Article

Factors associated with ED length of stay during a mass casualty incident

a b s t r a c t

Background: The aim of this study was to examine the factors associated with emergency department (ED) length of stay (LOS) using the patient registry data from a medical burns center during a Burn injury mass casualty in- cident (MCI) after a dust explosion in New Taipei City, Taiwan.

Methods: This was a retrospective cohort study conducted at an urban, tertiary care teaching hospital during an MCI event that occurred on June 27, 2015. A celebratory party was held at the Formosa Fun Water Park in New Taipei City, Taiwan. At 20:32, the was an explosion caused by an overheated spotlight accidentally igniting col- ored Cornstarch powder that had been sprayed on the stage. Factors associated with ED LOS were compared. Results: In total, 48 burn injury patients were enrolled for study analysis. The median total body surface area of second- to third-degree burns was 35.0% (interquartile range [IQR], 15.8%-55.0%). The median ED LOS was

121.5 minutes (IQR, 38.3-209.8 minutes). The output time interval accounted for the longest interval with a me- dian time of 56.0 minutes (IQR, 15.3-117.3 minutes). In multivariate analysis of the variables, triage level (level III; hazard ratio, 0.06; 95% confidence interval, 0.01-0.52) and output time (hazard ratio, 0.97; 95% confidence in- terval, 0.96-0.98) were significant influential factors. Conclusions: The triage level and output time intervals were significantly associated with ED LOS in a burn-related MCI. Time effectiveness analyses, using a patient flow model, might serve as an important indicator during a hos- pital MCI response.

(C) 2016

Introduction

A mass casualty incident (MCI) is defined an event where the avail- able Health care resources are unable to meet the demand of the inci- dent [1]. Both the number of patients and the severity of their injuries may exceed the capability of the medical facility and staff. Patients of an MCI are sent to an assigned hospital for management and, if neces- sary, admitted via the emergency department (ED). A mass casualty pa- tient load adversely affects the quality and decreases the level of care in the ED [2,3]. Such an incident, in which the demand for emergency ser- vices exceeds the ability to provide quality care within a reasonable time, shares many similarities with ED overcrowding [4]. Although ED overcrowding is a long-term, more complicated issue, both MCI and

? Grant: This study received no finical support or grant.

?? Conflict of interest: Nothing to declare.

* Corresponding author at: Department of Emergency Medicine, Chang Gung Memorial Hospital and Chang Gung University College of Medicine, No. 5 Fushing St, Gueishan Shiang, Taoyuan, Taiwan. Tel.: +886 3 3281 200×2505; fax: +886 3 3287 715.

E-mail addresses: [email protected], [email protected] (Y.-M. Weng).

ED overcrowding have patient loads that exceed the ability to provide adequate quality care [5].

Although ED waiting time and length of stay (LOS) are used as indi- cators of ED overcrowding [6-8], no study examines the use of these in- dicators during a hospital MCI response. Previous studies have focused on ED load, preparedness of the hospital response, and management during a hospital MCI response [9-12]. To monitor the quality of care during a hospital MCI response, analysis of time effectiveness using an ED patient flow model is important. The input-throughput-output con- ceptual model has been widely used for time effectiveness analysis in ED overcrowding scenarios [13-14]. Therefore, the input-throughput- output conceptual model may be a feasible method to evaluate and monitor time effectiveness during an MCI response in an ED. Patient flow models are used to identify which factors influence time effective- ness in the ED. This knowledge helps the system to allocate the staff and resources appropriately during an MCI. However, hospital disaster pre- paredness programs should be adapted based on evidence obtained from actual cases.

The aim of this study was to examine the factors associated with ED LOS using the patient registry data from a burns center during a burn in- jury MCI after a dust explosion in New Taipei City, Taiwan.

http://dx.doi.org/10.1016/j.ajem.2016.04.056

0735-6757/(C) 2016

Methods

Study design and setting

This was a retrospective cohort study conducted at an urban, tertiary care teaching hospital–Chang-Gung Memorial Hospital (CGMH)–in Linkou, Taiwan, during an MCI event that occurred on June 27, 2015. The study was approved by the CGMH Institutional Review Board and was exempted from requiring informed consent and full committee re- view. On June 27, 2015, a celebratory party was held at the Formosa Fun Water Park in New Taipei City, Taiwan. At 20:32, an explosion–caused by an overheated spotlight accidentally igniting colored cornstarch powder that had been sprayed on the stage–injured 499 people. Such an incident, one of the largest scale burn injury mass casualties, has rarely been reported on previously [9-11,15]. The Ministry of Health and Welfare activated the regional Emergency Medical Operation Cen- ters and informed designated hospitals to prepare for this burn injury MCI [16]. All injured patients were transported alive from the scene to 48 hospitals; 106, 96, 53, and 228 were admitted to the burn intensive care unit (ICU), surgical ICU, beds of burn ward, and surgical ward, at the first 48 hours after incident, respectively. Sixteen patients were discharged from ED after initial management. The mean burn size of hos- pitalized patients was 43.7%. Twenty-one and 248 patients had a burn total body surface area (TBSA) of greater than 40% and greater than 80%, respectively. There were total 12 fatalities and 107 hospitalized pa- tients (19 in an ICU) 3 months postincident (October 1, 2015) [16,17].

Chang-Gung Memorial Hospital, Linkou, is a large, urban, tertiary care teaching hospital with an estimated ED volume of 227 000 visits per annum. It is also the largest trauma and burns center and the prima- ry trauma referral center for Northern Taiwan. There are a total of 3700 beds, 1000 attending staffs, and 4738 other medical staff providing medical care in the hospital. The hospital has 19 burn ICU beds and 9 ward beds in an isolated positive pressure control area. The estimated surge capacity of the ED of the study hospital is 60 patients triaged as level I or II.

At around 21:00 on June 27, 2015, the hospital was notified of the burn injury MCI. After receiving the notification, the departmental duty chiefs and administrative staff started to prepare and to recall medical personnel. The first injured patient arrived at the ED at 21:21. A total of 49 patients were sent to the study hospital over the following 4 hours. One patient was transferred from the other hospital. The hospi- tal started an MCI response with an announcement of “Code 333” at 22:00. According to the study hospital’s protocol, 1 staff member from each department or ward should report to the ED once an MCI response has been activated. More than 30 physicians, 120 nurses, and other medical personnel reported to the ED at 22:10. A hospital incident com- mand system was set up at 22:37. Injured patients were triaged into 3 categories by experienced nurses and physicians: level I, red, emergen- cy; level II, yellow, urgent; level III, green, less urgent. These levels were based on a patient’s level of consciousness, respiration, involved body part, and TBSA of second- to third-degree burns at arrival. Emergency airway control, Central venous line access, and wound care were provid- ed by ED personnel. Patients were then either admitted to the burn ICU/ surgical ICU/ward or discharged, according to physician assessment. During the MCI response, a reserved ICU, which served as an adapted Burn unit, was used in addition to the burns ICU. The hospital MCI re- sponse ended, and the Incident Command System was dissolved at 01:04 on June 28, 2015.

Patient selection

All patients sent to CGMH from the burn injury MCI at Formosa Fun Water Park dust explosion on June 27, 2015, were enrolled. The patient list was abstracted via the hospital computer registry system. Patients who were transferred from other hospital post-initial evaluation and management with clear transfer indication and contact beforehand.

These patients might have different admission process and time matrix. Therefore, these patients were excluded from study analysis. Patients who lost to follow-up or had incomplete data were excluded.

Study protocol and data collected

The authors reviewed the medical records and collected data using a Standardized reporting template with clear definitions and codes. Pa- tient numbers and the time of their arrival were recorded. Background demographics of patients included age and sex (registered before acti- vation of hospital MCI response). Data on triage level, TBSA of second- to third-degree burns after evaluated by primary care physicians at ED, ED management with intubation and/or central venous access, ED disposition (admitted to ICU/ward, discharged, transferred to another hospital), and LOS in the ED (recorded in minutes) were collected. The patient flow in the ED was demonstrated using an input-throughput- output conceptual model. The input time interval was defined as the du- ration between ED registration and initial physician assessment. The throughput time interval was defined as the time from when the physi- cian gave management orders–including analgesic use, wound care, imaging studies, intubation, and peripheral or central venous access es- tablishment with fluid administration–until these orders were com- pleted. The output time interval was defined as the time from when a Disposition decision was made until the time the patient left the ED. Pa- tient accumulation was defined as the number of MCI patient registra- tions by arrival time at the ED; the number of ED patient registrations 30 minutes before and after patient arrival was counted. The primary outcome was ED LOS.

Primary data analysis

Data were analyzed using SPSS software (version 13.0 for Windows; SPSS, Chicago, IL). In the descriptive analysis, categorical variables were presented as numbers and percentages and were compared using the ?2 or Fisher exact test, as appropriate. Continuous variables were present- ed as the median and interquartile range (IQR). The Mann-Whitney U test was used for nonnormally distributed continuous variables. Factors associated with ED LOS were compared using univariate and multivari- ate Cox proportional hazards regression models. The proportional haz- ards assumption was tested using graphical methods; no significant violation was noted. Results are presented as Hazard ratios (HRs) with 95% confidence intervals (CIs). An HR of less than 1 represents a lesser probability of leaving the ED at a given time, indicating a longer ED LOS, and vice versa. P b .05 was considered statistically significant.

Results

In total, 48 burn injury patients were enrolled for study analysis. Fig. 1 shows the number of patients registered in each triage level in a sequence of 15-minute intervals in the ED. Patient registrations oc- curred continuously for 165 minutes. There was a peak of 20 patient registrations from 45 to 90 minutes after the first patient’s arrival. The characteristics of the enrolled patients are presented in Table 1. The me- dian TBSA of second- to third-degree burns was 35.0% (IQR, 15.8%- 55.0%). The number of patients stratified by different triage levels was 14, 19, and 15 for level I, II, and III, respectively. Seventeen patients (35.4%) required endotracheal intubation, 22 (45.8%) required central venous access, and 32 (66.7%) were admitted to the ICU. The median ED LOS was 121.5 minutes (IQR, 38.3-209.8 minutes). The output time interval accounted for the longest interval with a median time of 56.0 minutes (IQR, 15.3-117.3 minutes), followed by the throughput (medi- an, 35.0; IQR, 18.3-54.8 minutes) and input (median, 10.0; IQR, 6.0-14.8 minutes) time intervals (Table 2). There was a trend of decreasing input and throughput time intervals once the hospital MCI response had been fully activated and medical staff had reported for the response. In

Fig. 1. Emergency department patient load, over time.

contrast, the output time interval decreased but subsequently rebounded (Fig. 2).

Enrolled patients who registered before hospital MCI response acti- vation were not significantly associated with ED LOS (HR, 1.79; 95% CI, 0.85-3.74; Table 3). In the analysis of correlates associated with ED LOS, triage level (level II vs level I; HR, 0.29; 95% CI, 0.14-0.62; level III vs level I; HR, 0.25; 95% CI, 0.09-0.68), ED endotracheal intubation (HR, 2.68; 95% CI, 1.38-5.19), input time interval (HR, 0.95; 95% CI,

0.91-0.99), throughput time interval (HR, 0.98; 95% CI, 0.97-1.00), and output time interval (HR, 0.98; 95% CI, 0.97-0.99) were significant influ- ential factors. Patients’ sex (HR, 0.64; 95% CI, 0.33-1.22), age (HR, 1.00;

95% CI, 0.96-1.04), the patient registration load (HR, 0.96; 95% CI,

0.90-1.04), and patient accumulation (HR, 1.00; 95% CI, 0.98-1.03) were not significant influential factors. In multivariate analysis of the above significant influential factors, only triage level (level III vs level I; HR, 0.06; 95% CI, 0.01-0.52) and output time (HR, 0.97; 95% CI, 0.96-

0.98) were significant influential factors.

Discussion

The time effectiveness of patient flow might be comprehensive during a MCI response

Disaster preparedness, with an effective medical response for MCIs, has become an important issue for hospitals and emergency systems [3,18-19]. Several reports have evaluated the responses to large-scale

Table 1

Patient characteristics

Patients (n = 48)

Age in years, median (IQR) 22.0 (20.0-27.0)

Male, n (%) 20 (41.7)

MCIs and patient outcomes [20-24]. Early mitigation of ED crowding from an MCI might be important for quality of care, especially at individ- ual patient level [3]. No previous study evaluates time effectiveness using a patient flow model in the ED during a hospital MCI response. The time effectiveness of different stages of patient flow, which is wide- ly used in ED overcrowding analyses, might be more comprehensive than casualty load or scale of the MCI response. This is the first report to analyze the factors affecting ED LOS in a burn injury MCI response.

Output time is significantly associated with ED LOS in a burn-related MCI

From a time effectiveness perspective, this study identified that a hospital MCI response might initially improve the quality of care of pa- tients during an actual MCI by shortening the time from arrival in the ED to assessment and management. The patient load was not significantly associated with the ED LOS once the MCI response had been activated. Although the input and throughput time intervals initially reduced, the output time interval increased during the second half of the inci- dent. There was a delay in admitting patients to the wards, which in- creased their stay in the ED thereby prolonging output time. Instead of activation of the MCI response or the patient load, the output time inter- val and the severity of burn injuries were the major influential factors associated with the ED LOS. According to these study results, the surge capacity of a medical facility does not depend solely on the scale of the MCI response, but on the whole system of the care. The hospital MCI re- sponse should be activated early, with a dynamic distribution of the re- inforced personnel and resources. The indicators of dynamic plan of the distribution of personnel/resources should be considered, such as out- put time interval, or the number of patients boarding in the ED whilst awaiting admission. Further research using Simulated models and exer- cises to examine the optimized indicators is warranted.

Patients registered before hospital MCI response activation, n (%)

Triage, n (%)

Level I, emergent

14 (29.2)

Table 2

Time intervalsa, b, c

Level II, urgent

19 (39.6)

Level III, less urgent

15 (31.3)

%TBSA of second to third degree burns, median (IQR)

35.0 (15.8-55.0)

ED LOS (min), median (IQR)

121.5 (38.3-209.8)

ED management, n (%)

Inputa (min), median (IQR)

10.0 (6.0-14.8)

Endotracheal intubation

17 (35.4)

Throughputb (min), median (IQR)

35.0 (18.3-54.8)

Central venous access

22 (45.8)

Outputc (min), median (IQR)

56.0 (15.3-117.3)

12 (25%)

Disposition, n (%)

ICU 32 (66.7)

Ward 7 (14.6)

Discharge 8 (16.7)

Transferred to another hospital 1 (2.1)

a Input: time interval between ED registrations to initial physician assessment.

b Throughput: time interval between the physician orders to complete the management.

c Output: time interval between decisions of disposition until the patient was discharged from ED.

Fig. 2. Time intervals of input, throughput, and output for patients admitted to the ICU.

Explanations of study results

Several aspects of this study should be discussed. First, the output time interval was significantly associated with increased ED LOS after adjustment for associated variables. The study results suggest that the most important factor affecting ED LOS during such an incident is waiting for admission once a disposition decision has been made. The reason for the increasing output time interval might due to an inade- quate number of beds for multiple burn injury patients. Furthermore, medical staff and physicians had gathered in the ED from each depart- ment and ward, leaving fewer personnel to provide medical care to pa- tients admitted to the ICU or wards. Previous studies addressed adapting hospital bed usage to increase surge capacity by creating a spe- cial burns unit to resolve the problem of delayed output processing [9-12]. Although this is a possible solution to decrease the output time interval, available staff might be more crucial than actual space. In our case, there was a reserved ICU, which served as an adapted burns unit in addition to the normal burns ICU during the MCI response. However, the output time interval remained prolonged even after open- ing this reserved unit because of limited staff. No previous study could provide information about the time to create a special burns unit or when to shift personnel and resources from the ED to the ICU or wards during an MCI response. Based on this study’s results, identifying a cutoff value within the output time interval might be useful to deter- mine the timing of reallocation of personnel/resources.

Second, the input and throughput factors were not statistically sig- nificant for ED LOS after adjustment for associated variables. The

hospital MCI response was activated early with an adequate number of physicians and nurses reporting for the response. Despite the MCI at the ED, patients were seen by medical staff and treated timely, with- out significant delay. Therefore, the effects of input and throughput time intervals on ED LOS were subtracted. The hospital MCI response might improve the quality of care via shortening the time interval of input and throughput.

Third, Hirshberg et al [3] created a computer model that defines the quantitative relationship between an increasing casualty load and grad- ual degradation of the level of trauma care based on MCIs in 2005. The study concluded that the surge capacity depended on the rate of patient arrival, rather than the availability of beds. However, the computer model used a fixed number of medical staff and beds. In addition, pa- tient stasis in the ED–resulting in a prolonged output process–was not considered, which further exacerbates the condition. Input un- doubtedly influences the degradation of level of care and determines the surge capacity in such a fixed computer model. From another aspect, previous studies used casualty load and other variables, such as access to triage, quality of trauma, resuscitation bay, level of experience of the decision maker, composition and qualifications of the trauma team, and immediate access to the computed tomographic scanner and operating room, as evaluating factors for hospital management in an MCI [3,25]. A complex scale to measure the quality of care during an MCI response is not practical during an actual incident. Emergency department LOS and input-throughput-output models can be continu- ously monitored during a hospital MCI response. This could aid the decision-making processes around distribution of resources and per- sonnel over time. Data from an input-throughput-output model and data on time intervals are available from computer registry systems. These data are potential indicators during an MCI response.

Limitations

This study should be interpreted in the context of the following lim- itations. First, owing to the retrospective design and small sample size (48/499) from a single medical center, there might be selection bias. However, all enrolled patients had precise data for each variable, with no missing data. Second, there might be unmeasured confounders, such as time interval of waiting for ED registration, other patients (not from the burn injury MCI) requiring treatment in the ED, and the use of clinical judgment to decide on patient disposition. However, the au- thors did make every attempt to collect all data from a computer data- base and medical records. Third, this study was conducted at a university-affiliated teaching hospital in Linkou in northern Taiwan, which might limit the generalizability of the findings. In addition,

Table 3

Cox proportional hazards regression model to predict patient leaving ED

Univariate analysis

Multivariate analysis

HR (95% CI)

P

Adjusted HR (95% CI)

P

Sex

0.64 (0.33-1.22)

.174

Age

1.00 (0.96-1.04)

.917

Patients registered before hospital MCI response activation

1.79 (0.85-3.74)

.124

Triage

.002

.014

Level I

Reference

Reference

Level II

0.29 (0.14-0.62)

.001

0.46 (0.10-2.14)

.320

Level III

0.25 (0.09-0.68)

.007

0.06 (0.01-0.52)

.011

ED management

Endotracheal intubation

2.68 (1.38-5.19)

.004

3.53 (0.79-15.72)

.098

Central venous access

1.16 (0.61-2.22)

.644

Casualty load

Patient registrations 30 min before and after ED arrival

0.96 (0.90-1.04)

.314

Patient accumulation

1.00 (0.98-1.03)

.928

Time interval

Input

0.95 (0.91-0.99)

.008

1.00 (0.94-1.06)

.914

Throughput

0.98 (0.97-1.00)

.004

0.98 (0.96-1.00)

.163

Output

0.98 (0.97-0.99)

b.001

0.97 (0.96-0.98)

b.001

most of the patients from the dust explosion at Formosa Fun Water Park in New Taipei City, Taiwan, had isolated burn injuries. The number of patients sent to the study hospital did not exceed the surge capacity of the hospital. A comparison of the validity of these study results to results from a different hospital during an MCI response would be of interest.

Conclusion

The triage level and output time intervals were significantly associ- ated with ED LOS in a burn-related MCI. Time effectiveness analyses, using a patient flow model, might serve as an important indicator dur- ing a hospital MCI response.

References

  1. ACS Committee on Trauma. Mass casualties. Resources for Optimal care of the in- jured patient. Chicago, IL: American College of Surgeons; 1999. p. 87-91.
  2. Hirshberg A, Holcomb JB, Mattox KL. Hospital trauma care in multiple-casualty inci- dents: a critical view. Ann Emerg Med 2001;37:647-52.
  3. Hirshberg A, Scott BG, Granchi T. How does casualty load affect trauma care in urban bombing incidents? A quantitative analysis. J Trauma 2005;58(4):686-93 [discus- sion 694-5].
  4. Canadian Association of Emergency Physicians and the National Emergency Nurses Association Joint position statement: access to acute care in the setting of emergency department overcrowding, CJEM 2003;5(2):81-6.
  5. Lee JS, Franc JM. Impact of a two-step emergency department triage model with START, then CTAS, on patient flow during a simulated mass-casualty incident. Prehosp Disaster Med 2015;30(4):390-6.
  6. United States General Accounting Office GAO. Hospital emergency departments: crowded conditions vary among hospitals and communities; 2003[GAO-03-460].
  7. Schiff GD. System dynamics and dysfunctionalities: levers for overcoming emergen- cy department overcrowding. Acad Emerg Med 2011;18(12):1255-61.
  8. Weiss SJ, Rogers DB, Maas F, et al. Evaluating community ED crowding: the commu- nity ED overcrowding scale study. Am J Emerg Med 2014;32(11):1357-63.
  9. Potin M, Senechaud C, Carsin H, et al. Mass casualty incidents with multiple burn victims: rationale for a Swiss burn plan. Burns 2010;36(6):741-50.
  10. Cancio LC, Pruitt Jr BA. Management of mass casualty burn disasters. Int J Disaster Med 2004;2:114-29.
  11. Barillo DJ, Wolf S. Planning for burning disasters: lessons learned from one hundred years of history. J Burn Care Res 2006;27:622-34.
  12. Randic L, Carley S, Mackway-Jones K, et al. Planning for major burn incidents in the UK using accelerated Delphi technique. Burns 2002;28:405-12.
  13. Asplin BR, Magid DJ, Rhodes KV, et al. A conceptual model of emergency department crowding. Ann Emerg Med 2003;42:173e80.
  14. Fatovich DM. Emergency medicine. BMJ 2002;324:958e62.
  15. Cavallini M, Papagni MF, Baruffaldi Preis FW. Fire disasters in the twentieth century. Ann Burns Fire Disast 2007;20:101-3.
  16. Wang TH, Lin TY. Responding to mass burn casualties caused by corn powder at the Formosa Water Park in 2015. J Formos Med Assoc 2015;114(12):1151-3.
  17. Ministry of Health and Welfare Report on the Continued Oversight of Formosa Fun Coast Dust Explosion (1th Week of October), Ministry of Health and Welfare of Taiwan: 2105.
  18. Ciraulo DL, Frykberg ER, Feliciano DV, et al. A survey assessment of the level of pre- paredness for domestic terrorism and mass casualty incidents among Eastern Asso- ciation for the Surgery of Trauma members. J Trauma 2004;56:1033-9.
  19. Frykberg ER. Principles of mass casualty management following terrorist disasters. Ann Surg 2004;239:319-21.
  20. Biancolini CA, Del Bosco CG, Jorge MA. Argentine Jewish community institution bomb explosion. J Trauma 1999;47:728-32.
  21. Cushman JG, Pachter HL, Beaton HL. Two New York City hospitals’ surgical response to the September 11, 2001, terrorist attack in New York City. J Trauma 2003;54: 147-55.
  22. Feeney JM, Goldberg R, Blumenthal JA, et al. September 11, 2001, revisited: a review of the data. Arch Surg 2005;140:1068-73.
  23. Kuwagata Y, Oda J, Tanaka H, et al. Analysis of 2,702 traumatized patients in the 1995 Hanshin-Awaji earthquake. J Trauma 1997;43:427-32.
  24. May AK, McGwin Jr G, Lancaster LJ, et al. The April 8,1998 tornado: assessment of the Trauma system, response and the resulting injuries. J Trauma 2000;48:666-72.
  25. Shoher A, Chang DC, Efron D, et al. Multiple, simultaneous trauma patients: are they worse off? J Trauma 2006;61(3):611-5.

Leave a Reply

Your email address will not be published. Required fields are marked *