Journal of Critical Care (2012) 27, 422.e1–422.

e9

A model to predict short-term death or readmission after

intensive care unit discharge☆,☆☆,★
Islem Ouanes MD a,b , Carole Schwebel MD c , Adrien Français MSc d , Cédric Bruel MD a ,
François Philippart MD a,e , Aurélien Vesin MSc d , Lilia Soufir MD f ,
Christophe Adrie MD e,g,h , Maïté Garrouste-Orgeas MD a,d ,
Jean-François Timsit MD, PhD c,d , Benoît Misset MD a,e,⁎
and the Outcomerea Study Group 1
a
Intensive Care Unit, Saint-Joseph Hospital, Paris, France
b
Intensive Care Unit, Fattouma Bourguiba University Hospital, Monastir, Tunisia
c
Intensive Care Unit, Albert Michallon University Hospital, Grenoble, France
d
INSERM U823, Epidemiology of Cancer and Severe Illnesses, Albert Bonniot Institute Grenoble, France
e
Université Paris-Descartes, Sorbonne Paris Cité, Faculté de Médecine, Paris, France
f
Surgical Intensive Care Unit, Saint Joseph Hospital, Paris, France
g
Intensive Care Unit, Delafontaine Hospital, St Denis, France
h
Physiology Department, Cochin University Hospital, Assistance Publique-Hôpitaux de Paris, France

Keywords:
Abstract
discharge;
Objective: Early unplanned readmission to the intensive care unit (ICU) carries a poor prognosis, and
readmission;
post-ICU mortality may be related, in part, to premature ICU discharge. Our objectives were to identify
death;
independent risk factors for early post-ICU readmission or death and to construct a prediction model.
intensive care unit;
Design: Retrospective analysis of a prospective database was done.
score(or scoring system)
Setting: Four ICUs of the French Outcomerea network participated.
Patients: Patients were consecutive adults with ICU stay longer than 24 hours who were discharged
alive to same-hospital wards without treatment-limitation decisions.
Main results: Of 5014 admitted patients, 3462 met our inclusion criteria. Age was 60.6 ± 17.6 years,
and admission Simplified Acute Physiology Score II (SAPS II) was 35.1 ± 15.1. The rate of death or

☆
Conflicts of interest: The authors have no personal or financial conflicts of interest to declare.
☆☆
Financial support: Outcomerea is supported by nonexclusive educational grants from Pfizer, Aventis Pharma France, Wyeth France, and Ely Lilly and
by public grants from the Centre National de la Recherche Scientifique and Institut National de la Santé et la Recherche Medicale. The Outcomerea data
warehouse project was also supported by a grant from the Agence Nationale de Valorisation de la Recherche. These grants had no role in the design or conduct of
the study; the collection, management, analysis, or interpretation of the data; or the preparation, review, or approval of the manuscript.
★
The results reported in this manuscript were presented, in part, at the 2009 annual meetings of the French Society of Intensive Care Medicine, French
Society of Anesthesia and Intensive Care, and European Society of Intensive Care Medicine.
⁎ Corresponding author. Service de Réanimation, Fondation Hôpital Saint-Joseph, 185 rue Raymond Losserand, 75014 Paris, France. Tel.: +33 1 44 12 34
15; fax: +33 1 44 12 32 80.
E-mail address: bmisset@hpsj.fr (B. Misset).
1
The members of the Outcomerea Study Group are listed in the appendix.

0883-9441/$ – see front matter © 2012 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcrc.2011.08.003
422.e2 I. Ouanes et al.

ICU readmission within 7 days after ICU discharge was 3.0%. Independent risk factors for this outcome
were age, SAPS II at ICU admission, use of a central venous catheter in the ICU, Sepsis-related Organ
Failure Assessment and Systemic Inflammatory Response Syndrome scores before ICU discharge, and
discharge at night. The predictive model based on these variables showed good calibration. Compared
with SAPS II at admission or Stability and Workload Index for Transfer at discharge, discrimination was
better with our model (area under receiver operating characteristics curve, 0.74; 95% confidence
interval, 0.68-0.79).
Conclusion: Among patients without treatment-limitation decisions and discharged alive from the ICU,
3.0% died or were readmitted within 7 days. Independent risk factors were indicators of patients'
severity and discharge at night. Our prediction model should be evaluated in other ICU populations.
© 2012 Elsevier Inc. All rights reserved.

1. Introduction bed surgical ICU; and D, a 14-bed medical surgical ICU. A

Many critically ill patients experience clinical deteriora-
tion or death shortly after discharge from the intensive care
unit (ICU). In earlier studies, 8% to 10% of patients
discharged from the ICU died or required ICU readmission
during the same hospital stay [1-5]. Studies have demon-
strated that ICU discharge decisions depend on organiza-
tional factors such as workload and ICU bed availability [6-
8]. Furthermore, premature ICU discharge was responsible
for 22% to 42% of readmissions [9,10] and has led to rank
ICU readmission among the top indicators for ICU quality
[11]. Intensive care unit readmission has been associated
with worsening of the original disease process, higher
hospital costs, and increase in hospital mortality [5,7,10,12].
Therefore, knowledge of the risk factors for ICU readmission
may help to identify high-risk patients before determining
whether discharge is appropriate [7,9,12]. A rating scale
based on the subjective prognosis by attending intensivists
has been reported to predict hospital mortality after ICU
discharge [13]. In several countries, critical care outreach
teams are available for assessing patients being considered
for ICU admission [14]. A tool based on objective data could
help discriminating which patients should not be discharged
without risk of bad outcome and/or should undergo a special
surveillance after the ICU stay [15]. Our objective was to
develop a tool for prediction of early ICU readmission or
death during the same hospital stay of patients without
treatment-limitation decisions. Using a population-based
cohort ICU records, we hypothesized that objective factors
could discriminate between patients who were and were not
likely to die or be readmitted to the ICU within 7 days of
ICU discharge.
2. Methods
We retrospectively studied a prospective cohort of
patients from 4 ICUs (named A, B, C, and D) in tertiary
care hospitals filling the Outcomerea database. A is a 10-bed
medical surgical ICU; B, an 18-bed medical ICU; C, a 10-
and C are located in the same 460-bed nonprofit private
hospital, B is located in a 1500-bed public university-
affiliated hospital, and D is located in a 600-bed public
hospital. In each center, the study was active when each
consecutive patient of the unit could be collected in the
database. The study started in 1998 for center A and 2006 for
centers B, C, and D. The study ended in 2008 for centers C
and D and in 2010 for centers A and B. We included all
patients older than 18 years who spent more than 24 hours in
the ICU, had no prior ICU admissions, received no
treatment-limitation decisions, and were discharged alive
from the ICU to a ward in the same hospital.
2.1. Definitions and management policies
The primary outcome variables were unplanned ICU
readmission or death in a non-ICU care setting within 7 days
(including day 7) after ICU discharge [15]. Mortality was
assessed at the end of the ICU and hospital stay, respectively.
Hospital mortality includes patients who died during the ICU
stay and after the ICU stay. All 4 ICUs are “closed” units
composed of a minimal number of 6 full-time senior
intensivists. At least 1 intensivist is present overnight and
during weekends. Based on French regulations, the patient-
to-nurse ratio is 2:5, and the patient-to-nurse assistant ratio is
4:0. These requirements were fulfilled in the 4 participating
centers. Decisions to limit life-supporting interventions are
made during repeated meetings of ICU staff and relatives
[16]. All limitation decisions were collected prospectively in
the database. Patients were discharged to the referring wards
or to wards specialized in the relevant acute disease, at the
discretion of the attending intensivist. Patients experiencing
signs of organ dysfunction necessitating continuous moni-
toring and minor pharmacologic or device-related support
were discharged toward a multidisciplinary step-down unit
(only available in Saint Joseph hospital) or a coronary care
unit (all hospitals).
The policies of the 4 ICUs included that discharges have to
be planned and take place when the patient no longer required
continuous surveillance or organ supply. A discharge at night
was considered as a marker of bed shortage.
Short-term death or readmission after ICU discharge
2.2. Data collection and baseline data
In each participating ICU, a specifically trained physician
collected the data daily on a case report form using VIGIREA
and RHEA data capture software (OUTCOMEREA, Rosny-
sous-Bois, France). All codes and definitions were established
before study initiation. The following information were
recorded for each patient: age and sex, admission category
(medical, scheduled, or unscheduled surgery), McCabe score
[17], and origin (home, ward, or emergency department).
Severity of illness was evaluated on the first ICU day using the
Simplified Acute Physiology Score (SAPS II) [18] and
Logistic Organ Dysfunction (LOD) score [19]. Knaus scale
definitions were used to record preexisting chronic organ
failures [20]. Organ dysfunctions were quantified daily using
the Sepsis-related Organ Failure Assessment (SOFA) [21] and
LOD scores. Nursing workload was assessed with the Nine
Equivalents of nursing Manpower Score (NEMS) [22].
Discharge at night was considered between 9 PM and 7 AM
[23]. According to French law, this database study did not
require informed consent. The study was approved by the
institutional review board of the Clermont-Ferrand University
Hospital, Clermont-Ferrand, France (CECIC [Comité d’Ethi-
que des Centres d’Investigation Clinique] institutional review
board number 5891; approval number : 2007-16).
2.3. Database quality
The data capture software automatically conducted
iterative checks for internal consistency of most of the
variables upon entry into the database. Queries generated by
these checks were resolved with the source ICU before
incorporation of the new data into the database. At each
participating ICU, data quality was controlled by having a
trained physician from another participating ICU check a 2%
random sample of the study data.
2.4. Model development and statistical analysis
We developed a multivariable model for critical illness
during hospitalization in 3 steps: (1) assessment of candidate
variable quality and categorization of continuous predictors,
(2) construction of a parsimonious model, and (3) develop-
ment of a point score. When choosing candidate variables for
the model, we considered clinical relevance and generaliz-
ability (availability in the usual ICU records). Results are
expressed as number (percentage) and as mean (SD) or
median (interquartile [IQ]). Variables were compared using
the χ 2 or Fisher exact test and the Wilcoxon or Kruskal-
Wallis test as appropriate. Variables associated with death or
ICU readmission within 7 days after ICU discharge with P b
.2 by univariate analysis were introduced in the multivariate
model and subsequently selected to improve model devi-
ance. The Akaike information criterion was used to identify
the best model. The assumption that quantitative variables
422.e3
were linear in the logit was checked using cubic polynomials
and graphic methods. In the absence of log linearity,
continuous variables were transformed into qualitative
variables according to the slope of the cubic polynomial
functions and to the distribution of the variables. A pooled
test of clinically relevant 2-way interactions was performed
on the final model, and correlations between selected
variables were verified. We checked for potential colinearity
of the variables included in the final model. R values of less
than 0.2 were considered acceptable.
Our assessment of model performance was goodness of fit
as evaluated by the Hosmer-Lemeshow statistic and
calibration curves. Lower Hosmer-Lemeshow values and
higher P values (N.05) indicate better fit. We assessed
discrimination using the area under the curve (AUC) of the
receiver operating characteristics (ROC) curve.
The model was validated using bootstrap resampling with
500 replicates from the development cohort of 3462 patients.
The quartiles of the 500 ROC-AUC values provided the
confidence interval for model discrimination. The final
model was evaluated comparatively with other existing
models using the method of McNeil and Hanley [24] to
compare ROC-AUC values. Analyses were computed using
the SAS 9.1.3 package (SAS Institute, Cary, NC), R, and
Medcalc 5.00 (Medcalc, Ghent, Belgium).
3. Results
3.1. Study population
The patient selection is shown on Fig. 1. The age of the
3462 patients included in the study was 60.6 ± 17.6 years,
and their admission SAPS II was 35.1 ± 15.1 points. Before
ICU admission, they were either in a ward (51.4%), at the
emergency department (47.9%), or at home (0.6%). The
most common reasons for ICU admission were acute
respiratory failure (21.1%), shock (18.1%), coma (13.3%),
and acute renal failure (5.2%).
Of the 3462 patients, 224 (6.5%) were either readmitted or
died during the same hospital stay and 102 (3.0%) within 7
days after ICU discharge (Figs. 1 and 2). The characteristics
at first and second admissions of the 74 patients readmitted
within 7 days of ICU discharge are displayed on Table 1.
Eight (10.8%) died during the second ICU stay, and 3 died in
the hospital after the second ICU stay.
3.2. Risk factor identification and model for
predicting ICU readmission
In univariate comparisons (Table 2), during the first ICU
stay, patients who were readmitted or died within 7 days after
ICU discharge were older, were more likely to have
metastatic cancer, had higher severity scores at admission,
had undergone more frequently a venous or arterial catheters,
422.e4 I. Ouanes et al.

5218 ICU Admission

5014 Patients

1552 Patients excluded: 980 died in the

ICU, 335 orders to withdraw or withhold
intensive therapies before ICU discharge,
138 with ICU stay 24 hours and 99
transferred to another facility

3462 patients with ICU LOS > 24 h, discharged

alive from the ICU, without treatment limitation
decisions, to a ward in the same hospital

102 (3.0 %) 122 (3.5 %) 3238 patients

patients patients discharged alive
readmitted or readmitted or from the hospital
dead within 7 dead after 7 days after one ICU stay
days of ICU of ICU discharge after ICU
discharge discharge

74 (2.1%) 28 (0.8%) died 39 (1.1%) 83 (2.4%) died

readmitted within 7 days readmitted after after 7 days of
within 7 days of after ICU 7 days of ICU ICU discharge
ICU discharge discharge discharge

Fig. 1 Patient flowchart.

had more severe organ dysfunction at ICU discharge, and these patients had lower systolic arterial pressure, higher
had been more often discharged at night. Among the blood urea nitrogen, more hypo- or hyperleucocytosis, hypo-
subclasses of the SAPS II score at admission (e-Table 1), or hyperkalemia, and lower serum bicarbonates. They had a

50

45

40

35
Number of patients

Death Readmission
30

25

20

15

10

0
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 > 31

Days after ICU discharge

Fig. 2 Distribution of the daily death and ICU readmission after ICU discharge.
Short-term death or readmission after ICU discharge
Table 1 Characteristics at first and second admission of the 74 patients who were readmitted to the ICU within 7 days of ICU discharge
during the same hospital stay
SAPS II score at admission, median (IQ)
Main symptom at admission, n (%)
Acute respiratory failure
Cardiogenic shock
Coma
Acute renal failure
Septic shock
Hemorrhagic shock
Scheduled surgery
Type of ICU stay, n (%)
Medical
Emergent surgery
Scheduled surgery
The main symptom responsible for both the first and the second admissions was the same in 25 (34%) of 74 patients.
higher maximum value of Systemic Inflammatory Response
Syndrome (SIRS) score in the last 2 days of ICU stay and a
higher score for all the subclasses of SOFA on the day of
ICU discharge. The patients who died had more comorbid-
ities and severity scores at admission than those who were
readmitted (e-Table 2).
Through multivariate analysis, 6 variables independently
predicted death or ICU readmission within 7 days after ICU
discharge (Table 3): age, SAPS II at ICU admission, use of a
central venous catheter during the ICU stay, maximum value
of the SIRS score during ICU stay, SOFA score at ICU
discharge, and discharge at night.
The model exhibited good calibration (Hosmer-Lemeshow
χ 2, P = .93) and fair discrimination (ROC-AUC, 0.74; 95%
confidence interval, 0.68-0.79). Based on this predictive
model, we constructed the Minimizing ICU Readmission
(MIR) score, for minimized ICU readmission, calculated in
points. The scoring table is displayed in e-Table 3. A score
more than 150 points is associated with a positive predictive
value (risk of death or readmission within 7 days) of 8% and
negative predictive value of 98% (e-Table 4).
3.3. Comparison with previous models
We compared the performance MIR score to SAPS II at
admission and Stability and Workload Index for Transfer
(SWIFT) on the day of ICU discharge. Goodness of fit was
satisfactory for all models. Our MIR model showed better
discrimination than the other 2 models. The ROC-AUCs (95%
IQ range) were 0.74 (0.68-0.79), 0.64 (0.59-0.70), and 0.61
(0.55-0.67) for MIR, SAPS II, and SWIFT, respectively (Fig. 3).
4. Discussion
In our cohort of patients without treatment-limitation
decisions, the incidence of death or ICU readmission after
422.e5
First admission Second admission
40 (19) 36 (19)
20 (27.0) 31 (40.5)
3 (4.1) 11 (14.9)
11 (14.9) 11 (14.9)
7 (9.5) 7 (12.2)
16 (21.6) 12 (8.1)
3 (4.1) 2 (2.7)
14 (18.9) 0 (0)
49 (66.2) 54 (73)
8 (10.8) 20 (27)
17 (23) 0 (0)
ICU discharge was 6.6%. The incidence was 3.0% within 7
days of ICU discharge, and independent risk factors for this
outcome were the patients' age, indicators of severity during
the ICU stay, sepsis, organ dysfunction at ICU discharge,
and discharge at night. We developed a probability model
that predicted death or ICU readmission within 7 days after
ICU discharge with good calibration and fair discrimination.
A systematic review [25] showed that early ICU
readmission occurred in 5% to 10% of patients [1-5]. Our
relatively low rate may be because of exclusion of those
patients who had treatment-limitation decisions, a population
at high risk for rapid death [15,16]. We had decided not to
include patients with treatment-limitation decisions to
construct a score that helps to identify patients with a
preventable risk of life-threatening post-ICU ward events. In
other studies, premature ICU discharge was associated with
increased hospital mortality [26].
Among the independent risk factors for death or ICU
readmission within 7 days after ICU discharge in our study,
having a higher severity at ICU admission and at ICU
discharge is consistent with those factors identified in other
studies [15,16]. These authors found that these 3 factors
together with do-not-resuscitate orders were the main risk
factors for post-ICU inhospital mortality. When assessing
severity of illness measured with the APACHE or SAPS
systems at ICU admission or discharge, the risk of ICU
readmission may increase by 43% with each severity score
increase of 1 SD [27]. The use of a central venous catheter
is an indicator of therapeutic intensity and may represent
the persistence of severity during the ICU stay. A high
SIRS score during the last 2 days, because of either
community- or ICU-acquired infection demonstrates that all
the acute diseases of the patient may not have been solved
before discharge.
The decision to discharge the patient from the ICU is
usually based on the personal conviction of the physician at a
given time. A patient can be discharged before he/she has
422.e6 I. Ouanes et al.

Table 2 Characteristics of the patients during their first ICU stay according to their status on day 7 after ICU discharge
Alive and not readmitted to the ICU Dead or readmitted to the ICU P
Patients, n 3360 102
Age, mean (SD) 60.4 (17.6) 65.9 (14.2) .002
Female sex, n (%) 1295 (38.5) 30 (29.4) .06
Transfer from ward, n (%) 1677 (49.9) 57 (55.9) .23
Days from hospital to ICU admission, median (IQ) 0 (3) 1 (2) .70
Location before first ICU admission, n (%) .35
Ward 1723 (51.2) 59 (57.9)
Home 22 (0.7) 0 (0.0)
Emergency 1615 (48.1) 43 (42.2)
Type of patients, n (%) .70
Medical 2291 (68.2) 69 (67.6)
Emergent surgery 434 (12.9) 11 (10.8)
Scheduled surgery 634 (18.9) 22 (21.6)
Comorbidities, n (%)
Metastatic cancer 113 (3.4) 8 (7.8) .02
Immune deficiency 307 (9.1) 14 (13.7) .12
Diabetes 489 (14.6) 19 (18.6) .26
McCabe score, n (%) .48
No fatal disease 2297 (68.4) 64 (62.7)
Ultimately fatal disease 949 (28.3) 34 (33.3)
Rapidly fatal disease 112 (3.3) 4 (3.9)
Infection at admission, n (%) 525 (15.6) 21 (20.6) .17
Infection acquired in the ICU, n (%) 481 (14.3) 22 (21.6) .046
Days from last infection to ICU discharge, median (IQ) 5 (8) 6 (9) .70
Severity at first ICU admission, median (IQ)
LOD 3 (4) 5 (3) b.0001
SOFA 5 (5) 6 (4) b.0001
SAPS II 33 (20) 41 (21) b.0001
Procedures used during the ICU stay
Arterial catheter, n (%) 1051 (31.3) 50 (49.0) .0004
CVC, n (%) 1728 (51.4) 78 (76.5) b.0001
MV, n (%) 1780 (50.0) 60 (58.8) .09
MV duration, median (IQ) 4 (8) 5 (10) .17
NIV, n (%) 573 (17.1) 19 (18.6) .69
Days from MV weaning to ICU discharge, median (IQ) 2 (3) 3 (3) .36
Days from VA drug weaning to ICU discharge, median (IQ) 5 (8) 5 (8) .88
Severity on the day before discharge, median (IQ)
Pao2/FiO2 360 (200) 302 (168) b.0001
GCS 15 (0) 15 (1) .001
SOFA 2 (3) 3 (3) b.0001
LOD 1 (3) 3 (4) b.0001
NEMS 17 (7) 15 (7) .48
Maximal SIRS score in last 2 days before ICU discharge 2 (2-3) 3(2-3) .0001
ICU LOS, med (IQ) 4 (5) 5 (8) .45
Discharge at night, n (%) 143 (4.3) 12 (12) .002
Discharge during weekend, n (%) 514 (15.3) 17 (16.7) .67
Discharge when the ICU is full, n (%) 1627 (48.4) 46 (45.1) .55
Discharge to step down unit, n (%) 480 (21.6) 21 (26.3) .33
Discharge to a long-term facility 26 (1.2) 0 (0) 1
CVC indicates central venous catheter; MV, mechanical ventilation; NIV, noninvasive ventilation; PaO2/FiO2, ratio of partial pressure of arterial oxygen over
fraction of inspired oxygen; GCS, Glasgow Coma Scale score; LOS, length of stay; VA, vasoactive.

recovered enough autonomy to be admitted to a conventional Intensive care unit discharge is also influenced by the
ward because a patient in a more serious condition has to take necessity of high nurse staffing because patients may
his/her place in the ICU. Discharge at night is usually experience persistent high dependency after vital functions
considered as a marker of bed shortage [23]. have recovered. This may be taken on in down-step units that
Short-term death or readmission after ICU discharge 422.e7

Table 3 Independent predictors of death or ICU readmission within 7 days after ICU discharge as identified by multivariate logistic regression
Parameter Estimate SE Odds ratios 95% CI P
Intercept −5.59 0.32 – – – b.0001
Parameter related to the severity of the patient
SAPS II at admission 0.017 0.006 1.02 1.005 1.03 .007
Central venous catheter during stay 0.74 0.24 2.1 1.3 3.4 .003
SOFA at ICU discharge 0.19 0.03 1.2 1.1 1.3 b.0001
SIRS score N2 in last 2 days before ICU discharge 0.61 0.20 1.8 1.2 2.8 .003
Parameter related to admission and/or discharge policy
Discharged during night time 0.92 0.33 2.5 1.3 4.9 .006
CI indicates confidence interval.

are currently developing across the world [28,29]. In our
cohort, a high nursing workload at discharge, measured by
the NEMS score, was not a risk factor for death or
readmission. This may be because of the existence of
download units in the participating hospitals.
Other scores were constructed to predict ICU read-
mission. The Sabadell score, based on the physician's
subjective impression at ICU discharge, was found to predict
post-ICU ward mortality [13]. The SWIFT score study [15]
supports the usefulness of assessing disease severity at ICU
discharge, as respiratory and neurologic dysfunctions at
discharge were significant risk factors for ICU readmission.
Our MIR model has a better discrimination than the main
preexisting predictive models. For the SAPS II score, this is
because it was designed to predict hospital mortality and not
ICU readmission. For the SWIFT score, it may be because
the score was tested in an external cohort, whereas MIR was
tested in the cohort used for construction. The validity of the
MIR score needs to be assessed in external cohorts.
Eventually, as the ROC-AUC value is not close to 1, the
possibility of unknown confounders remains high and should
be searched in further cohorts. Finally, the MIR score is more
complex to calculate than the Sabadell and the SWIFT scores
because it includes a series of data from the entire stay in the
100
80
Sensitivity
ICU. Therefore, its calculation should be automatically
computed by the software of the ICU database. The absence
of digital medical record in several ICUs may be a limit to its
routine use.
In the absence of a treatment-limitation decision, to
discharge a patient before his/her physical autonomy allows
adequate care in a ward may be considered as premature. Our
score could alert the physician once he/she has decided to
discharge a patient to check if the patient has a high
probability to be shortly readmitted and then to reevaluate the
necessity to keep him/her under intensive surveillance and/or
supportive care. This should be addressed through interven-
tional trials in other ICU populations. A high rate of
discharged patients with a high MIR score could be
indicative of inadequate observance of discharge procedures
or of shortage of beds and help to adjust local policies.
5. Conclusion
In a cohort of ICU patients without treatment-limitation
decisions, the rate of death or ICU readmission within 7 days
after ICU discharge was 3.0%. The independent risk factors
for this outcome reflect disease severity during the ICU stay
and discharge at night. They are easy to identify in current
databases. Our MIR probability model is well calibrated and
better in terms of discrimination than prior published models.
Its validity should be evaluated in other ICU populations.
Supplementary materials related to this article can be
found online at doi:10.1016/j.jcrc.2011.08.003.

60 MIR
SAPS II
SWIFT
40

Contributions of the authors to the manuscript

20

Conception and design: IO, JFT, and BM.

0
0 20 40 60 80 100
Acquisition of data, analysis and interpretation of data:
All.
100-Specificity
Participation in writing the article: IO, CS, AF, MGO,
Fig. 3 Receiver operating characteristics curves for SAPS II at JFT, and BM.
ICU admission and for the MIR and SWIFT scores on the day Critical revision for intellectual content: All.
before ICU discharge. Approval of version to be published: All.
422.e8
Appendix A. Members of the Outcomerea
Study Group
Scientific committee
Jean-François Timsit (Hôpital Albert Michallon and
Institut National de la Santé et la Recherche Medicale
[INSERM] U823, Grenoble, France), Pierre Moine (Surgical
ICU, Denver, Colo), Elie Azoulay (Medical ICU, Hôpital
Saint Louis, Paris, France), Yves Cohen (ICU, Hôpital
Avicenne, Bobigny, France), Maïté Garrouste-Orgeas (ICU
Hôpital Saint- Joseph, Paris, France), Lilia Soufir (ICU,
Hôpital Saint-Joseph, Paris, France), Jean-Ralph Zahar
(Department of Microbiology, Hôpital Necker, Paris, France),
Christophe Adrie (Department of Physiology, Hôpital Cochin,
France), Adel Benali (Microbiology and Infectious Diseases,
Hôpital Saint-Joseph, Paris, France), Christophe Clec'h (ICU,
Hôpital Avicenne, Bobigny, France), and Jean Carlet (ICU,
Hôpital Saint-Joseph, Paris, France).
Biostatistical and informatics expertise
Jean-Francois Timsit (Epidemiology of Cancer and
Severe Illnesses, INSERM U823, Grenoble, France), Sylvie
Chevret (Medical Computer Sciences and Biostatistics
Department, Hôpital Saint-Louis, Paris, France), Corinne
Alberti (Medical Computer Sciences and Biostatistics
Department, Robert Debré, Paris, France), Adrien Francais
(Epidemiology of Cancer and Severe Illnesses, INSERM
U823, Grenoble, France), Aurélien Vésin (Epidemiology of
Cancer and Severe Illnesses, INSERM U823, Grenoble,
France), Sylvain Anselme (Epidemiology of Cancer and
Severe Illnesses, INSERM U823, Grenoble, France), Muriel
Tafflet (Outcomerea, France), Frederik Lecorre (Supelec,
France), and Didier Nakache (Conservatoire National des
Arts et Métiers, Paris, France).
Investigators of the Outcomerea database
Christophe Adrie (Department of Physiology, Hôpital
Cochin, France), Bernard Allaouchiche (Surgical ICU, Hôpital
Edouard Herriot, Lyon), Caroline Bornstain (ICU, Hôpital de
Montfermeil, France), Alexandre Boyer (ICU, Hôpital
Pellegrin, Bordeaux, France), Antoine Caubel (ICU, Hôpital
Saint-Joseph, Paris, France), Anh-Tuan Dinh-Xuan (Depart-
ment of Physiology, Hôpital Cochin, France), Christine
Cheval (SICU, Hôpital Saint-Joseph, Paris, France), Jean-
Pierre Colin (ICU, Hôpital de Dourdan, Dourdan, France),
Anne-Sylvie Dumenil (Hôpital Antoine Béclère, Clamart
France), Adrien Descorps-Declere (Hôpital Antoine Béclère,
Clamart France), Jean-Philippe Fosse (ICU, Hôpital Avicenne,
Bobigny, France), Samir Jamali (ICU, Hôpital de Dourdan,
Dourdan, France), Christian Laplace (ICU, Hôpital Kremlin-
Bicêtre, Bicêtre, France), Thierry Lazard (ICU, Hôpital de la
Croix Saint-Simon, Paris, France), Eric Le Miere (ICU,
Hôpital Louis Mourier, Colombes, France), Laurent Mon-
tesino (ICU, Hôpital Bichat, Paris, France), Bruno Mourvil-
lier (ICU, Hôpital Bichat, France), Benoît Misset (ICU,
Hôpital Saint-Joseph, Paris, France), Delphine Moreau
(ICU, Hôpital Saint-Louis, Paris, France), Roman Mounier
I. Ouanes et al.
(ICU, Hôpital Louis Mourier, Colombes, France), Etienne
Pigné (ICU, Hôpital Louis Mourier, Colombes, France),
Carole Schwebel (University hospital A Michallon, Grenoble,
France), Jean-Francois Timsit (University hospital A Michal-
lon, Grenoble, France ), Gilles Troché (Hôpital Antoine
Béclère, Clamart France), Marie Thuong (Agence de
Biomédicine, Saint Denis, France), Dany Golgran-Toledano
(CH Gonesse, France), and François Vincent (ICU, Hôpital
Avicenne, Bobigny, France).
Clinical research assistants
Caroline Tournegros (Hôpital Albert Michallon), Silvia
Calvino (Hôpital Albert Michallon), Loic Ferrand (Hôpital
Albert Michallon), Samir Bekhouche (Hôpital Saint Louis),
and Kaoutar Mellouk (Hôpital Saint-Joseph, Paris, France).
References
[1] Alban RF, Nisim AA, Ho J, Nishi GK, Shabot MM. Readmission to
surgical intensive care increases severity-adjusted patient mortality. J
Trauma 2006;60(5):1027-31.
[2] Campbell AJ, Cook JA, Adey G, Cuthbertson BH. Predicting death
and readmission after intensive care discharge. Br J Anaesth
2008;100(5):656-62.
[3] Cooper GS, Sirio CA, Rotondi AJ, Shepardson LB, Rosenthal GE. Are
readmissions to the intensive care unit a useful measure of hospital
performance? Med Care 1999;37(4):399-408.
[4] Ho KM, Knuiman M. Bayesian approach to predict hospital mortality
of intensive care readmissions during the same hospitalisation.
Anaesth Intensive Care 2008;36(1):38-45.
[5] Metnitz PG, Fieux F, Jordan B, Lang T, Moreno R, Le Gall JR.
Critically ill patients readmitted to intensive care units—lessons to
learn? Intensive Care Med 2003;29(2):241-8.
[6] Chen LM, Martin CM, Keenan SP, Sibbald WJ. Patients readmitted to
the intensive care unit during the same hospitalization: clinical features
and outcomes. Crit Care Med 1998;26(11):1834-41.
[7] Durbin Jr CG, Kopel RF. A case-control study of patients readmitted to
the intensive care unit. Crit Care Med 1993;21(10):1547-53.
[8] Wu CJ, Coyer F. Reconsidering the transfer of patients from the
intensive care unit to the ward: a case study approach. Nurs Health Sci
2007;9(1):48-53.
[9] Baigelman W, Katz R, Geary G. Patient readmission to critical care
units during the same hospitalization at a community teaching hospital.
Intensive Care Med 1983;9(5):253-6.
[10] Snow N, Bergin KT, Horrigan TP. Readmission of patients to the
surgical intensive care unit: patient profiles and possibilities for
prevention. Crit Care Med 1985;13(11):961-4.
[11] Guidelines for intensive care unit admission, discharge, and triage.
Task Force of the American College of Critical Care Medicine, Society
of Critical Care Medicine. Crit Care Med 1999;27(3):633-8.
[12] Franklin C, Jackson D. Discharge decision-making in a medical ICU:
characteristics of unexpected readmissions. Crit Care Med 1983;11(2):
61-6.
[13] Fernandez R, Baigorri F, Navarro G, Artigas A. A modified McCabe
score for stratification of patients after intensive care unit discharge:
the Sabadell score. Crit Care 2006;10(6):R179.
[14] Iyengar A, Baxter A, Forster AJ. Using medical emergency teams to
detect preventable adverse events. Crit Care 2009;13(4):R126.
[15] Gajic O, Malinchoc M, Comfere TB, et al. The Stability and Workload
Index for Transfer score predicts unplanned intensive care unit patient
Short-term death or readmission after ICU discharge
readmission: initial development and validation. Crit Care Med
2008;36(3):676-82.
[16] Azoulay E, Adrie C, De Lassence A, et al. Determinants of
postintensive care unit mortality: a prospective multicenter study.
Crit Care Med 2003;31(2):428-32.
[17] McCabe WR, Jackson GG. Gram negative bacteremia: etiology and
ecology. Arch Intern Med 1962;110:847-55.
[18] Le Gall JR, Lemeshow S, Saulnier F. A new Simplified Acute
Physiology Score (SAPS II) based on a European/North American
multicenter study. JAMA 1993;270(24):2957-63.
[19] Le Gall JR, Klar J, Lemeshow S, et al. The logistic organ dysfunction
system. A new way to assess organ dysfunction in the intensive care
unit. ICU Scoring Group. JAMA 1996;276(10):802-10.
[20] Knaus WA, Draper EA, Wagner DP, Zimmerman JE. APACHE II: a
severity of disease classification system. Crit Care Med 1985;13(10):
818-29.
[21] Vincent JL, Moreno R, Takala J, et al. The SOFA (Sepsis-related
Organ Failure Assessment) score to describe organ dysfunction/failure.
On behalf of the Working Group on Sepsis-Related Problems of the
European Society of Intensive Care Medicine. Intensive Care Med
1996;22(7):707-10.
422.e9
[22] Reis Miranda D, Moreno R, Iapichino G. Nine equivalents of nursing
manpower use score (NEMS). Intensive Care Med 1997;23(7):760-5.
[23] Priestap FA, Martin CM. Impact of intensive care unit discharge time
on patient outcome. Crit Care Med 2006;34(12):2946-51.
[24] McNeil BJ, Hanley JA. Statistical approaches to the analysis of
receiver operating characteristic (ROC) curves. Med Decis Making
1984;4(2):137-50.
[25] Frost SA, Alexandrou E, Bogdanovski T, et al. Severity of illness and
risk of readmission to intensive care: a meta-analysis. Resuscitation
2009;80(5):505-10.
[26] Daly K, Beale R, Chang RW. Reduction in mortality after
inappropriate early discharge from intensive care unit: logistic
regression triage model. BMJ 2001;322(7297):1274-6.
[27] Levy MF, Greene L, Ramsay MA, et al. Readmission to the intensive
care unit after liver transplantation. Crit Care Med 2001;29(1):18-24.
[28] Hutchings A, Durand MA, Grieve R, et al. Evaluation of moderni-
sation of adult critical care services in England: time series and cost
effectiveness analysis. BMJ 2009;339:b4353.
[29] Wunsch H, Angus DC, Harrison DA, et al. Variation in critical care
services across North America and Western Europe. Crit Care Med
2008;36(10):2787-93, e1-9.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.