians 60 s after regis- tration. In the first iteration of this system, physicians were assigned 3 consecutive patients at the beginning of their shift, were not assigned any patients in the last 90 min of their shift, and were not assigned more than 16 patients per shift. In its current iteration, physicians are assigned 4 consecutive patients at the beginning of their shift, are not assigned any patients in the last 120 min of their shift, and are not assigned more than 18 patients per shift.

In all iterations, after the initial allocation of patients physicians are placed into a strict rotation with other on-duty physicians. Assignments are made based solely on arrival time, without consideration of patient demographics, chief complaint, Emergency Severity Index score, provider patient load, or Acuity of patients recently assigned to the provider.

The typical shift length is 9 h. A more detailed discussion of the first iteration of this system has been reported previously [14].

Methods and measurements

We defined LOS as the time interval (in minutes) between registra- tion and the end of the ED visit. The end of the ED visit was defined as leaving the department or being placed into holding status, which gen- erally occurred if a bed had not been assigned to an admitted patient 4 h after a bed request was placed. We defined APT as the time from regis- tration until the time that a physician electronically “claimed” the pa- tient (an event that occurred after, and independent of, the patient being assigned to a certain physician). LBBS was defined as leaving be- fore evaluation by a physician. None of these working definitions or pa- rameters changed during the five year period.

Total nurse staffing hours was defined as the number of hours

worked by clinical nurse staff (excluding nurse administrators, nursing aides, and technicians) starting with the 7 AM shift, which was the first shift of the day. Nursing workload was defined as daily ED census/total nurse staffing hours. Total physician staffing hours was defined as the number of scheduled physician hours, beginning with the 6 AM shift, which was the first shift of the day. physician workload was defined as daily ED census/total physician staffing hours. ED holding is expressed as number of days in which any ED holding occurred. Effec- tive hospital occupancy was defined as the hospital census divided by the number of staffed beds (expressed as a proportion) at 6 AM.

Data for LOS, APT, LBBS and daily ED census were retrieved from the EMR. Data for nurse staffing were retrieved from handwritten logs and ANSOS One Staff(R) (McKesson, San Francisco, CA). Data for physician staffing were retrieved from Tangier(R) (Peake Software Labs, Sparks, MD). Data for effective hospital occupancy and ED holding were re- trieved from a customized Hospital operations report.

Non-parametric 95% bootstrap confidence intervals were produced to evaluate the differences between the means and medians for LOS and APT and the differences in percentage for LBBS. In each case we per- formed 1000 replications and calculated the 2.5th percentile and the

97.5th percentile to provide the lower and upper limits (respectively) of the 95% confidence intervals. Statistical significance was defined as a bootstrap confidence interval that did not include zero.

We performed a multiple linear regression to determine the change in LOS in the 4 years of rotational patient assignment vs. the last year of phy- sician self-assignment after accounting for confounders. We used a sim- plified model incorporating those variables (ED holding, effective hospital occupancy) that changed in a direction that might independently decrease LOS, and did not incorporate potential confounders that showed little or no change (such as physician and nurse staffing) or changed in a manner that would tend to increase LOS (such as ED volume).

Statistical analysis was performed with SAS version 9.3 (SAS Insti- tute Inc., Cary NC).

1. Results
   1. Patient and site characteristics

There were 140,673 visits over the 5 year period. We eliminated 1,772 visits (1.3%) with APT or LOS values that were missing or illogical (such as APT or LOS of less than zero). This left 138,901 visits eligible for analysis.

Yearly ED volumes and yearly study subjeCT characteristics are pre- sented in Table 1. There was little change in age or gender over the 5 year period, and a slight increase in acuity (manifesting primarily as

Table 1

ED volume and patient characteristics.

Characteristic Physician self-assignment Rotational patient assignment

	Last year		Year 1	Year 2	Year 3	Year 4
Volume Age (years) Median (IQR)	25,651 62 (44, 76)		26,381 62 (44, 76)	26,804 62 (44, 76)	29,613 62 (44, 75)	32,224 61 (43, 75)
Mean (SD)	58.6 (21.2)		58.8 (20.9)	59 (20.8)	58.7 (20.7)	58.1 (21)
Sex (percent female) ESI Score (percentages) Level 1	53.2 0.6		52.4 0.9	53.4 0.9	52.9 0.8	53.8 0.8
Level 2	9.7		13.6	13.9	13.6	14.6
Level 3	72.0		69.9	72.3	74	73.4
Level 4	16.8		14.9	12.3	11.1	10.6
Level 5	0.8		0.8	0.7	0.6	0.6

IQR, interquartile range; SD, standard deviation; ESI, emergency severity index.

an increase in ESI level 2 patients and a decrease in ESI level 4 patients) in all four years of rotational patient assignment vs. the last year of phy- sician self-assignment.

Daily departmental and facility characteristics are presented in Table 2. There was little change in physician or nurse workload over the 5 year period, but notable fluctuations in ED holding and effective hospital occupancy.

Main results

Data for LOS, APT and LBBS are presented in Table 3. LOS, APT and LBBS improved in every year of rotational patient assignment vs. the last year of physician self-assignment.

In a multiple Linear regression analysis incorporating ED holding and hospital occupancy, LOS improvements remained statistically signifi- cant (Table 4).

1. Discussion
   1. Limitations

Our long timeline of data acquisition (5 years) is central to our anal- ysis and conclusions, but raises concerns about confounding by opera- tional drift (improvements or declines in operations over time unrelated to the change being evaluated). We made no operational changes during the study period similar in scope to rotational patient assignment, made no significant changes in laboratory or radiology pro- cessing, introduced no significant changes to the Electronic Medical Re- cord (EMR), and did not implement any significant Hospital flow changes (such as smoothing of elective surgeries). Nonetheless, we can- not exclude the possibility that unmeasured or unintended changes contributed to the lower LOS that we found during follow-up.

We report simple comparisons, when ED LOS may depend on myri- ad factors. We confirmed the directionality of our results for LOS with a

regression analysis incorporating ED holding and effective hospital oc- cupancy, which both changed in a manner that would tend to improve LOS. However, we neither measured nor accounted for other variables that might affect LOS, such as ICU census [17] or daily percentage of ED patients who are admitted [17,18].

Our analysis is of a single site, which limits generalizability. Of note, we show durability of improvements only to a volume of approximately 32,200; the translation of our model to larger volume facilities is un- known. We also do not use advanced practice providers, and cannot comment about how to integrate advanced practice providers into a model such as this. In addition, staffing at our facility is relatively robust, and holding relatively rare. We note, however, that these factors may also make it easier to determine the effect of an intervention such as ours, without concern for confounding by staffing or holding concerns. In previous work, we found statistically significant improvements in LOS for both admitted and discharged patients in an unadjusted primary analysis, but only for discharged patients in a secondary analysis which was adjusted for operational confounders [14]. These findings may be important when facilities with admission rates different than ours

(31%) consider rotational patient assignment as an intervention.

We do not report patient satisfaction data, nor do we report surro- gate data such as complaint rate. An institutional change in vendors dur- ing the study period precluded any meaningful longitudinal analysis of patient satisfaction data. Our methodology for identifying and catalog- ing patient complaints also changed during the period of follow-up, pre- cluding meaningful analysis of that data as well.

We focus on ED operational data without considering non-LOS qual- ity outcomes. Unfortunately, such quality outcomes are difficult to iden- tify and quantify, particularly in studies such as this. There is substantial debate as to whether readily available metrics (such as returns to the ED within 72 h or returns to the ED within 72 h resulting in admission) are clear markers of quality [19-21].

We rely heavily on systems-generated data. We excluded unlikely or illogical data (i.e. LOS of zero or less than zero). Although we accepted

Table 2

Daily facility characteristics.

Characteristic Physician self-assignment Rotational patient assignment

	Last year (n = 366)		Year 1 (n = 365)	Year 2 (n = 365)	Year 3 (n = 365)	Year 4 (n = 366)
Physician workload	1.36 (0.19)		1.30 (0.17)	1.31 (0.18)	1.39 (0.17)	1.37 (0.17)
Nurse workload	0.33 (0.05)		0.32 (0.04)	0.35 (0.05)	0.37 (0.05)	0.35 (0.04)
Effective hospital occupancy	0.90 (0.09)		0.82 (0.10)	0.74 (0.09)	0.69 (0.08)	0.77 (0.15)
ED holding	36 (9.8%)		6 (1.6%)	3 (0.82%)	7 (2.0%)	27 (7.4%)

ED, Emergency Department.

Physician and Nurse workload expressed as mean (standard deviation) of ratio of total patients/total labor hours. Effective hospital occupancy expressed as mean (standard deviation) of percentage of occupied staffed beds at 6 AM. ED holding expressed as days (percent of days) in that year during which ED holding occurred.

Mathematical relationships may appear imperfect due to rounding.

Table 3

Main results.

Metric	Physician self-assignment		Rotational patient assignment
	Last year		First year	Second year	Third year	Fourth year
LOS (minutes)
Median (IQR)	232 (156-322)		208 (137-296)	213 (143-300)	217 (144-303)	211 (139-295)
Difference of medians (95% CI) Percentage change in medians			24 (22-27) -10%	19 (16-21) -8%	15 (12-17) -6%	21 (19-24) -9%
Mean (SD)	253 (1501)		230 (163)	235 (156)	238 (230)	232 (187)
Difference of means (95% CI) Percentage change in means			23 (20-26) -9%	18 (15-20) -7%	15 (12-18) -6%	20 (18-23) -8%
APT (minutes)
Median (IQR)	39 (22-70)		23 (12-41)	20 (9-41)	23 (11-47)	21 (10-40)
Difference of medians (95% CI)			16 (15-17)	19 (18-19)	16 (14-16)	18 (17-18)
Percentage change in medians			- 41%	- 49%	- 41%	- 46%
Mean (SD)	54 (48)		32 (34)	33 (48)	36 (39)	31 (32)
Difference of means (95% CI) Percentage change in means			21 (21-22) -39%	21 (21-22) -39%	18 (17-19) -33%	23 (22-23) -43%
LBBS
Percentage	0.95%	0.51%		0.20%	0.27%	0.30%
Difference of percentages (95% CI) Percentage change in percentages		0.44% (0.30%-0.58%) -46%		0.74% (0.62%-0.87%) -77%	0.67% (0.54%-0.80%) -71%	0.65% (0.52%-0.78%) -68%

LOS, length of stay; APT, arrival to provider time; LBBS, left before being seen. Values for LOS and APT rounded to nearest minute.

All differences are vs. last year of physician self-assignment; statistically significant changes in bold. Mathematical relationships may appear imperfect due to rounding.

the remainder of the data as accurate, we cannot be certain that there were no errors in the final data set, or that errors did not affect one group or one variable more than others.

Main discussion

Improvements in ED operations can yield multiple benefits. A reduc- tion in LOS and the associated improvED throughput may reduce Waiting times, which correlate strongly and negatively with patient sat- isfaction [22,23]. Operational improvements often decrease the rate of patients who leave before being seen, which is of benefit with respect to patient care, financial performance, and risk management. Improved ED throughput may also help to reduce ED crowding, which is associat- ed with Medication errors [24], negative patient outcomes [25,26], poor patient satisfaction [27] and negative perceptions of care [28].

The front-end of the ED is an area over which the ED often has com- plete or near-complete operational control, with few external stake- holders. As such, it is an area where changes designed to reduce LOS can be made relatively easily. Accordingly, numerous front-end innova- tions are described in the literature, including physician in triage, telemedical triage, bedside registration, and the use of a fast-track. These interventions and others are reviewed elsewhere [1,29].

Another front-end intervention is rotational patient assignment. In a system of rotational patient assignment, patients are assigned to physi- cians based on an algorithm, rather than leaving the decision to acquire new patients to physician discretion. Previous experiences with rotation- al patient assignment have found that it is associated with decreased ar- rival to provider time [10,11,14], increased patient satisfaction [11,13],a decrease in complaints [14], and lower overall LOS [12,14,15].

Table 4

decrease in LOS with rotational patient assignment, by year.

Year	Estimated decrease in LOS	95% CI
1	20.2	15.7-24.7
2	13.4	8.5-18.3
3	9.6	4.0-15.1
4	17.0	12.1-21.9

Change in LOS for each year expressed vs. last year of physician self assignment, in minutes. Regression model incorporates ED holding and effective hospital occupancy.

CI: confidence interval.

Deciding how or whether to proceed with ED front-end redesign may be challenging. There is scant literature available to help ED per- sonnel decide which front-end redesign (if any) is appropriate for their ED; the before-and-after nature of most Operational interventions means that interventions are often compared with a “standard practice” rather than directly with one another. Rarely, fortuitous timing may allow for a comparison between two different front end redesigns [30], but such comparisons are uncommon.

A second challenge lies in the lack of data to determine if changes ob- served in the immediate post-implementation period persist over time. The published results of most interventions describe the immediate fol- low-up period, almost always focusing on a period of less than or equal to one year, without reporting long-term results. One exception to this is a report regarding a physician in a triage model, which showed op- erational improvements that persisted in year 3 after implementation [31]. The lack of robust data regarding the durability of operational gains with front-end redesign strategies may be an impediment to front-end change in and of itself. Organizational change can be extraordinarily difficult [32], with multiple barriers at the level of both formulation and implementation [33]. The lack of follow-up Efficacy studies for front-end redesign efforts may give ED management personnel pause when deciding whether to ded-

icate resources or change management efforts to such projects.

We found that the operational gains in LOS, APT and LBBS that we previously reported in the first year of rotational patient assignment persisted in years 2-4, with approximate average yearly decreases in median LOS of 18 min (8% of baseline), median APT of 21 min (54% of baseline), and LBBS of 0.69% (72% of baseline). Importantly, we found that our improvements remained relatively consistent over the three year follow-up period, and that we did not observe a general trend back towards pre-implementation values. There did appear to be a mild decline in improvements in year 3, but these improvements essen- tially recovered in year 4. Improvements observed in year 4 were very similar to (and in some cases slightly better than) improvements ob- served in year 1. We believe that this demonstrates the durability of op- erational gains with rotational patient assignment.

Importantly, these gains have persisted without significant active management. Once this system was accepted by nurses and physicians (a process that was essentially complete within six months), ED man- agement has performed little or no work to keep it in place. Rotational patient assignment has become an integral and unquestioned part of our group’s operating culture.

We note that the institution of rotational patient assignment has abolished all concerns regarding differential productivity. Differential productivity may be a point of contention in any system (such as ours) where physicians are salaried, with no financial incentives for productivity.

Over time, we have made small changes to our underlying algorithm. We changed our initial patient allocation at the beginning of the shift from three patients to four, in order to increase the amount of work physicians perform at the beginning of their shift. We also changed the time of last patient assignment (“cut-off time”) from 90 min to 120 min before the end of a shift. We made this change based on an empiric observation that physicians were staying well past the nominal end of their shift with a cut-off time of 90 min. We have a strong group operating culture that limits sign-outs at the end of the shift, however, and note that groups without such a culture might not require such a long cut-off time. In order to increase productivity, we also changed our maximum number of pa- tients per shift to 18 from 16.

We note that our gains persisted despite a significant increase in daily ED volume. Compared to baseline (the last year of physician self- assignment), ED visit volumes in years 1-4 of rotational patient assign- ment rose by approximately 3%, 4%, 15% and 26%, respectively. During this time, we did not add any additional treatment space. While it is pos- sible that we would have been able to manage this increase in volume without rotational patient assignment, the subjective assessment of ED management was that rotational patient assignment has played a vital role in maintaining throughput, as this system assigned an ac- countable physician to each patient essentially upon patient arrival.

Although we have demonstrated durable success of rotational pa- tient assignment at our facility, we believe that this intervention would not be appropriate for every ED. For example, if a group has strong financial incentives for productivity, a system such as ours may generate resentment from physicians who lose autonomy in determin- ing their own balance of productivity vs. compensation. Autonomy is a known driver of Physician satisfaction [34].

We also note that rotational patient assignment may not be compat- ible with, or add value to, other successful front-end redesign efforts. The provider in triage model, for example, has demonstrated success at many institutions whether the provider is an attending physician [3,35], a resident physician [36], or a physician assistant [37]. Many of the operational gains associated with provider in triage are believed to accrue to the practice of a provider entering appropriate orders early in the encounter during an initial Screening examination.

In rotational patient assignment, physician ownership of patients oc- curs almost immediately upon registration. As such, physicians have a strong incentive to screen their own patients as soon after registration as possible, even if it means doing so in a triage area. Of note, although our facility noticed a benefit with use of a physician in triage, we elected not to use it concurrently with rotational patient assignment. We were concerned that combining the systems would lead to duplicative work, as data at our facility demonstrated that, after screening by a physician in triage, patients sent to the main ED and evaluated by a second physi- cian had additional imaging or testing ordered over 50% of the time [4]. We felt that the initial screening process was best left to the primary physician, as is the workflow in rotational patient assignment.

In conclusion, we found that rotational patient assignment was associat- ed with operational gains that persisted several years after implementation. While such a system is almost assuredly not appropriate for every depart- ment, our findings provide further evidence that rotational patient assign- ment is a valid and durable approach to ED front-end process redesign.

References

1. Wiler JL, Gentle C, Halfpenny JM, et al. Optimizing emergency department front-end operations. Ann Emerg Med 2010;55:142-60 [e1].
2. 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 emer- gency department patients. Qual Manag Health Care 2006;15:163-70.
3. Imperato J, Morris DS, Binder D, et al. Physician in triage improves emergency de- partment Patient throughput. Intern Emerg Med 2012;7:457-62.
4. Traub SJ, Wood JP, Kelley J, et al. Emergency department rapid medical assessment: overall effect and mechanistic considerations. J Emerg Med 2015;48: 620-7.
5. Traub SJ, Butler R, Chang YH, Lipinski C. Emergency department physician telemedical triage. Telemed J E Health 2013;19:841-5.
6. Tolia V, Castillo E, Guss D. EDTITRATE (Emergency Department Telemedicine Initiative to Rapidly Accommodate in Times of Emergency). J Telemed Telecare 2017;23(4): 484-8.
7. Gorelick MH, Yen K, Yun HJ. The effect of in-room registration on emergency depart- ment length of stay. Ann Emerg Med 2005;45:128-33.
8. Saghafian S, Hopp WJ, VO MP, JS D, SL K. Patient Streaming as a mechanism for im- proving responsiveness in emergency departments. Oper Res 2012;60:1080-97.
9. Wiler JL, Ozkaynak M, Bookman K, et al. Implementation of a front-end split-flow model to promote performance in an urban academic emergency department. Jt Comm J Qual Patient Saf 2016;42:271-80.
10. Patel PB, Vinson DR. Team assignment system: expediting emergency department care. Ann Emerg Med 2005;46:499-506.
11. Lau FL, Leung KP. Waiting time in an urban accident and emergency department - a way to improve it. J Accid Emerg Med 1997;14:299-301.
12. Hirshon JM, Kirsch TD, Mysko WK, Kelen GD. Effect of rotational patient assignment on emergency department length of stay. J Emerg Med 1996;14:763-8.
13. DeBehnke D, Decker MC. The effects of a physician-nurse patient care team on pa- tient satisfaction in an academic ED. Am J Emerg Med 2002;20:267-70.
14. Traub SJ, Stewart CF, Didehban R, Bartley AC, Saghafian S, Smith VD, Silvers SM, LeCheminant R, Lipinski CA. Emergency Department Rotational Patient Assignment. Ann Emerg Med 2016;67(2):206-15.
15. Song H, Tucker AL, Murrell KL. The diseconomies of queue pooling: an empirical in- vestigation of emergency department length of stay. Manag Sci 2015;61:3032-53.
16. 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.
17. Lucas R, Farley H, Twanmoh J, et al. Emergency department patient flow: the influ- ence of hospital census variables on emergency department length of stay. Acad Emerg Med 2009;16:597-602.
18. Wiler JL, Handel DA, Ginde AA, et al. Predictors of patient length of stay in 9 emer- gency departments. Am J Emerg Med 2012;30:1860-4.
19. Pham JC, Kirsch TD, Hill PM, DeRuggerio K, Hoffmann B. Seventy-two-hour returns may not be a good indicator of safety in the emergency department: a national study. Acad Emerg Med 2011;18:390-7.
20. Abualenain J, Frohna WJ, Smith M, et al. The prevalence of quality issues and adverse outcomes among 72-hour return admissions in the emergency department. J Emerg Med 2013;45:281-8.
21. Klasco RS, Wolfe RE, Wong M, et al. Assessing the rates of error and adverse events in the ED. Am J Emerg Med 2015;33:1786-9.
22. Nichol JR, Fu R, French K, Momberger J, Handel DA. Association between patient and emergency department operational characteristics and patient satisfaction scores in a pediatric population. Pediatr Emerg Care 2016;32:139-41.
23. Handel DA, Fu R, Vu E, et al. Association of emergency department and hospital char- acteristics with elopements and length of stay. J Emerg Med 2014;46:839-46.
24. Kulstad EB, Sikka R, Sweis RT, Kelley KM, Rzechula KH. ED overcrowding is associated with an increased frequency of medication errors. Am J Emerg Med 2010; 28:304-9.
25. Jo S, Kim K, Lee JH, et al. Emergency department crowding is associated with 28-day mortality in community-acquired pneumonia patients. J Infect 2012;64:268-75.
26. Pines JM, Pollack Jr CV, Diercks DB, Chang AM, Shofer FS, Hollander JE. The associa- tion between emergency department crowding and adverse cardiovascular out- comes in patients with chest pain. Acad Emerg Med 2009;16:617-25.
27. Pines JM, Iyer S, Disbot M, Hollander JE, Shofer FS, Datner EM. The effect of emergen- cy department crowding on patient satisfaction for admitted patients. Acad Emerg Med 2008;15:825-31.
28. Pines JM, Garson C, Baxt WG, Rhodes KV, Shofer FS, Hollander JE. ED crowding is asso- ciated with variable perceptions of care compromise. Acad Emerg Med 2007;14: 1176-81.
29. Saghafian GA S, Traub S. Operations research/managment contributions to emergen- cy department patient flow optimization: review and research prospects. IIE Trans Healthc Syst Eng 2015;5:101-23.
30. Traub SJ, Bartley AC, Smith VD, Didehban R, Lipinski CA, Saghafian S. Physician in tri- age versus rotational patient assignment. J Emerg Med 2016;50:784-90.
31. Rogg JG, White BA, Biddinger PD, Chang Y, Brown DF. A long-term analysis of physician triage screening in the emergency department. Acad Emerg Med 2013;20:374-80.
32. Kotter JP. Leading change: Why transformation efforts fail; 1995.
33. Pardo del Val M, Martinez Fuentes C. Resistance to change: a literature review and empirical study. Manag Decis 2003;41:148-55.
34. Schrijver I, Brady KJ, Trockel M. An exploration of key issues and potential solutions that impact physician wellbeing and professional fulfillment at an academic center. PeerJ 2016;4:e1783.
35. White BA, Brown DF, Sinclair J, et al. Supplemented triage and rapid treatment (START) improves performance measures in the emergency department. J Emerg Med 2012;42:322-8.
36. Svirsky I, Stoneking LR, Grall K, Berkman M, Stolz U, Shirazi F. Resident-initiated ad- vanced triage effect on emergency department patient flow. J Emerg Med 2013;45: 746-51.
37. Nestler DM, Fratzke AR, Church CJ, et al. Effect of a physician assistant as triage liaison provider on patient throughput in an academic emergency department. Acad Emerg Med 2012;19:1235-41.