High occupancy increases the risk of early death or readmission after transfer from intensive care.

Abstract

To determine whether a lack of intensive care unit beds was leading to premature patient discharge from the intensive care unit and subsequent early readmission or death.
Prospective cohort study.
A single Canadian tertiary care teaching hospital.
All intensive care unit admissions between January 1, 1989 and December 31, 1996 were collected prospectively for inclusion in a registry database.
None.
There was a positive correlation between early readmission or death and average quarterly intensive care unit percent occupancy (p = .001). During the study period, 8693 patients experienced 10,185 admissions to intensive care. Of the 8222 patients remaining under active treatment (patients under palliative care were excluded), there were 455 (5.5%) adverse events (431 intensive care unit readmissions and 24 deaths) in the first 7 days post intensive care unit discharge. Patients requiring a new surgical intervention with postoperative intensive care unit admission were not considered readmissions. In a multivariate analysis, significant risk factors for an adverse event included age >35 yrs, particular diagnoses (respiratory diagnoses, sepsis, neurosurgery, thoracic surgery, and gastrointestinal diagnoses), Acute Physiology and Chronic Health Evaluation II score, and intensive care unit length of stay. Discharge from the intensive care unit at a time of no vacancy was also a significant risk factor for intensive care unit readmission or unexpected death with an adjusted relative risk of 1.56 (95% confidence interval 1.05, 2.31).
Increased patient occupancy within an intensive care unit is associated with an increased risk of early death or intensive care unit readmission post intensive care unit discharge. Overloading the capacity of an intensive care unit to care for critically ill patients may affect physician decision-making, resulting in premature discharge from the intensive care unit.

Full text

Page 1

High occupancy increases the risk of early death or readmission
after transfer from intensive care*
Carla A. Chrusch, MD, MSc, FRCPC; Kendiss P. Olafson, MD, MPH, FRCPC;
Patricia M. Mcmillan, MD, FRCPC; Daniel E. Roberts, MD, FRCPC; Perry R. Gray, MD, FRCSC
C
to its population is finite (1), some degree
of patient prioritization and selection is
needed. This has led to the development
of statements regarding the allocation of
intensive care unit (ICU) resources and
guidelines for admission, discharge, and
aring for critically ill patients
requires a substantial com-
mitment of expensive physical
and human resources. As a ju-
risdiction’s ability to provide critical care
triage of patients (2–4). Optimal benefit
from intensive care depends on appropri-
ate admission and discharge decisions
based on a combination of need and po-
tential benefit. Intensive care is most
beneficial to severely ill, unstable patients
who are likely to improve with treatment.
Patients at the two extremes of illness—
either at low risk of death, such as pa-
tients admitted for monitoring (5), or
those patients with a high mortality de-
spite aggressive interventions—have lim-
ited benefit from an ICU admission (6). In
addition to need and benefit, the avail-
ability of resources can influence both
admission and discharge decisions (7, 8).
Resources vary from one institution to
another as well as within the same insti-
tution over time (9). Bed availability can
be influenced by system problems leading
to inefficient use of ICU beds (8, 10).
When ICU care is no longer appropri-
ate or required, patients are transitioned
to a lower intensity and less expensive
level of care. Premature transfer places
the patient at risk of an adverse event,
such as death or a readmission to ICU
that may have been prevented if the pa-
tient had remained in the ICU (11–13).
Patients who require readmission to the
ICU usually return with greater severity
of illness (14) and have a poorer progno-
sis with higher mortality rate and longer
hospital lengths of stay compared with
similar patients who do not require ICU
readmission (11, 13–19). Readmission
rates have therefore been proposed to be
important quality indicators for ICU care
(14, 20). Unnecessarily delaying transfer
from the ICU, however, decreases the
overall availability of ICU resources. ICU
physicians are challenged with the re-
sponsibility of both caring for patients
already under their care and patients in
their jurisdiction who require admission.
In practice, this means that patients al-
ready admitted to the ICU are in compe-
tition with the next patient who requires
ICU care.
Our hypothesis was that an inadequate
number of ICU beds was leading to pre-
mature patient discharge and subsequent
ICU readmission or unexpected death. We
sought to answer this question by exam-
*See also p. XXX.
From the Department of Critical Care Medicine,
University of Calgary, Calgary, AB, Canada; Depart-
ments of Medicine and Surgery, University of Mani-
toba, Winnipeg, MB, Canada; Department of Anaesthe-
sia, McGill University, Montreal, QC, Canada.
Supported, in part, from the Section of Critical Care
Medicine, University of Manitoba.
This work was performed at the Health Sciences
Centre, Winnipeg, MB, Canada.
The authors have not disclosed any potential con-
flicts of interest.
For information regarding this article, E-mail:
carla.chrusch@calgaryhealthregion.ca
Copyright © 2009 by the Society of Critical Care
Medicine and Lippincott Williams & Wilkins
DOI: 10.1097/CCM.0b013e3181a57b0c
Objective: To determine whether a lack of intensive care unit
beds was leading to premature patient discharge from the inten-
sive care unit and subsequent early readmission or death.
Design: Prospective cohort study.
Setting: A single Canadian tertiary care teaching hospital.
Patients: All intensive care unit admissions between January
1, 1989 and December 31, 1996 were collected prospectively for
inclusion in a registry database.
Interventions: None.
Measurements and Main Results: There was a positive corre-
lation between early readmission or death and average quarterly
intensive care unit percent occupancy (p ? 0.001). During the
study period, 8693 patients experienced 10,185 admissions to
intensive care. Of the 8222 patients remaining under active treat-
ment (patients under palliative care were excluded), there were
455 (5.5%) adverse events (431 intensive care unit readmissions
and 24 deaths) in the first 7 days post intensive care unit dis-
charge. Patients requiring a new surgical intervention with post-
operative intensive care unit admission were not considered
readmissions. In a multivariate analysis, significant risk factors
for an adverse event included age >35 yrs, particular diagnoses
(respiratory diagnoses, sepsis, neurosurgery, thoracic surgery,
and gastrointestinal diagnoses), Acute Physiology and Chronic
Health Evaluation II score, and intensive care unit length of stay.
Discharge from the intensive care unit at a time of no vacancy
was also a significant risk factor for intensive care unit readmis-
sion or unexpected death with an adjusted relative risk of 1.56
(95% confidence interval ? 1.05, 2.31).
Conclusions: Increased patient occupancy within an intensive
care unit is associated with an increased risk of early death or
intensive care unit readmission post intensive care unit dis-
charge. Overloading the capacity of an intensive care unit to care
for critically ill patients may affect physician decision-making,
resulting in premature discharge from the intensive care unit. (Crit
Care Med 2009; 37:000–000)
KEY WORDS: intensive care; readmission; mortality; transfer;
vacancy; quality of health care
1 Crit Care Med 2009 Vol. 37, No. 10

Page 2

ining the association between ICU bed
occupancy and other markers of unit ac-
tivity with the rate of early death or re-
admission after transfer from intensive
care in patients eligible for readmission
to the ICU. In addition, we attempted to
identify risk factors for and the conse-
quences of ICU readmission.
MATERIALS AND METHODS
The Health Sciences Centre is the primary
tertiary care teaching hospital in the Province
of Manitoba serving a population of 1.5 mil-
lion. The Centre treats trauma, neurosurgery,
and lung and bone marrow transplant patients
as well as general medical and surgical pa-
tients. There are separate medical and surgical
ICUs. The units are staffed by residents, fel-
lows, and intensivists with all admission, dis-
charge, and transfer decisions being made by
the attending intensivist. The medical ICU is a
closed unit. The surgical ICU is semiclosed,
meaning that the physician of record is the
attending surgeon, but medical care within
the unit is directed by the attending intensiv-
ist. Patients in the ICU are cared for by nurses
who have completed additional training in
critical care. The nurse/patient ratio in the
ICU is 1:1 or 1:2 in the case of stable lower
acuity patients, at the discretion of the nurse-
in-charge. Care on the wards is organized into
medical and surgical services composed of
medical students and resident physicians and
led by specialist medical staff. The Centre has
surgical step-down units where patients are
cared for by ward medical and nursing staff
with a nurse/patient ratio of 1:4. Patients are
transferred between the general ward and the
step-down unit at the direction of the ward
medical staff and housestaff. A rapid response
or formal follow-up team was not in place
during the study period. All patients admitted
to the medical or surgical ICU of the Health
Sciences Centre between January 1, 1989 and
December 31, 1996 were included in the study.
Cardiology patients not requiring mechanical
ventilation are cared for in a separate unit and
were not included in this study population.
For all patients admitted to the ICU, demo-
graphic information, admission diagnoses,
Acute Physiology and Chronic Health Evalua-
tion (APACHE) II score, and admission and
discharge times were collected prospectively
at the time of admission by trained data col-
lectors and entered into a database (Critical
Care Manager, Chelmsford, ON, Canada). Pa-
tient-related data were collected and main-
tained in a manner approved by the Research
Ethics Board of the University of Manitoba.
Medical admissions were grouped by primary
system involved and surgical admissions by pro-
cedure, resulting in 13 diagnostic categories.
Patients who were readmitted to the ICU
within 7 days of their transfer from the ICU
were identified in the database retrospectively.
Patients requiring a new surgical intervention
with postoperative ICU admission, those des-
ignated as palliative at the time of ICU dis-
charge, and those discharged from the hospi-
tal and readmitted were excluded. The hospital
medical records of patients who left the ICU
under active treatment and died within 7 days
of transfer were reviewed by one of the inves-
tigators. Patients were classified as under ac-
tive treatment if cardiac resuscitation was at-
tempted. Patients who died within 7 days of
transfer to the ward without resuscitative ef-
forts were designated as palliative and ex-
cluded from further analysis. The decision not
to resuscitate the patient had to be docu-
mented in the chart before the cardiac arrest.
Readmission or death occurring within 7 days
of discharge from the ICU, while still under
active treatment, was considered an adverse
outcome.
A number of markers of unit activity were
derived a priori. Unit occupancy at the time of
each patient’s admission and discharge from
the ICU was derived from discharge dates and
times and the number of operational beds. The
terms vacancy and no vacancy at the time of
discharge refer to the status of the unit after
that patient is discharged. That is, no vacancy
means that a patient is discharged from an
overcensus ICU. Vacancy means that there are
one or more beds available after that patient’s
discharge. The number of admissions in the 4
hrs preceding an admission, the time to the
next admission, and the time of day that a
discharge occurred (9 AM to 2 PM; 2 PM to 9 PM;
9 PM to 9 AM) were determined. It was hypoth-
esized that these variables may be markers for
premature discharge precipitated by the need
to admit another patient. Patients were also
retrospectively given a discharge rank for each
day based on their discharge time, specifically
either the first discharge of the day shift, the
second, the third, and so forth. It was assumed
that patients discharged earlier in the day
were judged to be more appropriate discharges
by the attending physician than patients dis-
charged later in the day. Interaction terms for
no vacancy and unit of discharge as well as
vacancy and unit activity markers were de-
rived post hoc.
The average adverse event rate was calcu-
lated as the total number of premature deaths
and ICU readmissions occurring within 7 days
of ICU discharge divided by the number of
discharges at risk for readmission. In addition,
the average percent ICU bed occupancy was
the average midnight ICU bed census divided
by the number of funded ICU beds within the
hospital. The average readmission and prema-
ture death rates and average percent ICU oc-
cupancy are calculated on a quarterly annual
basis.
Statistical Analysis
The relationship between average readmis-
sion rates and percent occupancy by calendar
quarter was examined, using a linear regres-
sion model.
Categorical variables were compared by
chi-square analysis, and continuous variables
were expressed as the mean and SD and were
compared using paired and unpaired Student’s
t tests as appropriate. Univariate and multivar-
iate logistic regression, using generalized es-
timating equations, were employed to model
the outcome of an adverse event (death or
readmission) within 7 days of discharge from
the ICU. Regression with generalized estimat-
ing equations adjusts the variance for corre-
lated observations, in this case, multiple ad-
missions of the same patient.
A p ? .05 was considered statistically sig-
nificant, as was a relative risk where the 95%
confidence intervals (CIs) did not cross zero.
Statistical analysis was performed, using SAS
version 9.1 (SAS Institute, Cary, NC).
RESULTS
During the study period, the number
of staffed ICU beds within the hospital
was ten in the surgical ICU whereas it
increased from eight to ten beds in the
medical ICU in 1992. The ICU bed occu-
pancy and adverse event rates by calendar
quarter for each unit are shown in Figure
1. The average midnight percent occu-
pancy was 71.2% (range ? 22%–115%).
A histogram of the midnight census is
shown in Figure 2. The relationship be-
tween average ICU bed occupancy and
readmission rates by calendar quarter
was statistically significant (r ? .55; p ?
.001 by linear regression) and is shown in
Figure 3.
A total of 8693 patients experienced
10,185 admissions to the ICUs. Patient
demographics, APACHE II scores, ICU
length of stay, and survival are shown in
Table 1. Overall ICU survival was 83.5%
with 8222 discharges to the ward still
receiving active treatment. Of the 8222
discharges remaining under active treat-
ment on the ward, there were 24 deaths
(0.3%) and 431 ICU readmissions (5.2%)
in the first 7 days post ICU discharge. In
the first 3 days, 74.7% of these events
occurred.
Patients who required readmission
had an APACHE II score of 22.0 ? 8.6
compared with an APACHE II score of
19.1 ? 6.6 on their prior admission (p ?
.01 by paired Student’s t test). The ICU
mortality of readmitted patients was
21.3%, which is substantially higher than
2 Crit Care Med 2009 Vol. 37, No. 10

Page 3

those who did not require readmission
who had a ward mortality of 0.3% (p ?
.01 by chi-square test). However, it is no
different from patients who had only a
single ICU admission (n ? 7538 patients)
with an ICU mortality of 19.0% (p ? .27
by chi-square test). The most common
reasons for readmission were respiratory
conditions (21.7%), cardiovascular prob-
lems (17.5%), and infections (16.8%).
The diagnosis on readmission was the
same as for the initial admission in 49%
of cases. The population at risk experi-
enced 71 cardiac arrests in 66 patients
within 7 days of discharge from the ICU.
Twenty-four of those patients died on the
ward. An additional 20 patients died sub-
sequently in the ICU.
Patient characteristics, admission
characteristics, and markers of unit ac-
tivity were examined regarding their re-
lationship to readmission or death after
transfer from the ICU. The results of the
univariate analysis are shown in Table 2
as the relative risk and 95% CIs. Age ?35
yrs and certain diagnoses (respiratory
conditions, sepsis, neurosurgical condi-
tion, thoracic surgery, and gastroenter-
ological disorders) were associated with
an increased risk on an adverse outcome.
Other important admission characteris-
tics included high APACHE II scores and
an ICU length of stay of ?3 days.
Markers of unit activity that were sig-
nificant in the univariate analysis in-
cluded discharge from an overcensus ICU
(relative risk ? 1.62; 95% CI ? 1.10,
2.39) and the rank of discharge. Specifi-
cally, the second or later patient dis-
charged from the ICU on a given day was
at increased risk (relative risk ? 1.21;
95% CI ? 1.00, 1.46). There were 402
discharges from the ICU under condi-
tions of no vacancy. The number of ad-
missions that preceded or followed a pa-
tient’s discharge from the ICU, the time
to the next admission, and time of dis-
charge were not significant risk factors
and, therefore, they were not included in
further analyses. The interactions be-
tween no vacancy and other unit activity
factors were not significant.
The adjusted relative risks of an ad-
verse event were derived, using a multi-
variable model and generalized estimat-
ing equations. Results are shown in Table
2. Age ?35 yrs remained significant (ad-
justed relative risk ? 1.46; 95% CI ?
1.02, 2.07) as did particular diagnoses
(respiratory diagnoses, sepsis, neurosur-
gery, thoracic surgery, and gastrointesti-
nal diagnoses), APACHE II score and ICU
length of stay. Being discharged from the
ICU when there was no vacancy was the
only unit activity variable that remained
significant with an adjusted relative risk
for an adverse outcome of 1.56 (95%
CI ? 1.05, 2.31). The risk of an adverse
event when discharged from an overcen-
sus surgical ICU was not statistically dif-
ferent from an overcensus medical ICU
(relative risk ? 1.71; 95% CI ? 0.75,
3.88). Of the 402 patients discharged
from an overcensus ICU, 34 patients
(8.5%) required ICU readmission within
7 days compared with 397 readmissions
(5.1%) in the 7398 patients discharged
when the ICU was not over census.
DISCUSSION
We have presented the early death and
readmission rates from ?10,000 ICU ad-
missions in a single institution over an
8-yr period. In this study, we have made a
number of refinements in the definition
of readmission as far as timing, active
care, and emergency surgery are con-
cerned. Previous studies considered any
subsequent ICU admission during the
same hospitalization a readmission (14,
15, 17, 21). The earlier a readmission
occurs post ICU discharge, the more
likely that readmission can be attribut-
able to the discharge decision. Our study
identified adverse outcomes that oc-
curred within 7 days of ICU discharge on
the assumption that, after this period, the
decision to transfer the patient could not
realistically be considered premature.
Care was taken to ensure that only pa-
tients still eligible for ICU readmission
were included in the at-risk population by
excluding those who were under pallia-
tive care on the ward. Patients may re-
quire readmission to the ICU because of
recurrence of their initial problem, ap-
pearance of a new problem, and elective
or emergency surgical procedure. We ex-
cluded operative interventions necessitat-
Figure 1. Intensive care unit (ICU) bed occupancy and adverse event rates by calendar quarter.
Figure 2. Histogram of daily midnight bed occupancy. ICU, intensive care unit.
3Crit Care Med 2009 Vol. 37, No. 10

Page 4

ing postoperative ICU care, as transfer
from the ICU to a ward does not create
indications for surgery.
Our study supports the previous find-
ings of increased age, ICU length of stay,
admission APACHE II score, and gastro-
intestinal bleeding as significant risk fac-
tors for early readmission or death (17,
21).. When compared with patients who
had a single ICU admission, the readmit-
ted patients did not experience an in-
creased ICU mortality. The reason for re-
admission was the same as the reason
for initial ICU admission in half of the
patients, a finding similar to that of
Chen et al (17).
In our examination of 8222 ICU pa-
tients eligible for readmission, we found
an effect of vacancy even after adjustment
for patient and admission characteristics
(adjusted relative risk 1.56; 95% CI ?
1.05, 2.31). Previous studies aimed at ex-
amining the effect of a lack of ICU beds
on patient outcomes were negative. The
study by Singer and colleagues included a
large proportion of coronary care unit
and monitoring admissions (9). As most
of these patients are at the low risk of
death, it is not surprising that rationing
had little impact. Strauss et al examined
1150 ICU admissions with an overall ICU
mortality of 12.7% (7). Patients dis-
charged when there was a lack of ICU bed
availability had higher severity of illness
scores at discharge, suggesting that bed
resources affected physician decision-
making. Although their readmission rate
was very high (14%), there was no asso-
ciation observed between bed availability
and adverse outcomes after ICU dis-
charge. Their ICU mortality, however,
was lower than our 16.5% ICU mortality
and perhaps a lack of effect was again
from selecting patients at a lower risk of
death.
In our institution, being discharged
from the ICU at night did not increase the
risk of an adverse outcome. This is differ-
ent from a previous study that described
an increased mortality for patients dis-
charged between 8 PM and 7:59 AM (22).
Another marker of unit activity, dis-
charge rank, was significant in the uni-
variate analysis, but not after adjustment
in the multivariable model. Our assump-
tion was that, in an overcensus situation,
discharge order reflects how suitable a
patient was for discharge, with the “best”
patients being discharged first. However,
in the case of multiple discharges leaving
to different wards, the ICU discharge rank
may be more dependent on the exact time
that a ward bed is ready than a reflection
of the patient’s clinical status.
Our hypothesis was that premature
discharge from the ICU was leading to an
increased rate of readmissions. It is chal-
lenging to account and adjust adequately
for all of the patient and hospital factors
that may affect the risk of readmission. In
this study, we assumed that patients dis-
charged from a nonoverfull ICU were ap-
propriate for discharge based on the judg-
ment of the attending intensivist. This
decision is based on the clinical status of
the patient as well as the attending phy-
sician’s intimate knowledge of capabili-
ties and limitations of the healthcare sys-
tem in which they work. Consideration
could be given to using a score to further
adjust for patient acuity at the time of
discharge, such as acute physiology
scores, nursing workload measures, or
organ dysfunction scores.
Acute physiology scores have been de-
signed to predict hospital mortality and
work very well for that purpose. However,
they award points for severe physiologic
abnormalities, such as hypotension, in-
fections, and metabolic derangements
that are generally resolved before ICU dis-
charge. Physiology scores, therefore, may
not add much to patient discrimination
at the time of discharge.
An important component of the risk of
ICU readmission is the care received out-
side of the ICU. In reality, this care is
dependent on the available medical and
Figure 3. Average intensive care unit (ICU) bed occupancy and adverse event rates per calendar
quarter.
Table 1. Characteristics of patients admitted to the ICU between January 1, 1989 and December 31, 1996
Both Units MedicalSurgical
Medical Versus
Surgical, p
Number of admissions
Age, yrs
Males
APACHE II
ICU length of stay, days
ICU survivors
Active treatment on ward
Readmitted within 7 daysa
Died within 7 daysa
10185
59.3 ? 18.1
6261 (61.5%)
18.6 ? 8.6
3.9 ? 6.5
8501 (83.5%)
8222
431 (5.2%)
24 (0.3%)
40526133
57.8 ? 18.9
2266 (55.9%)
21.5 ? 9.7
4.6 ? 7.4
3012 (74.3%)
2839
157 (5.5%)
13 (0.5%)
60.3 ? 17.5
3995 (65.1%)
16.7 ? 7.1
3.5 ? 5.8
5489 (89.5%)
5383
274 (5.0%)
11 (0.2%)
?.001
?.001
?.001
?.001
?.001
.40
.04
ICU, intensive care unit; APACHE, Acute Physiology and Chronic Health Evaluation.
aThe denominator is the number of discharges from the ICU and under active treatment on the ward.
4 Crit Care Med 2009 Vol. 37, No. 10

Page 5

nursing staff and the patient workload on
a given service or ward. For nursing,
there is a well-validated measure of work-
load. Therapeutic Intervention Scoring
System scores have been used to assess
ICU workload and resource allocation
(23, 24). Discharge Therapeutic Interven-
tion Scoring System scores may have
been helpful to adjust for patient acuity at
the time of discharge from the ICU. A
downside of the Therapeutic Intervention
Scoring System score is that, although it
describes what interventions a patient re-
ceived, the score does not indicate
whether the intervention was necessary.
For example, a patient ready for the reg-
ular ward may continue to undergo
hourly vital signs and continuous electro-
cardiographic monitoring during his/her
ICU stay even though the patient no
longer requires this type of monitoring
and will not be receiving it after transfer.
Medical staff workload is a combination
of the number of patients and their com-
plexity. An organ dysfunction score, such
as the Multiple Organ Dysfunction Score
(25), may be a way to adjust for patient
acuity on ICU discharge as well as be a
reflection of patient complexity.
In this study, we did not examine the
specific ward location from which pa-
tients were readmitted. In our institu-
tion, patients in step-down units and gen-
eral ward beds are cared for by the same
medical and nursing staff. Step-down
units provide a higher nursing ratio and
degree of vigilance than the general
wards. Without examining specific ward
locations and having detailed ward staff-
ing information, we cannot exclude that
the readmission rate was significantly in-
fluenced by a small number of hospital
wards.
This study is limited by the fact that it
represents a single institution. In addi-
tion, with a study of this duration,
changes to the general functioning of the
unit or institution could impact patient
outcomes. During the study period, both
ICUs continued to function as closed
units. All care was directed by trained
intensivists and provided by nurses with
specialty training in critical care. There
were no significant changes in staffing
ratios or protocols within the ICU beds.
Although we have attempted to account
for all factors in the ICU known to influ-
ence patient outcomes, we do not have a
measure for changes in medical or nurs-
ing practices or workloads on the hospital
wards, which may have influenced the
rate of ICU readmission. In addition, the
presence of a rapid response team may
mitigate the risk of readmission. Future
studies would benefit from including a
measure of patient complexity and work-
load at the time of ICU discharge.
Iapichino and colleagues have previ-
ously shown that an ICU occupancy of
?80% was associated with increased ICU
mortality (26). In our study, we observed
a positive association between occupancy
and readmission rates over a large range
of occupancy rates in contrast to a critical
percent occupancy above which adverse
events are increased.
Rationing of intensive care therapy oc-
curs based on patient needs combined
with the perceived potential benefit in
addition to resource availability. In-
creased ICU bed occupancy with limited
vacancy may adversely affect patient out-
come through either premature ICU dis-
charge or the late admission of appropri-
ate ICU patients. This study focuses only
on the outcomes of patients discharged
from the ICU with respect to unit occu-
pancy and therefore may underestimate
the effect of bed resources on patient
outcomes. In this study, discharge from
an ICU with no vacancy is significantly
associated with an increased risk of ICU
readmission. This finding raises the hy-
pothesis that bed resources influence
physician decision-making, resulting in
premature ICU discharge and adverse pa-
tient outcomes.
CONCLUSIONS
Over the 8 yrs of the study period, an
average of 5.2% patients experienced ei-
Table 2. Unadjusted and adjusted relative risk of early death or readmission to the intensive care unit
Covariate
Relative Risk
(95% CI)
Adjusted Relative Risk
(95% CI)
Patient factors
Male
Age ?35 yrs
Admission factors
MICU (vs. SICU)
Diagnostic group (vs. cardiac diagnoses)
Overdose/poisoning
Cardiac surgery
Vascular surgery
Medical miscellaneous
Trauma
Cardiac
General surgery
Neurological
Respiratory
Sepsis
Neurosurgery
Thoracic surgery
Gastroenterology
APACHE II score vs. APACHE II ?10
APACHE II 10–19
APACHE II 20–29
APACHE II 30–39
APACHE II ?40
LOS in ICU vs. LOS ?3 days
LOS 3–10 days
LOS ?10 days
Unit activity factors
Admission in previous 4 hrs vs. no admission
1 admission
?2 admissions
Time to next admission vs. no admission
within 8 hrs of a discharge
4–8 hrs
?4 hrs
Time of Discharge vs. 9 AM–4 PM
4 PM–9 PM
9 PM–9 AM
No vacancy at discharge
Second or later discharge of the day
0.80 (0.63, 1.00)
1.59 (1.12, 2.25)
1.01 (0.82, 1.24)
1.46 (1.02, 2.07)
1.14 (0.93, 1.39)0.85 (0.62, 1.16)
0.19 (0.07, 0.53)
0.35 (0.21, 0.55)
0.83 (0.53, 1.30)
0.92 (0.50, 1.70)
0.95 (0.62, 1.47)
1.00
1.49 (1.00, 2.21)
1.50 (0.68, 3.39)
1.64 (1.08, 2.51)
2.15 (1.42, 3.25)
2.29 (1.44, 3.62)
2.77 (1.75, 4.38)
2.86 (1.76, 4.64)
0.31 (0.11, 0.88)
0.40 (0.23, 0.70)
0.91 (0.53, 1.56)
0.96 (0.52, 1.78)
0.99 (0.58, 1.68)
1.00
1.53 (0.96, 2.42)
1.54 (0.68, 3.48)
1.73 (1.11, 2.68)
1.66 (1.08, 2.55)
1.95 (1.14, 3.33)
2.79 (1.64, 4.73)
2.55 (1.54, 4.25)
1.84 (1.24, 2.75)
3.55 (2.35, 5.35)
4.67 (2.72, 8.02)
8.05 (2.19, 29.6)
1.50 (1.00, 2.24)
2.16 (1.40, 3.33)
2.75 (1.52, 4.97)
4.57 (1.17, 17.8)
2.40 (1.94, 2.96)
3.42 (2.65, 4.42)
1.72 (1.35, 2.18)
2.22 (1.65, 2.98)
0.96 (0.76, 1.22)
1.02 (0.60, 1.74)
0.91 (0.69, 1.21)
1.06 (0.86, 1.30)
1.01 (0.80, 1.26)
0.69 (0.43, 1.11)
1.62 (1.10, 2.39)
1.21 (1.00, 1.46)
1.56 (1.05, 2.31)
1.16 (0.94, 1.42)
CI, confidence interval; MICU, medical intensive care unit; SICU, surgical intensive care unit; APACHE,
Acute Physiology and Chronic Health Evaluation; LOS, length of stay; ICU, intensive care unit.
5Crit Care Med 2009 Vol. 37, No. 10