a b s t r a c t

Study objectives: Facing increased utilization and subsequent capacity and budget constraints, ED’s must better understand bottlenecks and their effect on process flow to improve process efficiency. The primary objective of this study was to identify bottlenecks in obtaining a head CT and investigate patient waiting time based on those bottlenecks.

Methods: This observational study included all patients undergoing a head CT between July 1, 2013 and June 30, 2014 at a large, Urban academic ED with over 100,000 visits per year.

The primary study outcome was total cycle time, defined as the elapsed time between patient arrival and head CT preliminary report, divided into four components of workflow.

Results: 8312 patients who had a head CT were included in this study. The median cycle time from patient arrival to head CT preliminary report was 3 h and 13 min with 39 min of waiting time resulting from bottlenecks. In the 4-stEP model (time from patient arrival to head CT order, time from head CT order to head CT scheduled, time from head CT scheduled to head CT completed, and time from head CT completed to head CT preliminary report), each process was the bottleneck 30%, b 1%, 27%, and 42% of the time, respectively.

Conclusion: Demand capacity mismatch in head CT scanning has a significant impact on patient Waiting times. This study suggests opportunities to improve wait times through future research to understand the causes of de- lays in CT ordering, CT completion and timeliness of Radiology reports.

(C) 2017

Introduction

Background

Facing increased utilization and subsequent capacity and budget constraints, many emergency departments (EDs) are actively seeking new ways to streamline processes. However, this remains challenging, as most ED processes require coordinating a number of different func- tions [1]. In addition, while some ED processes can be completed in par- allel, many must be completed in series (i.e., the prior step must be completed before moving to the next step). According to the “theory of constraints”, a process is only as fast as its slowest step, known as the bottleneck. Better understanding of bottlenecks, and their effect

? Presented at American College of Emergency Physician’s research forum on October 16, 2016.

* Corresponding author at: 6431 Fannin Street, JJL 434, Houston, TX 77030, United States.

E-mail address: [email protected] (J.G. Rogg).

¹ Dr. Rogg’s present address is at University of Texas Health Science Center at Houston.

on process flow, can allow emergency physicians to improve Patient throughput and decrease wait times [2,3].

Owing to large variations in patient volume and acuity, and subse- quent resource demand and utilization, the ED is a unique place to ex- amine bottlenecks. Several studies have investigated methods of identifying bottlenecks in the ED and allocating resources to minimize wait times and optimize resource use [3,4]. Importantly, mean and me- dian wait times are not the only metrics with significant effects on the system, and for certain ED testing processes, outlier wait times have been shown to have a substantial impact on patient length of stay (LOS) [5].

Streamlining patient flow offers several potential benefits. First, faster testing and treatment may lead to the ability to treat patients sooner and provide care of higher quality. Second, treating patients more quickly allows for more-rapid Disposition decisions, potentially decreasing ED LOS. Third, for a given level of patient demand, shorter LOS correlates with smaller effective ED census, subsequently decreas- ing crowding. Such reductions in crowding allow limited resources such as bed space, nursing time, and physician time to be allocated to patients in need. This has been shown to improve patient safety and

http://dx.doi.org/10.1016/j.ajem.2017.04.059

0735-6757/(C) 2017

J.G. Rogg et al. / American Journal of Emergency Medicine 35 (2017) 1510-1513 1511

improve timely medication administration [6,7]. Fourth, wait times can also negatively impact patient experience [7,8]. Finally, extendED boarding in the ED negatively impacts hospital financial performance [8]. Improving patient flow and throughput thus has the potential to provide benefits to patients, providers, and hospitals alike.

Importance

Matching resource allocation more closely to resource demand re- duces throughput time. In a system with significant variability, like the ED, it is likely that the bottleneck is not the same for every patient, and the existence of shifting bottlenecks has the potential to play a crit- ical role in systems engineering solutions. However, the concept of shifting bottlenecks has not been well studied in the ED.

Computed Tomography (CT) scans of the head offer an instructive case study in understanding ED bottlenecks and variation. Head trauma and headache result in over 2.1 million ED visits annually; 14% of these visits include neuroimaging [9]. head CTs are ordered for as many as1 in 10 patients presenting to the ED [10]. In addition, the process of performing a head CT requires numerous coordinated steps between multiple role groups, including emergency physicians, radiologists, nurses, transport, and radiology technicians. Adding to the complexity, multiple patient, staff, and systEMS factors can impact the turnaround time of this process.

Objectives

The primary objective of this study was to identify bottlenecks in obtaining an ED head CT and investigate waiting time based on those bottlenecks.

Materials and methods

Study design, participants and setting

This retrospective observational study, approved by the Institutional Review Board, included all patients undergoing head CT between July 1, 2013 and June 30, 2014. The study setting was a large, urban academic ED with over 100,000 visits per year. This facility has 2 dedicated ED CT scanners that operate 24 h per day and primarily serve the ED but are also used for STAT inpatient CT scans.

Data collection and processing

Radiology logs were used to determine which ED patients had a head CT scan performed. This data was then cross referenced with the ED information system to collect the following time data: patient arrival time, attending assignment time, CT order time, CT completion time, CT preliminary report time (generated by radiologist, often a trainee), CT final report time, and disposition decision (e.g., discharge from ED, ad- mission to hospital) time. We also collected demographic data and the discharge location of the patients.

Outcome measures

Primary data analysis

Patients with deficient or missing data (i.e., those who had a head CT ordered, scheduled, or completed after dismissal and those with any data that had negative intervals recorded) were excluded. Patients who had long intervals that were unlikely to be accurate (i.e., N 6 h be- tween arrival and order, N 3 h between ordered and scheduled, N 6 h be- tween scheduled and complete, or N 8 h between completed and preliminary report) were also removed.

This study compares cycle times, the time to complete a defined pro- cess. The overall cycle time of patient arrival to head CT preliminary re- port time is made up of four components: the time from patient arrival to head CT order, the time from head CT order to head CT scheduled, the time from head CT scheduled to head CT completed, and the time from head CT completed to head CT preliminary report. The sum of the cycle times for a patient moving through each process at the median should be equal to the median cycle time of the overall process. However, in re- ality patients may move through each of the component processes faster or slower. This lack of coordination, such as moving though one process faster relative to that processes median time and another slower relative to that processes median time creates additional waiting time from capacity mismatch. This study compares the calculated pa- tient arrival to head CT preliminary report time (IE the sum of the me- dians of each component process cycle time) to the actual patient arrival to head CT preliminary time (the median time it actually took a patient to go through the entire process) to determine the additional waiting time. This entire process was calculated for 25th percentile and 75th percentile as well. We also examined the distribution of the bottleneck, which was defined as the longest interval among the four processes. All analyses were conducted using SAS version 9.4 (SAS Insti- tute, Cary NC) by a PhD biostatistician.

Results

The ED summary statistics during the study time frame are depicted in Table 1. A total of 8749 patient encounters had a head CT scan ordered during the study period. Five percent (437/8749) were excluded be- cause of missing/deficient data (1.4%) or unlikely long intervals (3.6%). This left 8312 patients included in the study. The average patient age was 59.6 years (SD 21.0).

Table 2 describes the median, 25th percentile, and 75th percentile times for the four workflow components of the head CT process. The ob- served median time from patient arrival until a preliminary head CT

Table 1

Summary statistics.

Patient population 8312

Age (mean, std) 59.6 (21)

Male sex (#, %) 4475 (53.8%)

Race (#, %)

Asian 276, 3.3%

African American 620, 7.5%

Hispanic 766, 9.2%

White 6367, 76.6%

Other 85, 1.0%

Unknown 198, 2.4%

The primary study outcome was total cycle time, defined as the elapsed time between patient arrival and head CT preliminary report, divided into four components of workflow: the time from patient arrival to head CT order, the time from head CT order to head CT scheduled, the time from head CT scheduled to head CT completed, and the time from head CT completed to head CT preliminary report. The disposition time for each patient was also recorded to understand how many patients had a disposition before the head CT preliminary report was available.

Disposition (#, %)

Length of stay (hours)

Discharge 3301, 39.7%

Inpatient 3484, 41.9%

ED observation 1448, 17.4%

Other 79, 1.0%

Median 6.4

Quartile 1 4.6

Quartile 3 9

1512 J.G. Rogg et al. / American Journal of Emergency Medicine 35 (2017) 1510-1513

Table 2

Wait times.

	(In hours:minutes)			% of time each interval is the bottleneck
	25th % (95% CI)	Median (95% CI)	75th % (95% CI)
aED patient arrival to head CT order	0:22 (0:21-0:22)	0:42 (0:40-0:43)	1:26 (1:23-1:28)	30%
aHead CT order to head CT scheduled	0:10 (0:10-0:10)	0:12 (0:12-0:12)	0:13 (0:13-0:13)	b1%
aHead CT scheduled to head CT complete	0:22 (0:21-0:23)	0:43 (0:42-0:45)	1:18 (1:17-1:20)	27%
aHead CT completed to preliminary report	0:36 (0:36-0:37)	0:57 (0:56-0:58)	1:35 (1:33-1:37)	42%
Actual ED arrival to head CT preliminary report	2:17 (2:15-2:18)	3:13 (3:10-3:16)	4:26 (4:22-4:29)
Calculated ED arrival to head CT preliminary report Difference	1:30 (1:29-1:31) 0:47 (0:44-0:49)	2:34 (2:32-2:37) 0:39 (0:35-0:42)	4:32 (4:28-4-:35) -0:06 (-0:11-0:00)

^a Calculated ED arrival to Head CT preliminary report is the arithmetic sum of the * intervals.

report was 3:13 (2:17-4:26). Three of the four steps each took roughly 45 min to an hour. 26.9% of patients had a disposition (discharged from the ED or had an inpatient or observation bed requested) before the pre- liminary head CT report was published (data not shown). The actual and the calculated patient arrival to head CT preliminary times differed by N 30 min from the actual arrival to preliminary times for the 25th per- centile and the median. For a patient who completed the head CT pro- cess in the median time for each step, the process should take 2:34, however the actual time was 3:13 (a difference of 39 min). For the 25th percentile, the calculated time was 1:30 compared with an actual time of 2:17 (a difference of 47 min). In contrast, the 75th percentile had a calculated time of 4:26 vs. 4:32 actual time. In the 4-step model (time from patient arrival to head CT order, time from head CT order to head CT scheduled, time from head CT scheduled to head CT complet- ed, and time from head CT completed to head CT preliminary report), each process was the bottleneck in 30%, b 1%, 27%, 42%, of the patients respectively.

Discussion

Findings

The purpose of this brief report was to use the process of performing a head CT in the ED to illustrate the concept of bottlenecks, and to dem- onstrate its importance. As the extant literature would suggest, Head CTs are frequently performed in the ED, (in this study, 8749 were per- formed during the study period) and visits associated with head CT accounted for about 8% of total ED volume. The median head CT cycle time was 3 h and 13 min from arrival until a preliminary report was available. However, over 26% of Disposition decisions were made before the CT preliminary report was done. This could be for several reasons. For these patients, the head CT was likely not important to make a dis- position decision. This could be because the patient was already going to be admitted for other reasons, such as other known traumatic injuries (e.g., femur fracture). It could also be a low-yield test to exclude unlikely diagnoses, for example a septic patient with altered mental status may undergo a head CT although it is unlikely to helpful for most patients in that population. It is also possible that the ED physician reviewed the head CT and felt sufficiently comfortable to determine a disposition before receiving the written preliminary report. However, about 74% of patients did not have a disposition before the preliminary head CT re- sults. The majority of patients for whom head CT is important for the disposition decision would benefit from faster arrival-preliminary re- port times.

Perhaps most interesting was the large difference in calculated vs. actual patient arrival-head CT preliminary report cycle times for the 25th percentile and median. The bottleneck in this model is not the same for all patients. It shifts between the time it takes emergency phy- sicians to order head CTs, the time it takes to move a patient to the CT scanner, and the time it takes radiologists to interpret a preliminary

head CT report. At the median, demand-capacity mismatch adds 39 min, equal to a 20% increase in wait time. Therefore, in this ED study population, with an annual head CT volume of 8312 patients,

5402 h (8312 patients * 0:39 min) or 225 days of patient waiting time are lost each year because of bottlenecks.

Limitations

This was a single center study conducted at a large urban academic hospital. The findings may not generalize to other hospitals with differ- ent volumes, patient characteristics, or ED imaging processes. This hos- pital also has a dedicated ED observation unit, so the high rate of observation use may impact ED physician time to disposition. Addition- ally, the use of physician trainees may impact how quickly emergency physicians order Head CT scans and how quickly they are reported by radiology.

Conclusion

In this ED, the median cycle time from patient arrival to head CT pre- liminary report was 3 h and 13 min. However, a median of 39 min of ad- ditional waiting time resulted from bottlenecks and suboptimal resource allocation in the simple 4-step process flow, accounting for over 225 excess days of patient waiting. Additionally, the bottleneck in this model shifts between different steps. Demand capacity mismatch in head CT scanning has a significant impact on patient waiting times. This study suggests opportunities to improve wait times through future research to understand the causes of delays in CT ordering, CT comple- tion and timeliness of radiology reports. Similar analysis could be used to explore bottlenecks in other ED functions such as laboratory tests and other imaging studies.

Funding and support

No funding or support was provided for this research project.

References

1. Kastner C, Tagg a. Improving the effectiveness of the emergency management of renal colic in a district general hospital: a completed audit cycle. Emerg Med J 2003;20:449-50.
2. Asaro PV, Lewis LM, Boxerman SB. The impact of input and output factors on emer- gency department throughput. Acad Emerg Med Mar 2007;14(3):235-42.
3. Rotstein Z, Wilf-Miron R, Lavi B, Seidman DS, Shahaf P, Shahar A, et al. Management by constraints: considering patient volume when adding medical staff to the emer- gency department. Isr Med Assoc J 2002;4(3):170-3.
4. Brenner S, Zeng Z, Liu Y, Wang J, Li J, Howard PK. Modeling and analysis of the emer- gency department at university of Kentucky Chandler Hospital using simulations. J Emerg Nurs 2010;36(4):303-10.
5. Holland LL, Smith LL, Blick KE. Reducing laboratory turnaround time outliers can re- duce emergency department patient length of stay: an 11-hospital study. Am J Clin Pathol 2005;124(5):672-4.
6. Wiler JL, Gentle C, Halfpenny JM, Heins A, Mehrotra A, Mikhail MG, et al. Optimizing emergency department front-end operations. Ann Emerg Med Feb 2010;55(2) Elsevier Inc. [142-160.e1].

   J.G. Rogg et al. / American Journal of Emergency Medicine 35 (2017) 1510-1513 1513

   Pines JM, Hollander JE, Localio aR, Metlay JP. The association between emergency de- partment crowding and hospital performance on antibiotic timing for pneumonia and percutaneous intervention for myocardial infarction. Acad Emerg Med Aug 2006;13(8):873-8.

7. Pines JM, Batt RJ, Hilton J a, Terwiesch C. The financial consequences of lost demand and reducing boarding in hospital emergency departments. Ann Emerg Med 2011; 58(4):331-40 Elsevier Inc.
8. Goldstein JN, Camargo C a, Pelletier aJ, Edlow J a. Headache in United States emer- gency departments: demographics, work-up and frequency of pathological diagno- ses. Cephalalgia 2006;26:684-90.
9. Lee. Computed tomography use in the adult emergency department of an Academic Urban Hospital from 2001 to 2007 (vol. 56, pg 591, 2010). Ann Emerg Med 2011;

   57(3):256 Elsevier Inc.