Outcome of patients undergoing prolonged mechanical ventilation

after critical illness*

Luca M. Bigatello, MD; Henry Thomas Stelfox, MD, PhD; Lorenzo Berra, MD; Ulrich Schmidt, MD, PhD;
Elise M. Gettings, RN

Objective: To examine the longitudinal outcome of a cohort of
mechanically ventilated patients admitted to an acute care respi-
ratory unit after critical illness.
Design, Setting, and Patients: Prospective, observational study
of 210 consecutive patients admitted to a respiratory unit of an
acute, tertiary care university hospital, who had an acute critical
illness with respiratory failure. The study was powered to develop
multivariate regression models to investigate the relationship
between patient characteristics and a) liberation from mechanical
ventilation and b) survival.
Interventions: None.
Measurements and Main Results: The median time to liberation
from mechanical ventilation after respiratory unit admission was 14
days (interquartile range, 6 –51). A total of 146 patients (69%) were
off mechanical ventilation at 6 months, and 123 patients (61%) were
alive at 1 yr. Patients who did not come off mechanical ventilation in
the respiratory unit were seven times more likely to die within a year
than those who did (odds ratio, 6.55; 95% confidence intervals,
4.04 –10.63; p < .001). At least 75% of deaths occurred by consen-
sual withdrawal of life support. Patient activity of daily living scores
(0 –100 scale) increased progressively from hospital discharge (24 ⴞ
6) through 3 (54 ⴞ 21) and 6 months (64 ⴞ 22) (p < .001). The
median cost of hospitalization for all study patients was $149,624
(interquartile range, $102,540 –225,843).
Conclusions: The majority of patients requiring prolonged me-
chanical ventilation in a respiratory unit after acute critical illness
are liberated from mechanical ventilation, survive, and have a
steady improvement in the activity of daily living during the first
6 months after discharge. However, a substantial fraction of these
patients does not wean from mechanical ventilation and dies from
consensual withdrawal of life support after a prolonged and
costly hospital stay. (Crit Care Med 2007; 35:2491–2497)
KEY WORDS: prolonged mechanical ventilation; prolonged criti-
cal illness; acute respiratory failure; weaning from mechanical
ventilation

A cute respiratory failure requir-
ing mechanical ventilation ac-
counts for approximately 30%
of admissions to intensive care
units (ICUs) (1). Early survival from
acute respiratory failure in the ICU has
increased during the past two decades
(2– 4), and a growing number of patients
receive mechanical ventilation for pro-
longed periods of time. These patients
*See also p. 2640.
From the Department of Anesthesia and Critical
Care, Massachusetts General Hospital, Harvard Medi-
cal School, Boston, MA (LMB, LB, US, EMG); and the
Department of Critical Care Medicine, University of
Calgary, Alberta, Canada (HTS).
The authors have not disclosed any potential con-
flicts of interest.
Supported, in part, through the Reginald and Mar-
gareta Jenney Fund for Anesthesia and Critical Care
Research of the Department of Anesthesia and Critical
Care of the Massachusetts General Hospital.
For information regarding this article, E-mail:
Lbigatello@partners.org
Copyright © 2007 by the Society of Critical Care
Medicine and Lippincott Williams & Wilkins
DOI: 10.1097/01.CCM.0000287589.16724.B2
often require protracted stay in acute
care hospitals because they necessitate
complex medical treatment and special-
ized nursing care not available in many
rehabilitation and long-term care facili-
ties (5– 8).
In many institutions, patients with
acute respiratory failure who are hemo-
dynamically and metabolically stable are
transferred from the ICU to units that are
specialized in completing the process of
weaning from mechanical ventilation,
optimizing the patients’ medical status,
and organizing their transfer to rehabili-
tation facilities (6, 8). However, these pa-
tients do not uniformly fare well, and
they may experience complications re-
quiring readmission to an ICU, reinstitu-
tion of various means of life support, and
further prolongation of mechanical ven-
tilation (9). Similarly, mechanically ven-
tilated patients transferred from an ICU
directly to a lower acuity facility had a
high rate of hospital readmission and a
high mortality rate (10).
Therefore, it is important to identify
the factors that are associated with pro-
longed mechanical ventilation and mor-
tality in this patient population. In a re-
view of 143 consecutive patients admitted
to our acute care respiratory unit (RU) in
its first year of activity (May 2001 to May
2002), we observed that patients who did
not wean from the ventilator while in the
RU (33%) had a high hospital mortality
rate (25%) (11). Hence, we designed a
prospective observational study to exam-
ine the long-term outcomes of patients
undergoing prolonged mechanical venti-
lation for acute respiratory failure.
MATERIALS AND METHODS
Study Questions. We asked six specific
questions:
1. What proportion of patients receiving
prolonged mechanical ventilation in
an acute care RU are liberated from
mechanical ventilation?
2. What patient characteristics are asso-
ciated with successful liberation from
mechanical ventilation?
3. What proportion of patients receiving
prolonged mechanical ventilation in

Crit Care Med 2007 Vol. 35, No. 11 2491

 an acute care RU are alive 1 yr after
admission?
4. What patient characteristics are asso-
ciated with 1-yr mortality?
5. What is the quality of life of these
patients during the first 6 months af-
ter discharge?
6. What are the direct medical costs of
providing care to patients who receive
prolonged mechanical ventilation in
an acute care RU?
Study Setting and Patients. The RU at the
Massachusetts General Hospital is a ten-bed
unit dedicated to the care of mechanically
ventilated patients who are hemodynamically
and metabolically stable. The great majority of
patients are admitted from one of the hospi-
tal’s ICUs and mechanically ventilated at the
time of admission. The medical staff includes
two co-directors and several staff intensivists
from the Departments of Anesthesia and Pul-
monary Medicine. Daily patients rounds are
performed by a team comprising a staff inten-
sivist, a critical care fellow, and a nurse prac-
titioner. Off-hours medical coverage is pro-
vided on site by a staff intensivist or critical
care fellow. The RU also has at least one ded-
icated registered respiratory therapist on site
at all times and a dedicated nursing staff that
includes a nurse manager and a clinical nurse
specialist. Nurses are assigned one or two pa-
tients, depending on patient acuity. Other ser-
vices providing integral contribution to daily
patient care include: pharmacy, physical ther-
apy, occupational therapy, clinical nutrition,
speech and language pathology, social ser-
vices, and case management. A dedicated phy-
sician from the physical medicine and rehabil-
itation service of an affiliated rehabilitation
institution is consulted in the majority of pa-
tients. These services meet with the RU nurs-
ing and medical staff during a weekly interdis-
ciplinary conference to discuss ongoing issues
and long-term plans.
The focus on our RU is to liberate patients
from artificial ventilation. Specific protocols
have been developed over the years, including
guidelines addressing weaning of mechanical
ventilation, downsizing tracheostomy tubes
for speech and oral feeding, and decannulating
the tracheostomy site. Since the inception of
the RU in 2001, the great majority of our
patients are ventilated with pressure-support
ventilation and undergo daily spontaneous
breathing trials.
The study was approved by the Human
Research Committee of the Massachusetts
General Hospital. We prospectively identified
consecutive adults undergoing mechanical
ventilation at the time of their admission to
the acute care RU at the Massachusetts Gen-
eral Hospital between January 2002 and March
2004. All patients who were receiving mechan-
ical ventilation through an endotracheal tube
or a tracheostomy tube were eligible for en-
rollment. Among patients who were receiving
noninvasive ventilation, we included in the
2492
study those with an acute exacerbation of their
underlying disease (e.g., chronic obstructive
pulmonary disease) and excluded those who
were on a stable regimen of home noninvasive
ventilation for conditions such as sleep apnea.
Patients who were not mechanically ventilated
were excluded. Written informed consent to
perform telephone follow-up was obtained
from the patient or the next of kin.
Each patient’s chart was reviewed for de-
mographic and clinical data at the time of
admission to the RU and subsequently during
their hospital stay to follow in-hospital dispo-
sition and outcome. For patients who had
multiple admissions to our RU, we computed
their RU length of stay as the sum of the days
of each admission; similarly, we computed
their hospital length of stay and days of me-
chanical ventilation before admission to the
RU as the sum of all the days that preceded RU
admission. After hospital discharge, we con-
tacted the consenting patients by telephone at
3 and 6 months postdischarge to determine
survival, disposition, and level of activity of
daily living (ADL) (12). We used an ADL scale
that ranged from 10 (moribund) to 100 (nor-
mal) in 10-point increments. For example, pa-
tients unable to care for themselves, or requir-
ing institutionalization, would receive scores
between 10 and 40, whereas patients with bet-
ter functional status, who could provide per-
sonal care at home but were still unable to
work, would receive scores between 50 and 70.
The Social Security Death Index was searched
to determine patients’ vital status 12 months
after RU admission (http://ssdi.genealogy.
rootsweb.com/).
Definitions. Liberation from mechanical
ventilation was defined as the ability to come
off ventilatory support within the 6 months of
the study follow-up period.
Comorbidities were defined prospectively
and extracted from the patient’s medical
record at the time of admission to the RU.
Respiratory comorbidity was defined as docu-
mented chronic obstructive pulmonary dis-
ease (COPD), asthma, or other symptomatic
chronic respiratory disorder. Cardiac comor-
bidity was defined as documented coronary
artery disease, congestive heart failure, valvu-
lar heart disease, or a significant arrhythmia-
atrial fibrillation and any ventricular arrhyth-
mia other than isolated premature ventricular
contractions. Chronic renal insufficiency was
defined by a serum creatinine concentration of
⬎1.3 mg/dL at admission to the hospital. Liver
insufficiency was defined as known liver cir-
rhosis or chronic hepatitis. Neurologic comor-
bidity was defined as a history of stroke, sub-
arachnoidal hemorrhage, or traumatic brain
injury (recent or remote, with persistent neu-
rologic deficit). Neuromuscular comorbidity
included all neuromuscular disorders (e.g.,
amyotrophic lateral sclerosis, multiple sclero-
sis) and paraplegia. Endocrine and metabolic
comorbidity included insulin-dependent dia-
betes mellitus, other endocrine disorders re-
quiring treatment, and morbid obesity. Also
recorded were the presence of active malig-
nancy and the recent use of drugs of addiction
or alcohol abuse.
Pathogenesis of acute respiratory failure
was also extracted from the medical record at
the time of admission to the RU and classified
in three categories: a) acute lung injury/acute
respiratory distress syndrome (13) from vari-
ous causes, such as pneumonia, sepsis, and
trauma, in patients without COPD; b) COPD:
acute respiratory failure from various causes
in patients with known COPD; and c) neuro-
logic and neuromuscular: acute respiratory
failure in patients with a predominant neuro-
muscular problem, such as a stroke causing
inability to protect the airway, Guillain-Barre
syndrome, severe polyneuropathy of critical
illness, a high spinal cord injury, and others.
The remaining causes of respiratory failure
were classified as miscellaneous (e.g., pulmo-
nary embolus, pneumonia without acute lung
injury/acute respiratory distress syndrome,
cardiac failure).
A medical admitting diagnosis identified
all patients coming from the medical ICU, or
followed by the medical ICU team in other
units, and patients with a neurologic injury
that was not secondary to trauma and who had
not had an operation. A surgical admitting
diagnosis identified all postoperative and
trauma patients.
Statistical Analysis. The strategy for the
primary analysis was to answer each of the six
specific study questions. We estimated that
200 patients would need to be enrolled for the
study to have a sufficient number of deaths to
allow development of stable regression models
with up to five independent variables. Our
calculations assumed that 33% of patients
would not be weaned from mechanical venti-
lation, 25% would die during their hospital-
ization, and that 10% would be lost to fol-
low-up after discharge.
Patient characteristics and outcomes were
compared using Student’s t-tests, Kruskal-
Wallis rank tests, chi-square tests, and Fisher’s
exact tests for outcomes with rare events. Lo-
gistic and Cox proportional hazards models
were developed to determine the relationship
between patient characteristics and both lib-
eration from mechanical ventilation and
death, respectively. Variables were first exam-
ined using univariate analyses. Variables that
were significant at a p value of ⱕ.20 were
included in the multivariable analyses. Vari-
ables were selected by means of backward
stepwise regression and comparison of the re-
gression sum of squares. A partial F-test was
used to compare linear models with quadratic
models. Statistical analyses were performed
using Stata (version 8.2, Stata Corp, College
Station, TX) with two-tailed significance levels
of .05.
RESULTS
We enrolled 210 consecutive mechan-
ically ventilated patients admitted to our
Crit Care Med 2007 Vol. 35, No. 11
Table 1. Characteristics of patients at admission to the respiratory unit (RU) telephone interview, and three were lost
after the 3-month interview.
Characteristic Patients (n ⫽ 210)
Descriptive statistics of the study pop-
Age in years, median (IQR) 66 (52–75) ulation are shown in Table 1. The most
Female sex 90 (43) common comorbidities were cardiac
Comorbidities (52%) and respiratory (48%). Of the ad-
Respiratory 100 (48)
Cardiovascular 109 (52)
missions, 56% were medical and 44%
Renal 16 (8) surgical. The median duration of me-
Neurologic 51 (24) chanical ventilation before admission to
Neuromuscular 20 (9) the RU was 21 days (interquartile range
Metabolic/endocrine 66 (31)
Malignancy 31 (15) [IQR], 12–32 days); 40% of the patients had
Substance abuse 25 (12) at least one documented failed tracheal ex-
Admitting diagnosis tubation before admission to the RU.
Medical 117 (56)
Surgical 92 (44)
Within 6 months of discharge from
Predominant pathogenesis of respiratory failure the RU, 146 patients (69%) were free
Neurologic and neuromuscular 36 (17) from mechanical ventilation. The major-
Acute lung injury/ARDS 113 (54) ity of these patients (n ⫽ 136) weaned in
Chronic obstructive pulmonary disease 52 (25)
Miscellaneous 9 (4) the RU within a median time of 14 days
Length of hospitalization before RU admission, days, median (IQR) 23 (13–38) (IQR, 6 –51 days) (Fig. 1). Nine additional
Days of ventilator use before RU admission, median (IQR) 21 (12–32) patients (out of 18 patients discharged
Glasgow Coma Scale, median (IQR) 11 (10–12)
APACHE II score, median (IQR) 14 (11–17) while still on the ventilator after a median
Documented failed extubation before RU admission of 32.5 days in the RU [IQR, 19 – 47]) were
0 125 (59) weaned in rehabilitation institutions. Inde-
1 62 (29)
⬎1 23 (11)
pendent predictors of weaning failure were
Ventilatory settings a preexistent respiratory comorbidity (odds
Driving pressure, cm H2O, median (IQR) 10 (10–15) ratio, 0.50; p ⬍ .05) and readmissions to
PEEP, cm H2O, median (IQR) 5 (5–7) the RU (odds ratio, 0.36; p ⬍ .05) ( Table 2).
FIO2, cm H2O, median (IQR) 0.40 (0.40–0.45)
Independent predictors of weaning success
IQR, interquartile range; ARDS, acute respiratory distress syndrome; APACHE, Acute Physiology were a neurologic comorbidity (odds ratio,
and Chronic Health Evaluation; PEEP, positive end-expiratory pressure. 3.84; p ⬍ .01) and unsuccessful extubation
Results are expressed as number (percentage) unless otherwise indicated. attempts before admission to the RU (odds
ratio, 2.67; p ⬍ .01).
After discharge from the RU, 123 pa-
tients (61%) were alive at 12 months
.8

(Fig. 2). A total of 60 patients (75% of the

deaths) died after consensual withdrawal
Proportion of Patients Weaned

of support, and seven (9%) died of an

.6

acute cause; in 11 patients who died after

hospital discharge, we were unable to ob-
tain the appropriate information. Inde-
.4

pendent predictors of mortality during

the 365 days after RU admission (Table 3)
included an older age (increments of 10
.2

yrs; odds ratio, 1.31; p ⬍ .001), pathogen-

esis of acute respiratory failure of acute
lung injury/acute respiratory distress syn-
0

0 20 40 60 80 100 drome (odds ratio, 2.55; p ⬍ .05) or COPD

Days After Admission To Acute Care Respiratory Unit
(odds ratio, 3.60; p ⬍ .01) vs. primary neu-
Figure 1. Proportion of patients liberated from mechanical ventilation during admission to our acute romuscular, and the inability to wean from
care respiratory unit after critical illness. Filled circles, actual data points, connected by a solid line; mechanical ventilation (odds ratio, 6.55;
dashed lines, 95% confidence interval. p ⬍ .001).
The median hospital length of stay for
the overall study population was 49 days
RU during a 26-month period. There transfer from another institution. A total (IQR, 33–72 days) and increased to 58
were 27 patients admitted to the RU dur- of 165 (79%) patients were ventilated days (IQR, 32–72 days) for the patients
ing this time who were excluded because through a tracheostomy at admission; 40 who did not wean from mechanical ven-
they either were not receiving ventilatory patients (19%) were ventilated through tilation. A total of 180 patients (86%)
support (n ⫽ 25) or were on a stable an orotracheal tube, and 19 of them re- were discharged alive from the RU, with
schedule of home noninvasive ventilation ceived a tracheostomy during their RU an average ADL of 24 ⫾ 6 (Table 4). There
(n ⫽ 2). There were 196 patients (93%) stay; five patients (3%) were ventilated were 80 patients (38%) discharged di-
admitted from one of the hospital’s ICUs via a face mask. Nine patients had incom- rectly to a rehabilitation institution (n ⫽
and 14 (7%) from a general ward or in plete follow-up: six did not consent to the 77) or home (n ⫽ 3), and 108 patients

Crit Care Med 2007 Vol. 35, No. 11 2493

Table 2. Univariate and multivariate predictors of liberation from mechanical ventilation before ity rate), 49 patients (23%) were in the
discharge from the respiratory unit (RU)a hospital or in a rehabilitation institution,
80 (38%) were at home, and the overall
Univariate Odds Multivariatea Odds
ADL was 54 ⫾ 21. At 6 months after dis-
Variable Ratio (95% CI) [p Value] Ratio (95% CI) [p Value]
charge, nine additional patients had died
Age (increments of 10 yrs) 0.99 (0.84, 1.17) [.925] (40% mortality rate), 20 patients (10%)
Female sex 0.76 (0.43, 1.34) [.338] were in the hospital or in a rehabilitation
Admitting diagnosis, medical (vs. surgical) 0.45 (0.25, 0.81) [.008] 0.55 (0.28, 1.09) [.087] institution, 100 patients (48%) were at
Prehospital comorbidities home, and the overall ADL was 64 ⫾ 23
Respiratory 0.44 (0.24, 0.78) [.005] 0.50 (0.26, 0.96) [.037]
Cardiovascular 1.12 (0.64, 1.98) [.684] (p ⬍ .001 for trend from admission).
Renal 0.68 (0.24, 1.90) [.461] The median cost of hospitalization for
Neurologic 3.81 (1.68, 8.64) [.001] 3.52 (1.48, 8.37) [.004] all study patients was $149,624 (IQR,
Neuromuscular 1.01 (0.38, 2.66) [.981] $102,540 –225,843). Patients who weaned
Metabolic/endocrine 1.02 (0.56, 1.89) [.936]
Malignancy 1.68 (0.71, 3.97) [.237]
from mechanical ventilation in the RU
Substance abuse 1.84 (0.70, 4.83) [.215] exhibited slightly lower costs ($144,566;
Predominant pathogenesis of respiratory IQR, $100,637–217,772) than those who
failure failed to wean ($187,806; IQR, $104,440 –
Neurologic and neuromuscular 1.00 (reference group) 255,694; p ⫽ .25).
Acute lung injury/ARDS 1.81 (0.83, 3.94) [.136]
Chronic obstructive lung disease 0.71 (0.30, 1.68) [.442]
Miscellaneous 5.71 (0.64, 50.65) [.117] DISCUSSION
Length of hospitalization before RU 1.03 (0.94, 1.13) [.485]
admission (increments of 7 days) The most important finding of our
Days of ventilator use before RU 0.94 (0.86, 1.03) [.194] study is that a substantial portion of pa-
admission (increments of 7 days) tients admitted to an acute care RU does
Glasgow Coma Scale (increments of 1 unit) 0.96 (0.86, 1.07) [.475]
APACHE II score (increments of 5 units) 0.93 (0.68, 1.27) [.625] not wean from the ventilator and has a
Failed extubation attempt(s) 1.86 (1.03, 3.37) [.042] 2.52 (1.27, 5.01) [.008] high mortality rate. Death usually occurs
Ventilatory settings by consensual withdrawal of care after a
Driving pressure (increments of 5 cm H2O) 0.76 (0.57, 1.01) [.055] prolonged hospital stay.
PEEP (increments of 1 cm H2O) 1.02 (0.90, 1.17) [.721]
FIO2 (increments of 0.10) 0.92 (0.70, 1.20) [.542]
Specialized units that admit mechan-
RU readmission(s) 0.41 (0.22, 0.75) [.004] 0.36 (0.14, 0.90) [.029] ically ventilated patients after a critical
illness report ventilator weaning rates of
CI, confidence interval; ARDS, acute respiratory distress syndrome; APACHE, Acute Physiology and 40 – 60% and hospital survival rates of
Chronic Health Evaluation; PEEP, positive end-expiratory pressure. 50 –90% (6, 7, 9, 14). This variability is in
a
Odds ratios of ⬍1.0 indicate that the variable is negatively associated with liberation from part due to the different administrative
mechanical ventilation; odds ratios of ⬎1.0 indicate variable is positively associated with liberation and clinical makeup of each unit, which
from mechanical ventilation.
may be dedicated mainly to an inpatient
population (like our RU) or function as a
regional referral center (8, 15). Just as
1

important, this variability is also due to

Proportion of Patients Surviving

the lack of consensus on definitions, in-

.75

cluding what constitutes weaning success

and what constitutes prolonged mechan-
ical ventilation. Although weaning suc-
.5

cess has been given a number of time-

based definitions, from 48 hrs to 1 wk
(15, 16), we elected to score as success
.25

only the patients who remained free from

mechanical ventilation within the time
limit of our study (6 months); a similar
0

0 100 200 300 400 philosophy was used in the most recent
Days After Admission To Acute Care Respiratory Unit and comprehensive observational study
Figure 2. One-year Kaplan-Meier survival curve of the 210 patients admitted to our acute care on prolonged mechanical ventilation to
respiratory unit during the 26-month enrollment period. Dashed lines, 95% confidence intervals. date (14). Although prolonged mechani-
cal ventilation was recently defined as
⬎21 days of mechanical ventilation (16),
(51%) were admitted to other hospital the RU, 18 of them died (58% mortality most available studies examine, like ours,
units—short line 34 to an ICU and 74 to rate) either in the RU (ten patients) or in unit-based rather than time-based patient
a general ward. Eight of the 108 patients other hospital units (eight patients). The populations, resulting, like ours, in the
were counted twice because they were median hospital length of stay of these 31 inclusion of patients ventilated for ⬍21
initially discharged from the RU and later patients was 111 days (IQR, 71–137 days). days (5, 7, 9, 17). However, analyzing a
readmitted and died. Of the 31 patients At 3 months after RU discharge, an ad- unit-based population also has advantages.
(17%) who had at least two admissions to ditional 45 patients had died (36% mortal- First, it reflects clinical practice, and sec-

2494 Crit Care Med 2007 Vol. 35, No. 11

Table 3. Univariate and multivariate predictors of mortality
Variable
Age (increments of 10 yrs)
Female sex
Admitting diagnosis, medical (vs. surgical)
Prehospital comorbidities
Respiratory
Cardiovascular
Renal
Neurologic
Neuromuscular
Metabolic/endocrine
Malignancy
Substance abuse
Pathogenesis of respiratory failure
Neuromuscular
ARDS vs. neuromuscular
Obstructive lung disease vs. neuromuscular
Miscellaneous
Length of hospitalization before RU
admission (increments of 7 days)
Glasgow Coma Scale (increments of 1 unit)
APACHE II score (increments of 5 units)
Failed extubation attempt(s)
Ventilatory settings
Driving pressure (increments of 5 cm H2O)
PEEP (increments of 1 cm H2O)
FIO2 (increments of 0.10)
Days on ventilator before RU
admission (increments of 7 days)
Unable to wean from mechanical ventilation
RU readmissions
CI, confidence interval; ARDS, acute respiratory distress syndrome; RU, respiratory unit; APACHE,
Acute Physiology and Chronic Health Evaluation; PEEP, positive end-expiratory pressure.
Hazard ratios ⬎1.0 indicate that the variable is associated with an increased risk of mortality.
ond, it provides more homogeneous study
populations (common triage criteria, diag-
nostic procedures, and weaning protocols)
that are highly suitable for observational
studies.
Our rate of discontinuation of mechan-
ical ventilation— 69% at 6 months—is
within the range reported by others (5–7,
14, 15, 17–19). Our observed predictors of
failure to discontinue mechanical ventila-
tion—a medical diagnosis, COPD, and mul-
tiple unit admissions (Table 2)—are in
agreement with previous observations (6,
9). Our predictors of successful weaning—a
neurologic comorbidity and failing tracheal
extubation before admission to the RU—
are likely to be indicators, respectively, of a
less severe pulmonary injury and of a more
advanced weaning process.
Of the patients who weaned from me-
chanical ventilation in the RU, 65% did
so within 10 days of admission and 10%
within 2 days (Fig. 1). This temporal pat-
tern has notable implications. First, it
suggests that a number of patients were
ready to come off mechanical ventilation
at the time of admission to the RU, im-
Crit Care Med 2007 Vol. 35, No. 11
Univariate Hazard Multivariate Hazard
Ratio (95% CI) [p Value] Ratio (95% CI) [p Value]
1.35 (1.16, 1.57) [⬍.001] 1.31 (1.12, 1.52) [⬍.001]
1.29 (0.84, 1.97) [.243]
1.70 (1.08, 2.67) [.021] 1.50 (0.95, 2.38) [.083]
1.35 (0.88, 2.07) [.167]
1.05 (0.68, 1.60) [.836]
1.89 (0.98, 3.67) [.058]
0.86 (0.52, 1.43) [.559]
0.51 (0.21, 1.26) [.143]
0.76 (0.47, 1.22) [.254]
1.20 (0.67, 2.12) [.540]
1.02 (0.53, 1.97) [.960]
1.00 (reference group) 1.00 (reference group)
1.85 (0.87, 3.94) [.113] 2.55 (1.17, 5.56) [.019]
3.87 (1.78, 8.39) [.001] 3.60 (1.63, 7.95) [.002]
1.65 (0.44, 6.21) [.460] 4.05 (1.01, 16.22) [.048]
0.97 (0.90, 1.04) [.351]
1.01 (0.95, 1.13) [.443]
1.39 (1.10, 1.76) [.006]
0.95 (0.61, 1.47) [.817]
1.12 (0.92, 1.37) [.270]
0.93 (0.83, 1.03) [.165]
1.16 (0.96, 1.41) [.118] 1.19 (0.96, 1.46) [.105]
1.02 (0.95, 1.08) [.632]
5.89 (3.76, 9.24) [⬍.001] 6.55 (4.04, 10.63) [.001]
1.35 (0.96, 1.89) [.085]
plying that in the preceding ICU stay,
discontinuing mechanical ventilation was
not always a priority. Despite the evi-
dence that daily performance of a spon-
taneous breathing trial may decrease the
duration of mechanical ventilation (20),
this practice is still not consistently car-
ried out in ICUs (1). Conversely, in a
dedicated RU, the focus of care shifts
from the life-threatening issues charac-
teristic of ICU care to ventilatory man-
agement and progress toward rehabilita-
tion. Second, it is unlikely that the sharp
demarcation of the weaning rate around
the fourth to fifth week of RU stay (Fig. 1)
reflects a purely physiologic phenome-
non. Rather, as patients stay longer in an
RU, additional factors, including social
and administrative factors, may influence
patient disposition (21–23). Thus, the
overall duration of mechanical ventila-
tion of our patient population was signif-
icantly affected by nonphysiologic factors
at both ends of the RU stay. This obser-
vation offers a different perspective to the
important question of how long should
weaning attempts be continued in pa-
tients undergoing prolonged mechanical
ventilation (16). Rather than recom-
mending a specific time frame, which will
inevitably prove wrong in a number of
patients (of the 146 patients who were
liberated from mechanical ventilation
within the 6-month study period, at least
eight were ventilated for ⬎100 days), we
suggest that an optimal ventilator wean-
ing strategy should be implemented in-
dependently of the previous duration of
mechanical ventilation and should always
include two fundamental considerations.
First, does the patient need to be on a
ventilator? Our data show that a number
of patients may wean rapidly once venti-
latory management becomes the focus of
care in a specialized RU and confirm a
similar observation by Vitacca et al. (15)
in tracheostomized COPD patients. Sec-
ond, as the duration of mechanical ven-
tilation extends in time, will the patient
ever overcome his or her barriers to
weaning? Our data suggest that, once the
focus on care moved to ventilator wean-
ing and rehabilitation in a specialized RU,
a decision of disposition (including the
withdrawal of vital support, see below)
happened within a reproducible time. It
is reasonable to construe that if realistic
care plans were formulated with patients
and families early in the stay, time in the
RU would be individualized, length of
stay may decrease, and patient satisfac-
tion may increase (21, 24).
Our 6-month survival rate (59%) is
within the range reported by others (5, 7,
14, 18, 19). Independent determinants of
death (Table 3) included advanced age, a
medical admitting diagnosis, a primary
pulmonary pathogenesis of respiratory
failure, and by far the strongest predictor,
the inability to wean from mechanical
ventilation. Although similar findings
have been reported (18, 19), our analysis
discloses a remarkable correlation be-
tween the failure to wean from mechan-
ical ventilation and death: the 74 patients
who did not wean during the study period
had a seven-fold likelihood of death (Fig.
3), either in the RU (29 patients) or
shortly thereafter (22 patients). In con-
trast to other studies showing a steady
decrease in the number of surviving pa-
tients over time (5, 9, 10, 17, 18), our
death rate reaches a plateau at the
3-month follow-up, with minimal in-
crease at 6 and 12 months. Accordingly,
ⱖ75% of the study deaths (60 patients)
occurred after consensual withdrawal of
life support. The importance of this ob-
2495
Table 4. Disposition of patients admitted to the respiratory unit (RU) at discharge, 3-month follow-up, 31 patients who had multiple RU admis-
and 6-month of follow-upa sions, ten died in the RU (included in the
30 RU deaths in Table 4) and an addi-
All Patients Weaned Nonweaned
tional eight after discharge from the RU.
Time and Patient Location (n ⫽ 210) (n ⫽ 136) (n ⫽ 74)
Hence, a significant number of patients
Discharge from RU 180 135 45 moved from one hospital unit to another
Death 30 (14) 1 29 (39) hospital unit before eventually dying.
Rehabilitation hospital 77 (37) 61 (45) 16 (22) These patients would be counted as sur-
Home 3 (1) 1 (1) 2 (3) vivors in each of these units before the
Acute care hospital 108b (51) 72c (53) 36d (49)
ICU 34 (16) 7 (5) 27 (36) one in which they eventually died, thus
Ward 74 (35) 65 (48) 9 (12) diluting the mortality rate of each indi-
Multiple RU admissions 31 (17) 11 (8) 20 (27) vidual ward. However, when followed lon-
ADL score, mean ⫾ SD 24 ⫾ 6 25 ⫾ 6 20 ⫾ 4 gitudinally, the 31 patients who were ad-
3 mos after discharge from RU 129 108e 21f
Death 75 (36) 23 (17) 51 (69)
mitted at least twice to the RU had a
Rehabilitation hospital 40 (19) 31 (23) 9 (12) median hospital length of stay of 111 days
Home 80 (38) 68 (50) 12 (16) and a 3-month mortality rate of ⱖ58%
Acute care hospital 9 (4) 9 (7) 0 (0) (seven patients were lost to follow-up af-
ADL score, mean ⫾ SD 54 ⫾ 21 56 ⫾ 21 48 ⫾ 18 ter hospital discharge), most of which
6 mos after discharge from RU 120 102g 18
Death 84 (40) 29 (21) 55 (74) occurred by consensual withdrawal of
Rehabilitation hospital 17 (8) 16 (12) 1 (1) support. In a similar analysis (data not
Home 100 (48) 81 (60) 17 (23) shown) of the 34 patients who were trans-
Acute care hospital 3 (1) 3 (2) 0 (0) ferred at least once from the RU to an
ADL score, mean ⫾ SD 64 ⫾ 22 65 ⫾ 22 61 ⫾ 22
ICU, 20 (59%) were dead at 3 months.
ICU, intensive care unit; ADL, activity of daily living. This is a conservative estimate of the size
a
Results are expressed as number (percentage) unless otherwise indicated; beight patients are of this particular patient population be-
counted twice: they were discharged alive from the RU and were later readmitted; cone weaned cause it does not include: a) patients who
patient discharged from the RU was later readmitted on mechanical ventilation and did not wean; this were admitted to multiple units without
patient is counted in the nonweaned group; dnine nonweaned patients are counted twice: they were returning to the RU and b) time spent at
discharged from the RU and were later readmitted; efour patients lost to follow-up; ftwo patients other institutions between successive ad-
lost to follow-up; gone patient lost to follow-up. missions to the RU or after discharge
from the RU. We were able to collect
length-of-stay data from one rehabilita-
1.00

tion institution: the ten patients who died

Weaned
at that institution after discharge from
Proportion of Patients Surviving

our RU spent an additional median of 63

0.75

days there before dying.

There are several limitations in this
study. First, our data were collected in
0.50

Failed To Wean one RU only and cannot be universally

applied to all patients undergoing pro-
0.25

longed mechanical ventilation. However,

as the reports on specialized units like
ours increase (5–7, 9, 14, 15, 18), it seems
0.00

that, though the absolute outcome rates

0 100 200 300 400
Days After Admission To Acute Care Respiratory Unit
Figure 3. One-year Kaplan-Meier survival curves of the study patient population according to weaning
status.
servation goes beyond the fact that pro- Of the 180 patients discharged from
longed mechanical ventilation is a the RU during the study period, 108 were
marker and a cause of increased morbid- transferred to another in-hospital unit
ity and mortality. In addition to that, our (Table 4). Although some of these pa-
findings draw attention to a population of tients were considered ready for rehabil-
mechanically ventilated patients who itation at the time of transfer to a general
were in the hospital for a prolonged pe- care ward, their average ADL was low
riod of time, often in multiple hospital (Table 4), indicating that this was still a
units, probably at high emotional and needy patient population, as it has been
financial costs, and who eventually died, shown for patients who survive severe
generally by consensual withdrawal of acute respiratory failure, such as acute
support. respiratory distress syndrome (25). Of the
may vary, the physiologic, social, and eth-
ical challenges faced in caring for this
patient population are consistent. Sec-
ond, it would have been helpful to have
access to additional data on length of stay
and cause of death after hospital dis-
charge. It would have also been interest-
ing to gather additional information
during the telephone interview, specif-
ically, information regarding patients
and family satisfaction, end-of-life
wishes, and financial effect of their ill-
ness.
CONCLUSIONS
A hospital-based, acute care RU can
provide effective care to mechanically

2496 Crit Care Med 2007 Vol. 35, No. 11

ventilated patients recovering from criti-
cal illness. Six months after discharge,
50% of our study patients were at home,
with a reasonable average ADL. Once ad-
mitted to an RU, most patients who are
going to wean from mechanical ventila-
tion do so within a reproducible time, in
our case, 4 –5 wks. Patients who do not
wean within that time frame have a very
high mortality. Most of them die of con-
sensual withdrawal of support after pro-
longed stay in multiple hospital units and
considerable cost.
These data call to implement studies,
procedures, and practices that promote
early care planning with patients under-
going prolonged mechanical ventilation.
Specialized RU teams should assist pa-
tients and families in understanding the
implications of their prolonged illness
and matching it with their expectations
of quality of life.
ACKNOWLEDGMENTS
We thank the staff of the Respiratory
Acute Care Unit at the Massachusetts
General Hospital for their tireless com-
mitment to patient care, Rhodes Berube
from the Department of Medicine for pro-
viding the financial data published in this
report, and Hui Zheng, PhD, for statisti-
cal consultation.
REFERENCES
1. Esteban A, Anzueto A, Frutos F, et al: Char-
acteristics and outcomes in adult patients
receiving mechanical ventilation. JAMA
2002; 287:345–355
2. Milberg JA, Davis DR, Steinberg KP, et al: Im-
proved survival of patients with acute respira-
tory distress syndrome (ARDS): 1983–1993.
JAMA 1995; 273:306 –309
3. Abel SJC, Finney SJ, Brett SJ, et al: Reduced
Crit Care Med 2007 Vol. 35, No. 11
mortality in association with the acute respi-
ratory distress syndrome (ARDS). Thorax
1998; 53:292–294
4. Rubenfeld G: Epidemiology of acute lung in-
jury. Crit Care Med 2005; 31:S276 –S284
5. Carson S, Bach P, Brzozowski L, et al: Out-
come after long-term acute care: An analysis
of 133 mechanically ventilated patients. Am J
Respir Crit Care Med 1999; 159:1568 –1573
6. Scheinhorn D, Chao D, Stearn-Hassenpflug
M, et al: Post-ICU weaning from mechanical
ventilation: The role of long-term facilities.
Chest 2001; 120:482S– 484S
7. Pilcher D, Bailey M, Treacher D, et al: Out-
comes, cost and long term survival of pa-
tients referred to a regional weaning centre.
Thorax 2005; 60:187–192
8. Scheinhorn D, Hassenpflug M, Votto J, et al:
Ventilator-dependent survivors of cata-
strophic illness transferred to 23 long-term
care hospitals for weaning from prolonged
mechanical ventilation. Chest 2007; 131:
76 – 84
9. Stoller J, Xu M, Mascha E, et al: Long-term
outcomes for patients discharged from a
long-term hospital-based weaning unit.
Chest 2003; 124:1892–1899
10. Nasraway S, Button G, Rand W, et al: Survi-
vors of catastrophic illness: Outcome after
direct transfer from intensive care to ex-
tended care facilities. Crit Care Med 2000;
28:19 –25
11. Bigatello L, Gettings E, Hess D: Outcome of
patients undergoing prolonged mechanical
ventilation for acute respiratory failure. Am J
Respir Crit Care Med 2003; 167:A457
12. Katz S, Ford A, Moskowitz R, et al: The index
of ADL: A standardized measure of biological
and psychosocial function. JAMA 1963; 185:
914 –991
13. Artigas A, Bernard G, Carlet J, et al: The
American-European Consensus Conference
on ARDS: Part 2. Am J Respir Crit Care Med
1998; 157:1332–1347
14. Scheinhorn D, Hassenpflug M, Votto J, et al:
Post-ICU mechanical ventilation at 23 long-
term care hospitals: A multicenter outcomes
study. Chest 2007; 131:85–93
15. Vitacca M, Vianello A, Colombo D, et al:
Comparison of two methods for weaning pa-
tients with chronic obstructive pulmonary
disease requiring mechanical ventilation for
more than 15 days. Am J Respir Crit Care
Med 2001; 164:225–230
16. MacIntyre N: Management of patients requir-
ing prolonged mechanical ventilation: Re-
port of a NAMDRC consensus conference.
Chest 2005; 128:3937–3954
17. Scheinhorn DJ, Chao DC, Stearn-Hassen-
pflug M, et al: Post-ICU mechanical ventila-
tion: Treatment of 1,123 patients at a re-
gional weaning center. Chest 1997; 111:
1654 –1659
18. Schonhofer B, Euteneur S, Nava S, et al:
Survival of mechanically ventilated patients
admitted to a specialised weaning centre. In-
tensive Care Med 2002; 28:908 –916
19. Engoren M, Arslanian-Engoren C, Fenn-
Beuderer N: Hospital and long-term out-
come after tracheostomy for respiratory fail-
ure. Crit Care Med 2004; 125:220 –227
20. Ely EW, Baker A, Dunagan D, et al: Effect on
the duration of mechanical ventilation of
identifying patients capable of breathing
spontaneously. N Engl J Med 1996; 336:
1864 –1869
21. Cook D, Rocker G, Marshall J, et al: With-
drawal of mechanical ventilation in anticipa-
tion of death in the intensive care unit.
N Engl J Med 2003; 349:1123–1132
22. Lee K: Patient preference and outcomes-
based surgical care among octogenarians and
nonagenarians. J Am Coll Surg 2006; 202:
356 –372
23. White D, Curtis J, Lo B, et al: Decisions to
limit life-sustaining treatment for critically
ill patients who lack both decision-making
capacity and surrogate decision-makers. Crit
Care Med 2006; 34:2053–2059
24. Lloyd C, Nietert P, Silvestri G: Intensive care
decision making in the seriously ill and el-
derly. Crit Care Med 2004; 32:649 – 654
25. Herridge M, Cheung A, Tansey C, et al: One-
year outcomes in survivors of the acute re-
spiratory distress syndrome. N Engl J Med
2003; 348:683– 693
2497