Girou E, Schortgen F, Delclaux C, Brun-Buisson C, Blot F, Lefort Y, Lemaire F, Brochard L. Association of Noninvasive Ventilation With Nosocomial Infections and Survival in Critically Ill Patients. JAMA. 2000;284(18):2361-2367. doi:10.1001/jama.284.18.2361
Author Affiliations: Unité d'Hygiène et Prévention de l'Infection (Drs Girou and Brun-Buisson), Service de Réanimation Médicale (Drs Schortgen, Delclaux, Brun-Buisson, Blot, Lefort, Lemaire, and Brochard), Institut National de la Santé et de la Recherche Médicale U492 (Drs Delclaux, Lemaire, and Brochard), Hôpital Henri Mondor, Assistance Publique-Hopitaux de Paris, Créteil, France.
Caring for the Critically Ill Patient Section
Editor: Deborah J. Cook, MD, Consulting Editor, JAMA.
Context Invasive life-support techniques are a major risk factor for nosocomial
infection. Noninvasive ventilation (NIV) can be used to avoid endotracheal
intubation and may reduce morbidity among patients in intensive care units
Objective To determine whether the use of NIV is associated with decreased risk
of nosocomial infections and improved survival in everyday clinical practice
among patients with acute exacerbation of chronic obstructive pulmonary disease
(COPD) or hypercapnic cardiogenic pulmonary edema (CPE).
Design and Setting Matched case-control study conducted in the medical ICU of a French
university hospital from January 1996 through March 1998.
Patients Fifty patients with acute exacerbation of COPD or severe CPE who were
treated with NIV for at least 2 hours and 50 patients treated with mechanical
ventilation between 1993 and 1998 (controls), matched on diagnosis, Simplified
Acute Physiology Score II, Logistic Organ Dysfunction score, age, and no contraindication
Main Outcome Measures Rates of nosocomial infections, antibiotic use, lengths of ventilatory
support and of ICU stay, ICU mortality, compared between cases and controls.
Results Rates of nosocomial infections and of nosocomial pneumonia were significantly
lower in patients who received NIV than those treated with mechanical ventilation
(18% vs 60% and 8% vs 22%; P<.001 and P = .04, respectively). Similarly, the daily risk of acquiring an infection
(19 vs 39 episodes per 1000 patient-days; P = .05),
proportion of patients receiving antibiotics for nosocomial infection (8%
vs 26%; P = .01), mean (SD) duration of ventilation
(6  vs 10  days; P = .01), mean (SD) length
of ICU stay (9  vs 15  days; P = .02), and
crude mortality (4% vs 26%; P = .002) were all lower
among patients who received NIV than those treated with mechanical ventilation.
Conclusions Use of NIV instead of mechanical ventilation is associated with a lower
risk of nosocomial infections, less antibiotic use, shorter length of ICU
stay, and lower mortality.
The occurrence of nosocomial infections is a major source of morbidity
and mortality in critically ill patients.1- 4
Among the various risk factors for acquiring an infection in the intensive
care unit (ICU), the use of invasive devices, such as intravenous catheters,
urinary catheters, and endotracheal tubes, is the leading factor.5- 7 Whereas many of these
devices cannot be avoided in the routine care of the patient, intubation of
the trachea can be avoided in selected groups of patients by delivering ventilation
through a full face mask or a nasal mask.8,9
This approach may have several advantages from the standpoint of infections.
It may reduce the risk of nosocomial pneumonia by maintaining the natural
barriers provided by the glottis and the upper respiratory tract and also
by reducing the duration of mechanical assistance, the need for sedation,
and the length of stay in the ICU.10- 13
All these factors may lead to a decrease in other types of nosocomial infections
by reducing the overall invasiveness of the care delivered to the patient.
Noninvasive ventilation (NIV) has been shown to reduce the need for
endotracheal intubation in patients with acute respiratory failure11,14 and to reduce mortality in selected
groups of patients with chronic obstructive pulmonary disease (COPD).11,13,15 The beneficial impact
of this technique on nosocomial infections has been suggested by several prospective
although the low event rates did not result in a significanct difference.
More importantly, the results of carefully conducted clinical trials in limited
groups of patients could differ from everyday practice. This is partly because
the optimum use of this technique requires specific staff training and patient
monitoring to achieve its potential efficiency.16,17
This may also explain why results vary among centers.18,19
In the medical ICU of Henri Mondor hospital, Créteil, France,
NIV has been progressively implemented over the last decade, especially for
patients with COPD or severe hypercapnic cardiogenic pulmonary edema (CPE).11 During these years, some of the patients did not
receive this technique despite a clinical status making them potentially eligible
for NIV, either because they had received endotracheal intubation before the
ICU admission, because the policy of out-of-hospital emergency teams or of
other departments was not to use NIV, or because the nurses and physicians
in charge of the patients were not yet familiar with this technique. Because
of this progressive implementation, we had the unique opportunity to compare
the different effects of NIV and endotracheal intubation on the infection
and survival rate out of the context of a randomized clinical trial. We therefore
performed a retrospective matched case-control study to compare outcomes for
similar patients admitted for acute exacerbation of COPD or severe CPE who
were treated with NIV or received endotracheal intubation and conventional
mechanical ventilation (MV).
The study was conducted in the medical ICU of Henri Mondor university
hospital, a 26-bed ICU that receives patients from the community, from several
other wards and specialized ICUs in the hospital, and from ICUs of other hospitals.
We performed a pairwise, retrospective case-control study with 1:1 matching.
The study period for cases extended from January 1, 1996, to March 31, 1998,
during which 2441 patients were admitted to the ICU.
Eligible patients were those who were admitted for acute exacerbation
of COPD or severe CPE and who received NIV for at least 2 hours as first-time
ventilatory support. We chose a minimum of 2 hours of NIV treatment to define
cases to exclude patients in whom the need for ventilatory support was questionable
because of quasi-immediate improvement and patients who rapidly required endotracheal
intubation after a brief attempted treatment of NIV because of associated
contraindications, such as refractory shock, or because of the inability of
the staff to adequately deliver NIV.
In addition, the following exclusion criteria were applied: patients
having a do-not-resuscitate order; patients with cancer or hematologic malignancy
with a poor short-term prognosis or who declined or were denied intubation;
patients having acute lung injury and COPD (presence of bilateral lung infiltrates
on a chest radiograph, a PaO2/FIO2 ratio <300 mm
Hg, and the absence of confirmed or suspected left cardiac failure); patients
having a contraindication to NIV; and patients receiving NIV during weaning
from mechanical ventilation. Noninvasive ventilation primarily was applied
intermittently, usually for periods of 2 to 6 hours, with several periods
delivered per day. Ventilatory settings during NIV delivery included pressure
support ventilation and positive end-expiratory pressure. The usual settings
were 15 to 20 cm H2O of pressure support and 0 to 5 cm H2O of positive end-expiratory pressure. Noninvasive ventilation was started
exclusively in the ICU and always within the first 72 hours of patient admission
to the ICU.
Control patients had to be admitted for exacerbation of COPD or severe
CPE and treated with conventional MV within the first 72 hours of ICU admission.
Postoperative patients, patients with asthma, and patients with metastatic
cancer or hematologic malignancy with a poor short-term prognosis or with
acute lung injury were excluded from the control group as well as patients
having potential contraindications for NIV as explained below.
A computer-generated list of potential controls was obtained from a
database that included 6264 patients during a 6-year period (from 1993-1998).
Controls were selected over a longer period than cases because the use of
NIV was less frequent 5 or 6 years ago compared with the most recent period
when most eligible patients were treated with NIV. Controls were chosen according
to the following matching criteria: same diagnosis on admission, age (±5
years), and Simplified Acute Physiology Score (SAPS) II20
(± 6 points) and Logistic Organ Dysfunction score21
(±3 points), which evaluate the severity of illness and both were calculated
within the first 24 hours of ICU admission. In addition, controls had to have
no contraindication for being treated with NIV, such as coma with swallowing
dysfunction, severe shock, or acute lung injury. Severe
shock was defined by the administration of epinephrine or norepinephrine
within the first 48 hours of ICU admission. Coma
was defined by a Glasgow Coma Scale score22
of less than 10 on admission. The list of potential controls was reviewed
for the best possible match, giving a ranking of priority to diagnosis, absence
of contraindication to NIV, SAPS II, Logistic Organ Dysfunction score, and
age. When multiple possible controls existed to match one case, the patient
with the date of ICU admission closest to that of the case patient was selected.
The following variables were collected: age, sex, dates of admission
and of discharge from the ICU, location before ICU admission, worst value
of the PaO2/FIO2 ratio within the first 72 hours of
ventilatory support, and primary diagnosis on admission. Arterial pH and PCO2 values were recorded before implementation of mechanical ventilation
in COPD patients when available. The total duration of ventilatory support,
including the days when NIV was delivered, was recorded. Therapeutic activity
was evaluated using the Omega score23 at discharge.
The Omega score is composed of therapeutic items, and it is divided into 3
categories accorded 1 to 10 points each as follows: category 1, items entered
only at the time of their first application; category 2, items entered at
each application; and category 3, items entered every day of application.
The total score, which covers the entire length of stay, is calculated by
adding the points obtained in the 3 categories.
The sites and dates of diagnosis of each nosocomial infection were recorded
as well as antibiotic regimens given during the ICU stay. Pneumonia, urinary
tract infection (UTI), primary bacteremia, and central venous catheter–related
infection, occurring at least 48 hours after ICU admission, were collected
according to the following definitions. Patients with new and persistent lung
infiltrates on chest radiographs, with a temperature greater than 38°C,
and with macroscopically proven purulent tracheal secretions were suspected
of having nosocomial pneumonia acquired while receiving either conventional
MV or NIV.
In such patients who received conventional MV, a diagnosis of ventilator-associated
pneumonia was ascertained by the positivity of a quantitative protected plugged
catheter culture, defined as at least 1 microorganism recovered at a significant
concentration (≥103 colony-forming units [CFU]/mL).24 In patients clinically suspected of having pneumonia
but treated with NIV, the positivity of a quantitative protected plugged catheter
culture at the same significant threshold as described above, when available,
or the sole administration of new antibiotics in the absence of other sites
of infection was used to characterize the presence of NIV-associated pneumonia.
A different definition was used in both groups, since the rationale for obtaining
protected bacteriological sampling of the lungs with quantitative cultures
is based on the presence of colonization in the airway of patients who were
intubated. This does not strictly apply to patients who were not intubated,
in whom treatment can be administered based on clinical, radiological, and
laboratory criteria, without bacteriological sampling of the lung.
A UTI was defined by the association of the
2 following criteria: pyuria (≥10 white blood cells/µL) and a urine
culture growing 1 × 105 CFU/mL in patients with clinical
signs of infection (temperature >38°C, leukocytosis, abnormal macroscopic
appearance of urine, and presence of urinary nitrites). Catheter-related infection was defined by a positive quantitative tip
culture with a significant threshold of 1 × 103 CFU/mL25 in the presence of localized signs of infection at
the access site and/or systemic signs of infection, such as fever or elevated
white blood cell count. Primary bacteremia was defined
by at least 1 positive blood culture result (2 or more blood cultures when
coagulase-negative staphylococci were isolated) in the absence of infection
focus growing the same microorganism or the isolation of the same organism
from a catheter segment quantitative culture and from a peripheral blood culture.
Surveillance was performed prospectively by a physician (C.B.-B.) who reviewed
all clinical and microbiological information weekly for each patient.
Categorical variables were expressed as percentage and continuous variables
as mean (SD). A P value ≤.05 in a 2-tailed test
was considered to indicate statistical significance. Percentages were compared
with use of the χ2 test and means with the t test. Nonparametric tests were used when the conditions for parametric
tests were not fulfilled (ie, Mann-Whitney test for continuous variables).
Failures of NIV treatment were retained in the NIV group and analyzed as cases.
Kaplan-Meier curves were used to determine the probabilities of remaining
free of nosocomial infection during the ICU stay in the 2 groups of patients;
curves were compared using the log-rank test. The statistical analysis was
performed using the Statistica 4.5 software (Statsoft, Inc, Tulsa, Okla).
During the study period, 134 patients received NIV (12.9 per 100 patients
needing ventilatory support). Of these 134 patients potentially eligible as
cases, 77 were excluded as indicated in Figure
1. Of the 57 case-patients enrolled in the study, matching was possible
for only 50 (88%) of them because of an insufficient number of suitable control
patients available in our database.
The results of matching for the criteria listed above were as follows:
All control patients were matched to cases for diagnosis on admission and
contraindication to NIV. Forty-nine (98%) of the 50 case-control pairs were
matched for SAPS II, 43 (86%) for age, and 45 (90%) for the Logistic Organ
Dysfunction score. Overall, matching was successful for 237 (94.8%) of 250
variables used. Several other variables also were compared in the 2 patient
groups on admission, and no difference was found between cases and controls,
except for the number of patients who received antibiotics on admission (P<.001) (Table 1).
All patients were hypercapnic on admission except 2 pairs of patients with
CPE. In the NIV group, 6 patients (12%) eventually required endotracheal intubation
As shown in Table 2, the
50 case-patients developed significantly fewer infections (P = .006) during their ICU stay than controls and received fewer antibiotics
for nosocomial infection (P = .01). The sites of
infection in cases were 4 pneumonia (among which 3 were diagnosed in patients
requiring endotracheal intubation), 3 UTIs, 1 primary bacteremia, and 1 central
venous catheter–related infection. The sites of infection in controls
were: 11 pneumonia, 10 UTIs, 5 primary bacteremia, and 4 central venous catheter–related
infections. For each site of infection, the rates were significantly lower
(P = .04, P = .03, and P = .002, respectively) in NIV patients than in controls
Figure 3 shows the percentage of patients
who remained free of nosocomial infection during the ICU stay. The probability
of remaining free of infection was significantly higher in patients receiving
NIV than in those receiving conventional MV (P = .04). The mean (SD) delay before onset of infection was 8 (3) days (median,
10 days) in the NIV group and 11 (10) days (median, 10 days) in the conventional
MV group (P = .52). Incidence densities, which reflect
the daily risk of acquiring an infection or receiving antibiotics, also were
significantly lower in cases than in controls (P = .05 and P<.001, respectively; Table 2). There was no significant difference in incidence densities
for nosocomial pneumonia (P = .40; Table 2).
The ICU mortality rate was significantly lower in patients treated with
NIV (2 deaths [46%] vs 13 deaths [26%]; P = .002).
Durations of ICU stay and of ventilation also were significantly shorter in
NIV patients (P = .02 and P
= .01, respectively; Table 2).
To ensure that the 7 nonmatched NIV patients were similar to the paired
cases, we compared their clinical characteristics with those of the 50 matched
cases and found no significant difference (data not shown). Moreover, when
we included them in the analysis together with hypothetical controls assumed
to have had no nosocomial infection, no antibiotic therapy, and to have survived,
the impact of NIV remained unchanged (data not shown).
This study describes the infectious complications associated with intubation
and conventional MV in patients who could have been treated by NIV. The results
show that the use of NIV is associated with lower rates of nosocomial pneumonia,
other acquired infections, and antibiotic administration. This lower nosocomial
infection rate is likely to contribute to the reduced mortality in this group
Because randomized controlled trials have demonstrated the benefits
of NIV on intubation and mortality in patients with COPD, the case-control
study can be considered an appropriate study design to assess the relation
of NIV with nosocomial infections in routine practice. It is of particular
relevance to assess the use of NIV treatment in everyday clinical practice
and to assess whether the results of randomized controlled trials mirror current
practice, for such a ventilatory technique in which the motivation and the
skills of the clinicians applying it are so important. Therefore, different
and complementary information is generated by the case-control study than
is generated by randomized clinical trials.
We used a careful matching process to avoid selecting more seriously
ill patients in the conventional MV group. Patients considered to have contraindication
to NIV,26 those presenting with coma, severe
shock, or acute lung injury were not enrolled as controls. Similarly, patients
treated with NIV but who had a do-not-resuscitate order or who declined or
were denied intubation were not enrolled as cases, since these patients would
not have received endotracheal intubation in any case and could not have been
matched with controls.
One limitation of this design, however, may come from the lack of sufficient
control patients or from the limited number of matching variables. The possibility
that the choice of endotracheal intubation was based on a higher clinical
severity may exist in some cases. However, this was not apparent when comparing
laboratory data or severity indices on admission. The only difference between
cases and controls concerned the number of patients receiving antibiotics
on admission. This was frequently related to suspected bronchitis, however,
and not to documented infection. This lack of difference in severity indices
suggests that if any difference in severity still existed, it was modest and
largely insufficient to explain most of the major differences in nosocomial
infection rates observed in this study and that corroborated the results of
Lower rates of nosocomial pneumonia in patients receiving NIV has been
suggested in several studies.27,28
In a randomized controlled trial of patients with COPD, Brochard et al11 found that treatment with NIV reduced the total number
of adverse events associated with mechanical ventilation and length of stay
in the ICU and that the rate of nosocomial pneumonia was reduced from 17%
to 5%. While this difference in rate of pneumonia was not significant, crude
mortality was significantly reduced. Nava et al13
used NIV as a means to shorten the duration of invasive mechanical ventilation
by switching to NIV after 48 hours of mechanical ventilation. In the group
of 25 patients treated with NIV early on, no patient developed nosocomial
pneumonia vs 28% in the group who were weaned from invasive MV; this approach
was associated with a reduced mortality at 2 months. Recently, Antonelli et
al10 showed that NIV could be used instead
of endotracheal intubation in severely hypoxemic patients. Adverse events
were significantly lower in the NIV group. When pooling together nosocomial
sinusitis and pneumonia, these infections were significantly reduced by the
use of NIV. All these results are consistent with our data on the relation
between treatment with NIV and nosocomial pneumonia.
Close monitoring of nosocomial infections through a continuous surveillance
system implemented in our ICU since 1993 made it most unlikely that an episode
of infection would be missed in this patient population. We used definitions
of nosocomial infections routinely applied in ICUs and applied strict criteria
to define bacteriologically proven ventilator-associated pneumonia in patients
who were treated with invasive mechanical ventilation. Pneumonia acquired
during NIV treatment could be based on the administration of new antibiotics
in patients with no other documented site of infection and who met the clinical
criteria for pneumonia. While this definition was chosen to avoid omitting
any possible case of pneumonia in patients treated with NIV, it may have resulted
in overestimating the rate of pneumonia in the NIV group.
In addition to pneumonia, infections at all other sites were significantly
reduced in patients treated with NIV independent from the patient's severity
of illness. This finding is in accordance with those reported in a recent
study.28 This may reflect a less frequent use
or shorter duration of invasive devices in those patients.5,28
Craven et al5 analyzed risk factors for nosocomial
infections in surgical ICU patients. A greater exposure to invasive devices
and procedures of all types was found to be associated with a higher incidence
of nosocomial infections, independent from the underlying severity of illness.
More recently, a case-control study assessing risk factors for nosocomial
infections and in which patients were carefully matched on initial severity
of illness, showed that a persistent high level of therapeutic activity was
associated with the acquisition of nosocomial infection in ICU patients.3 In our study, patients treated with NIV were administered
fewer antibiotics for nosocomial infection and had shorter length of ICU stay
and duration of ventilation. This might explain the reduced frequency of all
types of nosocomial infections observed in the NIV patients.
Noninvasive ventilation cannot be administered to all patients with
respiratory failure.9 Patients with acute exacerbation
of chronic respiratory failure, especially COPD and CPE, constitute the groups
more likely to benefit from NIV treatment.11,14,15
Several randomized controlled studies have shown reductions in mortality in
COPD patients. 11,13,29
Patients with community-acquired pneumonia also have been shown to benefit
from the use of NIV, especially in patients with COPD.30
The other group represented in our study included patients with CPE,
a disease that has been shown to respond well to NIV treatment,31- 34
especially when associated with hypercapnia and ventilatory failure.31,35 In our study, all but 2 NIV cases
with CPE were hypercapnic on admission. Although patients with hypoxemic respiratory
failure represent a larger scope of diseases, not all of these patients are
good candidates for NIV therapy, and the selection of appropriate patients
is still a matter of research. Wysocki et al35
have shown that among hypoxemic patients, only patients with associated ventilatory
failure and acute hypercapnia really benefited from treatment with NIV. Antonelli
et al10,36 showed that in selected
patients, NIV could be beneficial in patients with severe hypoxemia or in
patients with solid organ transplant. Other authors reported that severe hypoxemic
pneumonia was a poor indication for the technique.9,19
Our study showed a lower mortality associated with the use of NIV. This
can be explained in part by the reduction not only of nosocomial infections
but also by other complications previously reported to be associated with
prolonged length of stay and not reported here.1- 4
Our results confirmed the results of trials that showed a reduction in mortality
with NIV therapy.11,13 It is important,
however, to evaluate whether this observed reduction in mortality is observed
in everyday practice, since the results of carefully conducted, concealed
but nonblinded studies evaluating selected group of patients may not always
be replicated during routine application of the technique.
In conclusion, our study found that critically ill patients treated
with NIV were less likely to acquire pneumonia and other nosocomial infections
than similar patients treated with conventional MV and that the use of NIV
is associated with a shortened stay in the ICU and reduced mortality.