Professional Documents
Culture Documents
In our ED, all patients with significant traumatic injuries (Table 1) ordered was divided into five groups: none, 1, 2 to 3 (typical trauma
receive a standardized "trauma system" evaluation. Each patient is series), 4 to 5, and 6 or more. The number of consultants was di-
assessed by a dedicated trauma team consisting of three nurses, vided into three groups: none, 1, and 2 or more. A consultant was
respiratory therapy, a second-year surgery resident, the ED attend- defined a priori as a specialist not normally performing initial patient
ing physician, and a representative from the trauma service (typi- evaluations in the ED (such as an orthopedist or neurosurgeon).
cally a third-year or more senior resident). Initial resuscitation, Each ED chart was reviewed in the same manner to minimize any
blood testing, and radiographs are protocol-driven using preprinted observer bias, and the information was entered directly into a cus-
order sheets. Using a separate critical care flow sheet developed by tomized computer database (Epi Info, version 5.01b~4). Each re-
the ED staff, the primary nurse is responsible for documenting the viewer was given a brief training session in computerized data entry
patient's physical examination, fluids and medications adminis- using Epi Info written by one of us (J.D.), and regular meetings
tered, consultants called, testing performed, and other significant between the investigators and reviewers were held to discuss prob-
events (such as vomiting episodes and allergic reactions). Major lems that arose during data collection and entry.
patient management decisions, such as hospital admission, specialty Summary statistics for total LOS, ED LOS, and holding times
consultation, and the need for ED-based testing (eg, CT scanning, were calculated. Because of the authors' interest in those factors
peritoneal lavage) are made jointly by the ED faculty and the senior that affected the ED-based work-up, the analysis focused on factors
representative from the admitting service. associated with ED LOS (instead of total LOS, which included hold-
Although based on the specific criteria listed in Table 1, activation ing time).
of the trauma system varies by the attending and resident physicians Univariable comparisons of ED LOS were made for dichotomous
on duty, time of day, and the current ED census. In particular, some and ordinal values of all major variables. To assess the effect of
providers activate the system more liberally than others for ex- multiple simultaneous factors affecting ED LOS, a Cox proportional
pected patients who appear to be clinically stable and have only hazards model of time-to-ED discharge was created (PHREG pro-
single-organ system involvement (eg, hip fracture, isolated facial cedure,~5 Statistical Applications Software [SAS]). All variables met
fracture requiring surgery, isolated eye injury). Cox assumptions, ie, log( - log(S)) functions were linear and parallel
Each study chart was evaluated by a single reviewer (B.D. or for all discrete variable levels. Starting with all variables in the
S.S.). Information was extracted on demographics, illness severity model, and controlling for demographics and admitting service, a
as defined by trauma score) 3 disposition, and admitting service. succinct model was created by successively eliminating factors that
Total LOS was defined as the time from triage until report was did not reach at least borderline significance (P ~< .2). All factors in
called to the OR or ICU. The reviewer also determined the ED LOS, the final model were retained at P < .05 level (except for the number
defined as the time from triage until the ED work-up was completed of radiographs, which was retained at a P = .08 level of signifi-
(ie, the patient was simply waiting for an OR suite or ICU bed). cance).
Holding time was defined as the time from completion of the ED Survival curves were then generated to compared the independent
work-up until report was called. For 58 charts, two reviewers inde- effect of each variable that was a significant predictor of ED LOS in
pendently determined the ED LOS variable to examine its reliabil- the final model. These curves showed the proportion of patients
ity. For patients in whom two ED LOS values were available be- with their ED work-up not yet completed with and without the vari-
cause of overlap in reviewing, the value from the principal reviewer able of interest (eg, CT versus no CT) over a given interval of ED
(B.D.) was used. LOS, adjusting for the simultaneous effect of all other variables in
Information was extracted from each chart on use of CT scanning the model. The proportion of patients with their ED work-up not yet
(head, abdomen, or both), and other ED-based special procedures complete at 2 and 4 hours after triage was estimated from this func-
(eg, radiology diagnostic tests, diagnostic or therapeutic maneuvers tion for the typical patient with and without the characteristic of
performed in the ED). The number of plain radiographs and con- interest (eg, CT versus no CT).
sultants was recorded, and was divided into ordinal categories after Hazard ratios were calculated for each factor in the final model.
a preliminary review of the data. The number of plain radiographs They represent the relative likelihood that a patient with a given
characteristic would have completed their ED work-up during a
given thne interval, simultaneously adjusting for all other factors in
TABLE1. Guidelines for Activating Trauma System*
the model. A hazard ratio greater than one signifies a shorter ED
Trauma score <13 LOS, because the likelihood of the ED work-up being completed
Glasgow coma score <11 over time is greater. Likewise, a hazard ratio of less than one sig-
Hypotension or unstable vital signs nifies a longer ED LOS.
Traumatic arrest
RESULTS
High energy dissipation or rapid deceleration injury
Falls of more than 10 feet O n e h u n d r e d e i g h t y - f o u r p a t i e n t s w e r e a d m i t t e d to t h e O R
Near drowning, hypothermia, or hyperthermia
or I C U directly f r o m t h e E D d u r i n g t h e s t u d y period. O f
High-risk motor vehicle crash:
t h e s e , 15 p a t i e n t s (8%) w e r e e x c l u d e d b e c a u s e o f p o o r -
Extrication time more than 20 minutes
Passenger compartment invasion more than 1 foot quality d o c u m e n t a t i o n or missing c h a r t i n f o r m a t i o n . T h e re-
Ejected from vehicle m a i n i n g 169 p a t i e n t s w e r e i n c l u d e d in t h e study.
Death of other passenger T h e overall E D L O S a n d h o l d i n g t i m e s are s h o w n in T a b l e
Any pedestrian struck by vehicle 2. M o s t p a t i e n t s s t a y e d in the E D a b o u t 2 to 3 h o u r s a n d
Other multisystem or major injuries w a i t e d 30 to 60 m i n u t e s for e i t h e r a n O R suite or I C U bed.
Penetrating trauma to head, neck, torso, groin, or midthigh H o l d i n g t i m e s w e r e e x t r e m e l y s k e w e d ; 47% o f t h e p a t i e n t s
Amputation above ankle or wrist h a d h o l d i n g t i m e s o f less t h a n 30 m i n u t e s , w h e r e a s 6 %
Burns (>20% 2nd or 3rd degree) w a i t e d m o r e t h a n 4 h o u r s . T h e m e d i a n v a l u e s a n d t h e inter-
Paralysis or paresthesias
quartile r a n g e (ie, 25th to 75th p e r c e n t i l e values) give a b e t t e r
Assaults, abuse, or battering injuries
picture o f the typical p a t i e n t .
* Adapted from the Duke ED Patient Care Handbook, Joint Com- S e l e c t e d p a t i e n t c h a r a c t e r i s t i c s are s h o w n in T a b l e 3. T h e
mittee for Clinical Practice and Standards, July 1992. m e a n p a t i e n t age w a s 36 --- 22 y e a r s (range, 0 to 95 years),
DAVIS ET AL • FACTORS AFFECTING ED LENGTH-OF-STAY 497
TABLE 2. LOS and Holding Times* (Minutes) of Surgical Table 3 also shows the corresponding values of ED LOS
Critical Care Patients (N = 169) for each variable studied. In this univariable comparison,
four factors were strongly associated with an increased ED
Holding Time
ED LOS for OR/ICU Total LOS
LOS: use of CT scanning, special procedures, four or more
plain radiographs, and two or more consultants. Use of our
Mean ± SD 191 ± 141 69 ± 94t 260 ± 169 protocol-driven trauma evaluation system ( " t r a u m a sys-
Median 157 35 232 tem") was not associated with a shorter ED LOS in the
Range 5-855 0-480 10-930 crude data. However, a greater proportion of the trauma
Mode 120 0 270
system patients underwent CT scanning (65% versus 11%)
* Time waiting for an OR suite or ICU bed. and special procedures (26% versus 9%) than in the non-
1" Large departure from normal distribution. trauma system group, which would tend to conceal a bene-
ficial effect of the trauma system in the unadjusted ED LOS.
with a median age of 36 years. Most of the patients were This was unmasked by examining the effect of the trauma
men, and three quarters had normal trauma scores (ie, score system on ED LOS in each of these groups. F o r the group
of 16). The CT scanning was performed in one third of the with (N = 54) and without (N = 115) CT scanning, the
population. Special procedures consisted of 19 diagnostic trauma system decreased the mean ED LOS from 346 to 249
minutes (P = .02, Kruskal-Wallis test) and from 173 to 75
radiology procedures (including nine aortic arch arterio-
minutes (P < .001, Kruskal-Wallis test), respectively.
grams, four intravenous pyelograms, three ultrasound exam-
A Cox proportional hazards model of time to completion
inations, and one magnetic resonance imaging [MRI]), four
of the ED work-up (ie, ED LOS) was constructed to identify
neurosurgical procedures (two ventriculostomies and two
the independent predictors of prolonged ED LOS. This
Halo placements) and three complex fracture reductions.
method adjusts for the effect of multiple factors determining
Several additional patient characteristics are helpful in de-
ED LOS. The final model, shown in Table 4, corrected for
scribing the study population. One hundred and twenty-
demographics, admitting service, and illness severity (as
seven patients (75%) were sent to the operating room, and
measured by trauma score). Use of either CT scanning or
the remaining 42 (25%) were admitted to the ICU. Most of
special procedures was associated with longer ED LOS (haz-
the patients were admitted by the orthopedic surgery (36%),
ard ratios [HR] less than one), and use of the trauma system
trauma surgery (20%), or neurosurgery (18%) services. Only
was associated with shorter ED LOS (HR greater than one).
10 patients (6%) had immediate indications for surgery be-
This means, for example, that patients who received CT
cause of penetrating trauma to the abdomen (N = 8) or neck
scanning were 4.5 times less likely (ie, 1 divided by .22) to
(N = 2).
complete their ED work-up within a given time interval than
those who did not. By contrast, individuals who went
TABLE 3. Selected Variables and ED LOS through the trauma system were five times more likely to
ED LOS P Value complete their ED work-up within a given time interval than
Variable Number (Minutes) (pair)* those who did not. The number of plain radiographs and
consultants had a smaller, less significant effect. Admitting
Age NS service had only a small effect on ED LOS (data not shown).
/>65 27 (16%) 159 Figures 1 and 2 are survival (time-to-event) curves show-
<65 142 197 ing the effect of CT scanning and the trauma system on
Sex NS
Male 116 (69%) 178 completion of the ED work-up, after correcting for other
Female 53 219 factors that were observed to affect ED LOS. In contrast to
Trauma score NS most survival analysis curves, a lower lying curve is better
16 125 (74%) 186 for L O S ; this means that a greater proportion of patients
13-15 18 261 have had the event of interest (ie, completion of ED work-
~<12 26 168 up) over time. In our data, more than half (54%) of the pa-
CT scan <.001 tients who received CT scanning were still in the ED after 4
Yes 54 (32%) 271 hours, compared with only 7% of those who did not undergo
No 115 154 the test. Even more dramatically, only 1% of the patients
Special procedure .002
Yes 26 (15%) 290
No 143 173 TABLE 4. Selected factors affecting completion of ED
Plain radiographs <.001 Work-up
/>4 24 (14%) 302 Hazard 95% Confidence
1-3 145 172
Variable Ratio Interval
Consultants <,001
/>2 28 (17%) 304 Age > = 65 years 1.54 0.95-2.48
0-1 141 168 Sex = male 1.40 0.98-2.02
Trauma system NS Trauma score 13-15 (versus 16) 0.48 0.27-0.84
Yes 65 (38%) 187 Trauma score ~<12 (versus 16) 1.21 NS
No 104 193 CT scanning 0.22 0.13-0.41
Special procedures 0.32 0.20-0.54
* Comparison of ED LOS among pairs using Kruskal-Wallis test. Trauma system 5,1 2.9-9.0
Abbreviation: NS, not significant.
498 AMERICANJOURNALOF EMERGENCYMEDICINE• Volume13, Number5 • September1995
1
2 hrs: CT scan: 87% still in ED
0.9 No CT scan: 55% still in ED
¢z
= 0.8
"~ 4 hrs: CT scan: 54;~oS::::I inn ED
o ~ 0.7
FIGURE 1. Effect of CT
~. 0.6
UJE scanning on ED LOS. Surgical
o 0.s OR and ICU admissions ad-
justed for age, sex, admitting
= ~ 0.4 service, trauma score, special
.o "6 procedures, and trauma sys-
= 0.3 tem.
o.
2
n
0.2
0.1
0
60 120 180 240 300 360 420 480
Length of ED Stay (minutes)
who went through the trauma system were still in the ED dures, have the largest independent effect on ED LOS in our
after 4 hours, compared with 42% of those who did not. (The population of surgical critical care patients. These types of
10 patients with immediate indications for surgery, all of ED-based decisions may be especially important for deter-
whom received the trauma system, were omitted from this mining overall LOS in critical care patients, whose holding
graph to improve comparability of the two groups.) The times (ie, time waiting for a bed) tend to be short relative to
graphical effect of special procedures on ED LOS is not the length of their ED work-up. It is likely that testing deci-
shown, and was similar to that of CT scanning. sions and use of protocols are important determinants of ED
We examined the reliability of the ED LOS variable in a LOS for other groups of patients.
subset of the charts. Fifty-eight charts were reviewed by two ED-based causes of prolonged ED LOS are only part of
independent observers to determine ED LOS. There was the larger problem of ED overcrowding. Many other factors
perfect agreement (to the minute) in 41% of the cases. The external to the ED are precipitating ED overload nation-
remaining cases had disagreements of 1 to 60 minutes (28%), wide, such as substantial numbers of poor and uninsured
1 to 2 hours (22%), and more than 2 hours (9%). In 69 % of patients, increasing substance abuse and violence, and
the cases, there was disagreement of less than one hour. greater emphasis on outpatient management of sicker pa-
Cases with larger disagreement tended to be those with tients. ~-8 Suggestions to deal with the overcrowding problem
longer total LOS, but the correlation was weak (r = . 14). have focused on improved/flexible staffing, hospital policy
changes, and crisis management. 2'9 Increasing the speed and
DISCUSSION
efficiency of the ED work-up is an important adjunct to these
This study shows that ED-based management decisions, proposals so that EDs can better handle the continuing in-
particularly use of CT a n d other ED-'based special proce- crease in patient volume and acuity that most EDs expect in
"'<_2'
~ 0.3 ting service, trauma score,
special procedures, and CT
2 0.2
0. scanning.
0.1
0
60 120 180 240 300 360 420 480
Length of ED Stay (minutes)
DAVIS ET AL • FACTORS AFFECTING ED LENGTH-OF-STAY; 499
the near future. More importantly, emergency department on the decision to perform CT scanning and special proce-
physicians and nurses may have more control over decisions dures that were clinically indicated. We did not attempt to
involving the ED workup, a luxury less often enjoyed when correct for the availability of ICU beds or OR suites because
pursuing policy and other administrative changes at the hos- these would principally affect the holding time, but not the
pital level. length of the ED-based work-up.
Knowing what factors have the greatest effect on ED LOS A final limitation was our definition of ED LOS. In order
can focus our efforts for change in the ED itself. For this to study factors affecting ED LOS, one would ideally like to
group of surgical critical care patients, it was clear that CT isolate the time required for the ED work-up from the hold-
scanning and other time-consuming diagnostic procedures ing time. However, what defines the point when the ED
performed outside the department, typically in radiology, work-up is complete, and how accurately can we determine
were a major problem (the common parlance is "lost in this time from a chart review? In our data, we had fair agree-
x-ray"). Solutions may include use of ultrafast helical CT ment between two observers using a common-sense proto-
scanning, performing CT scanning or special procedures in a col, supporting the validity of the variable. However, there
CT/fluoroscopy suite located in the ED, and postponing were several cases of large disagreement in ED LOS, and
elective special procedures such as Halo placement and not always in cases where total LOS was excessively long.
nonurgent fracture manipulation until the patient has been Better definition of what constitutes ED LOS versus holding
triaged to an appropriate critical care unit. Less effort might time, as well as prospective data collection, may help im-
be spent trying to limit the number of plain radiographs and prove the validity of the ED LOS variable and the conclu-
consultants, because these theoretical causes of prolonged sions drawn from its use.
ED LOS appeared to be less important.
The use of our ED-based "trauma system" was associated CONCLUSION
with dramatically shorter ED LOS. However, the patients
who received the trauma system still might not be fairly In addition to administrative and societal factors that af-
compared with the remaining population of surgical critical fect ED overcrowding, ED-based patient management deci-
care patients, even after correcting for other factors ob- sions such as use of CT and ED-based special procedures
served to affect ED LOS (and omitting those patients with strongly affected ED LOS in a population of surgical critical
immediate indications for surgery). Nevertheless, the large care patients. Such factors may have a larger effect on over-
difference in ED LOS between the two groups strongly sug- all LOS in this group of patients, who tend to have shorter
gests that the trauma system was at least in part responsible ED holding times. Factors identified as independent predic-
for the observed effect. What is less clear is how the system tors of prolonged ED LOS are candidates for ED-based in-
exerted its beneficial effect. We believe that an automated, terventions to reduce LOS. These interventions might in-
systematic initial evaluation of the patient and a focused clude faster procedures (ultrafast CT scanning), performing
effort by ED staff were both essential to its success. This the tests in the ED proper, or delaying certain procedures
underscores the need for adequate (and if necessary, flexi- until after the patient is transferred to another critical care
ble) ED staffing. A protocol-based trauma system may in- suite. By contrast, factors associated with shorter ED LOS,
crease efficiency, but only if supported by enough ED per- such as a protocol-driven trauma evaluation system, can be
sonnel. It is important to note that our trauma system does implemented.
not encourage less diagnostic testing, nor does it give pa-
tients a higher priority than they would otherwise have by
REFERENCES
virtue of their illness or injury. 1. Andrulis DP, Kell'ermann A, Hintz EA, et al: Emergency
There are several important limitations to our study. First, department and crowding in United States teaching hospitals.
we did not correct for admitting diagnosis or mechanism of Ann Emerg Med 1991 ;20:980-986
injury in examining the factors affecting ED LOS. Certain 2. Lynn SG, Kellermann AL: Critical decision making: Man-
types of injuries or illness might be given higher priority by aging the Emergency Department in an overcrowded hospital.
Ann Emerg Med 1991 ;20:287-292
the ED staff and get through their ED work-up faster. How-
3. Gallagher EJ, Lynn SG: The etiology of medical gridlock:
ever, correcting the survival curves for age, sex, illness se- Causes of Emergency Department overcrowding in New York
verity, and admitting service probably accounted for some of city. J Emerg Med 1990;8:785-790
this diagnostic mix effect. It might be that the beneficial 4. Buesching DP, Jablonowski A, Vesta E, et al: Inappropri-
effect of our trauma system was partly caused by differences ate Emergency Department visits. Ann Emerg Med 1985;14:672-
in diagnosis or mechanism of injury; by omitting the patients 676
with immediate operative indications, we attempted to min- 5. Shesser R, Kirsch T, Smith J, et al: An analysis of Emer-
gency Department use by patients with minor illness. Ann Emerg
imize this bias. Med 1991 ;20:743-748
In addition to problems of diagnostic mix, we were unable 6. Hoffman DM: Hospital admissions through the emergency
to correct for how busy the ED was on a particular day. The department (letter to the editor). JAMA 1992;267:1609
number and acuity level of patients in the ED at any given 7. Stern RS, Weissman JS, Epstein AM: The Emergency De-
time would likely have a strong effect on the length and partment as a pathway to admission for poor and high-cost pa-
efficiency of the ED work-up, and might introduce a bias. tients. JAMA 1991 ;266:2238-2243
For example, use of the trauma system might be less likely 8. Olson CM: Hospital admission through the Emergency
Department. The obstructed pathway. JAMA 1991 ;266:2274
on busy days, which would tend to overstate the system's 9. American College of Emergency Physicians: Measures to
benefit in reducing ED LOS. However, it is likely that the deal with emergency department overcrowding. Ann Emerg
level of patient activity in the ED did not have much effect Med 1990;19:944-945
500 AMERICAN JOURNAL OF EMERGENCY MEDICINE • Volume 13, Number 5 • September 1995
10. American College of Emergency Physicians. Emergency Moore EE, Mattox KL, Feliciano DV (eds): Trauma, ed 2. Nor-
Department observation units. Ann Emerg Med 1988;17:95-96 walk, CT, Appleton and Lange, 1991, pp 139-150
11. Derlet RW, Nishio DA, Cole LM, et al: Triage of patients 14. Dean AG, Dean JA, Burton AH, et al: Epi tnfo, Version 5: A
out of the emergency department: Three year experience. Am J word processing, database, and statistics program for epidemi-
Emerg Med 1992;10:195-199 ology on microcomputers. USD, Incorporated, Stone Mountain,
12. Fromm RE, Gibbs LR, McCallum WB, et al: Critical care in GA, 1990
the Emergency Department: A time-based study, Crit Care Med 15. The PHREG Procedure, In: SAS Technical Report P-229,
1993;21:970-976 SAS/STAT Software: Changes and Enhancements, Release 6.07.
13. Champion H, Socco WJ, Copes WS: Trauma scoring. In: Cary, NC, The SAS Institute, 1992, pp 433-480