Full text of DOI:10.1111/j.1553-2712.2001.tb01121.x

ACADEMIC EMERGENCY MEDICINE • November 2001, Volume 8, Number 11
1091
other patients presenting for care.
In the setting of ED overcrowding,
returns may be a significant contri-
bution to an already strained set of
resources.
Recently Discharged Inpatients as a Source
of Emergency Department Overcrowding
We conducted a retrospective ob-
servational study of a one-month
sample of returns in our institution
to characterize this group. We esti-
mated the resources that this group
of patients required in this second
ED/hospital presentation on a basis
of ED charges and ED LOS. We also
evaluated potential preventability of
returns with individual chart re-
view. None of the previous literature
has attempted to quantify and char-
acterize this patient population as a
possible source of ED overcrowding.
R
YAN B. BAER, MD, JOEL S. PASTERNACK, MD, P
HD,
F
RANK L. ZWEMER ., MD, MBA
JR
Abstract. Objective: To assess the impact on the emergency depart-
ment (ED) of recently discharged inpatients and how they contribute
to and worsen the current situation of ED overcrowding. Methods:
Retrospective, observational study of medical records and billing data
of all patients presenting to the ED within seven days of inpatient
discharge from the hospital (‘‘returns’’) in September 2000. The data
were collected from electronic logs. Billing charges were used to esti-
mate ED resources. Medical records were reviewed to classify visits:
1) new problem, 2) related problem, likely preventable, 3) related prob-
lem, not likely preventable, 4) unable to classify, or 5) incomplete
chart. Results: One-hundred seventy-four returns occurred among
6,290 total ED visits (3%). Significant differences between returns and
total ED patients were noted for length of stay (LOS) (6.58 vs 5.22
hours, p = 0.000), percent admitted (47% vs 19%, p = 0.000), and ED
billing ($1,415.67 vs $391.00, p = 0.000). The highest rate of admission
was for patients presenting 48–72 hours after inpatient discharge
(65.4%). Admission rate was higher for patients presenting >48 hours
than <48 hours (54% vs 33%, p = 0.01). A review of the medical records
(117/174) revealed: 15 new problems (13%); 16 related, likely prevent-
able (14%); 72 related, not likely preventable (62%); 4 unable to assess
(2%); and 10 incomplete charts (9%). Conclusions: The ED is appro-
priately utilized as a safety net for discharged inpatients. Though ‘‘re-
turns’’ are a small percentage of ED patients, they have longer LOSs,
have higher ED charges, and are more frequently admitted. Returns
increase the strain on an already overcrowded ED. Key words: emer-
gency department; overcrowding; readmission. ACADEMIC EMER-
GENCY MEDICINE 2001; 8:1091–1094
METHODS
Study Design. We conducted a
retrospective observational study of
medical records and billing data.
Study approval was obtained from
the research subjects review board
of the university.
Study Setting and Population.
The study was conducted at a uni-
versity teaching hospital with an an-
nual ED census of 72,000 visits. The
facility serves as a regional referral
center. The ED inpatient admission
rate is 17%, and ED patients com-
prise 40% of all inpatient admis-
sions, and 70% of intensive care unit
(ICU) admissions. The payer status
of ED patients is Medicare 16.4%,
Medicaid 21.8%, other insurance
49.8%, and self-pay 11%.
Emergency department (ED) over-
crowding is an increasing problem
both nationally and globally.1,2 This
problem has led to an in-depth
search for causes and solutions.3–7
One factor in this overcrowding that
has not been analyzed in detail is
‘‘bounce-back’’ patients, recently dis-
charged inpatients who return to the
ED within seven days for further
care (‘‘returns’’).
Study Protocol. For the month of
September 2000, data were re-
viewed for all patients who pre-
sented to the ED following discharge
from the inpatient service at the
same facility within seven days of
ED registration.
It is not directly clear what the
relationship is between returns and
changes in inpatient length of stay
(LOS). We do know that in response
to economic pressures, the average
LOS in hospital has decreased over
the past two decades (Table 1).8,9 Pa-
tients are not able to recuperate and
From the Department of Emergency
Medicine, University of Rochester,
Strong Memorial Hospital, Rochester, rehabilitate as long in hospital as in Measurements. The diagnosis,
NY (RBB, JSP, FLZ).
the past; they are discharged earlier disposition, and time of ED stay
to home or to a skilled nursing facil- were obtained from the electronic
ity to continue healing. Patients dis- ED log. Length of stay was mea-
charged prematurely return to the sured from the time of ED registra-
ED for further assessment, treat- tion to time of departure from the
ment, and readmission as necessary. ED. Time to ED presentation was
Appropriately, the ED serves as a calculated from hospital discharge
safety net to these patients to en- time (first visit) to ED presentation
Received March 24, 2001; revision re-
ceived June 28, 2001; accepted June 28,
2001.
Address for correspondence and reprints:
Frank L. Zwemer Jr., MD, MBA, Vice-
Chair for Clinical Affairs, Department
of Emergency Medicine, University of
Rochester, 601 Elmwood, Box 655, Roch-
ester, NY 14642. Fax: 716-473-3516;
e-mail: frank࿝zwemer@urmc.rochester.
edu
sure that they receive proper care.
time (second visit). Time data of the
However, returns require re- study group were compared with
sources that then cannot be used for those for the entire ED population

1092
READMISSIONS
Baer et al. • RECENTLY DISCHARGED PATIENTS AS SOURCE OF OVERCROWDING
TABLE 1. Average Inpatient Length of Stay (Days)
1980
1985
1990
1992
1994
1996
1997
1998
All U.S. hospitals
Strong Memorial Hospital
7.1
NA
6.3
NA
6.2
NA
6.0
7.5
5.6
7.4
5.1
6.8
5.0
6.6
4.9
6.4
over the same time period. Sub- (to previous inpatient diagnosis), as a metric of ED performance and
groups in each category (admissions likely preventable 16 (14%); related,
and discharges) were compared with likely not preventable 72 (62%); un-
function.
Although the percentage of re-
each other and the ED population.
able to classify 4 (2%); and incom-
turns (as a part of overall ED vol-
ume) in this study was relatively
small (3%), it is a very complex
group of patients. These patients
spent more time being evaluated
and treated than other ED patients
(6.58 vs 5.22 hours) and their aver-
age charges were more than 350% of
those of all ED patients. The returns
were more than twice as likely to be
admitted to the hospital as other ED
patients (47% and 19%, respec-
tively). Therefore, the resources
needed to evaluate and to treat
these patients are relatively high
when compared with those for the
‘‘average’’ ED patient.
Inpatient readmission rates are
used widely as a quality indicator.
This metric is usually characterized
as a rate of inpatient readmission 15
or 30 days after inpatient discharge.
One national 30-day rate for 2000 is
approximately 5.1%.10 In contrast,
‘‘returns’’ are patients who present
within seven days of inpatient dis-
charge. At our institutions, returns
comprised 2.8% of inpatient dis-
charges in the sample period.
Billing data were obtained from plete charts 10 (9%). Admission
hospital billing system and were rates for the groups were not signif-
identified as institutional charges icantly different (all p > 0.05), with
for patients on the day of ED regis- the exception of the group with in-
tration.
complete charts (p = 0.011) (Table 3).
To further classify returns, avail- Of the patients with related, likely
able ED records, discharge summa- preventable visits, 37.5% presented
ries, and other medical records were within 24 hours. However, this was
reviewed by two clinicians. After not statistically significant (p = 0.26)
review and discussion, consensus (Table 4).
was reached and patients were cat-
egorized. A third investigator was
used when consensus could not be
reached. Visits were classified as
one of the following: new medical
problem (unrelated); related, likely
preventable; related, not likely pre-
ventable; unable to classify; or in-
complete chart.
The highest rate of admission
was for patients presenting from
48–72 hours (65.4%). Comparing
later presentation (>48 hours) with
earlier (<48 hours), late presenta-
tions had
a significantly higher
readmission rate (54% vs 33%, p =
0.01) (Table 5).
DISCUSSION
Data Analysis. Data were entered
into Excel 97 (Microsoft, Inc., Red-
mond, WA) and imported into Systat
8.0 (SPSS Inc., Chicago, IL). Analy-
ses were performed with one-way
and two-way chi-square, and inde-
pendent t-tests.
The ED appropriately functions as
society’s health care safety net, and
as society and medical practice
change, so does the role of the ED.
There are no data available on na-
tional performance or trends for the
ED resources needed for returns.
Our goal was to characterize returns
We chose a shorter time frame (7
RESULTS
During the month of September
2000, there were 174 returns in
6,290 ED visits (3%). Of the returns,
82/174 (47%) were admitted, com-
prising 7% of the total number of ED
admissions for the month (Table 2).
There were significant differences
between returns and total ED pa-
tients for LOS (6.58 hours vs 5.22
hours, p = 0.000), percent admitted
(47% vs 19%, p = 0.000), and ED
charges ($1,415.67 vs $391.00, p =
0.000) (Table 2). Of note, 21 patients
presented twice to the ED within the
month (42/174 visits). No one pre-
sented more than two times over the
study period.
T
ABLE 2. Characteristics of Returning Patients Compared with All Emergency
Department (ED) Patients for September 2000
Returns
All ED Patients
p-value
Total number
Admissions
% Admissions
ED mean length of stay (hours)
Mean ED charges
174
82
6,290
1,173
N/A
N/A
47%
6.58
$1,415.67
18.65%
5.22
$391.00
0.000
0.000
0.000
TABLE 3. Classification of Returning Patients by Medical Record Review
Total
Number Admissions Treat/Release p-value
New problem
15
16
72
4
7
9
37
0
8
7
35
4
0.796
0.617
0.814
1.00
Related, likely preventable
Related, not likely preventable
Unable to assess
Investigators reviewed 117 of
174 patient visits and placed them
into the following categories: new
medical problem 15 (13%); related
Incomplete chart
10
1
9
0.011

ACADEMIC EMERGENCY MEDICINE • November 2001, Volume 8, Number 11
1093
T
ABLE 4. Likely Preventable Returns
days for returns vs 30 days for na- not be identified that were shared by
tional data) with the interest of ex- all 16 patients. An alternative
amining whether inpatients were method of determining preventabil-
being discharged prematurely, ex- ity would be to apply a time/dispo-
pecting that a shorter time frame sition standard; e.g., any patient
would be more specific for prema- readmitted within 72 hours of dis-
ture discharge. The relationship be- charge could have been prevented.
tween returns and inpatient LOS is Our sample data (Table 4) do not
not clear, and we were unable to suggest any obvious trends, though
make definitive conclusions about it. they describe only 16 patients. In-
However, it seems likely that the dividual chart review (our method),
number of returns has increased as despite being imprecise, labeled as
by Hours after Inpatient Discharge
(n = 16)
Hours
from
Number of
Likely
% of All
Likely
Hospital
Discharge
Preventable Preventable
Visits
Visits
0–24
24–48
48–72
72–96
96–120
120–144
6
1
2
1
3
3
0
37.5%
6.25%
12.5%
6.25%
18.75%
18.75%
0%
hospital financial pressures con- likely
tinue to reduce inpatient LOS.8
From society’s perspective, there tion measure (16 vs 34 at 72 hours).
is benefit to be gained if the overall Our data on the time of ED pre-
preventable
significantly
fewer returns than a time/disposi- 144–168
TOTAL
16
100%
cost of health care can be reduced sentation after discharge also argue
(e.g., reduce inpatient days and cost) against a time/disposition classifica-
for some increased level of ED utili- tion system (Table 5). Those patients study might also look at the patients
zation. What are the savings of a re- who returned to the ED more than who are readmitted through routes
duced global inpatient LOS for a 48 hours after being discharged other than the ED (i.e., readmissions
limited increase in ED volume? It is were readmitted more frequently via ED compared with hospital read-
not clear at what rate of returns the than those who returned earlier missions). That would help to better
health system is failing.
(54% vs 33%). Evaluating discrete delineate the role that the ED plays
We were surprised at the small 24-hour periods, patients presenting as a safety net to other services in
number of likely preventable re- to the ED at more than 48 hours are the health care system.
turns (n = 16), given that 88 (75%) readmitted fairly consistently (about
The relationship between rate of
of the returns had presenting prob- 50% each for days 3–7). Therefore, returns and inpatient LOS is not
lems related to inpatient care and 48-hour and 72-hour measures are clear, and should be evaluated.
diagnoses (‘‘related’’ was subdivided not sensitive indicators for identify-
into likely and not likely prevent- ing preventable readmissions.
able). An incremental 16 patients
Deciding which admissions were
likely preventable was based on sub-
jective criteria. Standard criteria
with minimal interobserver variabil-
ity could be developed and used. The
likely preventable returns did not
appear to be admitted more fre-
over a month period would not over-
LIMITATIONS AND
FUTURE
crowd any ED. Without a set of firm
criteria, it is possible that we under-
estimated likely preventable visits.
QUESTIONS
All returns, regardless of appro- This study only begins to look at the quently or present earlier than other
priateness of care in hospital, con- role of the ED as a safety net for in- patient groups. A larger sample with
tribute to the strain of resources in patient bounce-backs within the more objective criteria will be
an already overcrowded ED. To eval- health care system. It is limited by needed for further evaluation.
uate the matter of appropriate care, its small sample size (one month), at
A larger sampling will have to be
this study addresses the issue of pre- a single institution. Our hospital is performed to better evaluate
ventable visits. Preventability was a large regional referral center so whether a possible time/disposition
determined after individual medical many patients seen in our ED may endpoint might be useful. In our
record review and discussion. Each be bounce-backs from other institu- one-month sample, the patients pre-
of the 16 likely preventable cases tions; we have not accounted for senting later were admitted more
was complex and carefully consid- these patients. To further under- frequently than those presenting
ered. A single set of criteria could stand the matter of returns, a future early. However, with a larger sample
TABLE 5. Likelihood of Inpatient Admission When Returning to the ED after Inpatient Discharge (n = 168)
Hours after Inpatient
Discharge
Total No.
Patients
Patients
Admitted
Patients
Discharged
% (Admitted/Total)
% Discharged/Total)
0–24
24–48
48–72
26
26
26
22
21
28
19
9
8
34.6%
30.8%
65.4%
50%
52.4%
50%
17
18
9
11
10
14
8
65.4%
69.2%
34.6%
50%
47.6%
50%
17
11
11
14
11
72–96
96–120
120–144
144–168
57.9%
42.1%
TOTAL
168*
81*
48.2%
87
51.7%
*Six patients presenting during the study period did not have exact times available. One of these six was readmitted.

1094
READMISSIONS
Baer et al. • RECENTLY DISCHARGED PATIENTS AS SOURCE OF OVERCROWDING
4. Derlet RW, Richards JR. Overcrowd-
of data it is possible that there could sion rate for returns is for those pa-
be significantly more readmissions tients presenting 48–72 hours after
ing in the nation’s emergency depart-
ments: complex causes and disturbing ef-
fects. Ann Emerg Med. 2000; 35:63–8.
5. Kellermann AL. De´ja` vu. Ann Emerg
Med. 2000; 35:83–5.
within the first 48 or 72 hours.
inpatient discharge. The ED care of
Hospital billing data were used returns is resource-heavy and con-
as a measure of ED resources. Emer- tributes significantly to ED over-
gency department charges are fre- crowding.
6. Grumbach K, Keane D, Bindman A.
Primary care and public emergency de-
partment overcrowding. Am J Health.
1993; 83:372–8.
7. American College of Emergency Phy-
sicians. Measures to deal with emer-
gency department overcrowding. Ann
Emerg Med. 1990; 19:944–5.
8. Baxt WG. Crisis at America’s teach-
ing hospitals. Ann Emerg Med. 2000; 36:
145–8.
9. National Center for Health Statistics,
Centers for Disease Control and Preven-
tion, US Department of Health and Hu-
man Services, Washington, DC, 2001.
10. Comparative Data Base, University
HealthSystem Consortium (UHC), Oak-
brook, IL, 2001.
quently poorly correlated with ED
care, and an alternative measure
could be used.
References
1. Graff L. Overcrowding in the ED: an
international symptom of health care
system failure. Am J Emerg Med. 1999;
17:208–9.
CONCLUSIONS
Recently discharged inpatients re-
turn to the ED for care in significant
numbers. Compared with other ED
patients, returning patients have a
longer length of stay, incur higher
ED charges, and are admitted much
2. Shih F-Y, Ma MH-M, Chen S-C, et al.
ED overcrowding in Taiwan; facts and
strategies. Am J Emerg Med. 1999; 17:
198–202.
3. Derlet RW, Richards JR, Kravitz RL.
Frequent overcrowding in U.S. emer-
gency departments. Acad Emerg Med.
more frequently. The highest admis- 2001; 8:151–5.
ⅷ
Instructions for Contributors to Reflections
Reflections submissions should be presented as original nowledgment of manuscript and photograph acceptance
poems, short stories, or black-and-white photographs that will be made in writing to the contributor.
portray the humanistic (joys, sorrows, struggles, chal-
If a photograph is submitted for consideration, the most
lenges, and realities) aspect of the practice of emergency original image available (slide, negative, or photograph) and
medicine. Reflections will be published on a space-available two 5 × 7-inch black-and-white prints should be sent. The
basis.
original image(s) will be returned if requested. Each print
Poems and short stories must have a title and body. and slide should be labeled with the last names of the con-
Two hard copies and one electronic copy should accompany tributors and an arrow indicating the top of the image.
each submission. A cover letter should identify the submis-
It is the responsibility of the contributor to obtain pa-
sion as Reflections. Photographs should have a title and tient consent for use of the photograph in a publication if
may have a caption of no more than 50 words. All submis- a person (patient, health care worker, family member) is
sions must be accompanied by a signed copyright release identifiable in the photograph. A copy of this written con-
and author disclosure form.
sent must be sent with the submission.
Contributors must provide the names, highest aca-
Send submissions and images to Academic Emergency
demic degrees, addresses, 3-mail addresses, and phone/fax Medicine, % Society for Academic Emergency Medicine,
numbers of all contributors (including photographer). Ack- 901 North Washington Avenue, Lansing, MI 48906.