1. sports
2. football

CHEST
Original Research
CRITICAL CARE MEDICINE
Ventilator Advisory System Employing
Load and Tolerance Strategy
Recommends Appropriate Pressure
Support Ventilation Settings*
Multisite Validation Study
Michael J. Banner, PhD; Neil R. Euliano, PhD; Neil R. MacIntyre, MD;
A. Joseph Layon, MD, FCCP; Steven Bonett, RRT; Michael A. Gentile, RRT;
Zoheir Bshouty, MD, PhD, FCCP; Carl Peters, MD; and Andrea Gabrielli, MD
Background: Loads on the respiratory muscles, reflected by noninvasive measurement of the
real-time power of breathing (POBn), and tolerance of these loads, reflected by spontaneous
breathing frequency (f) and tidal volume (VT), should be considered when evaluating patients
with respiratory failure. Pressure support ventilation (PSV) should be applied so that muscle loads
are not too high or too low. We propose a computerized, ventilator advisory system employing a
load (POBn) and tolerance (f and VT) strategy in a fuzzy logic algorithm to provide guidance for
setting PSV. To validate these recommendations, we performed a multisite study comparing the
advisory system recommendations to experienced physician decisions.
Methods: Data were obtained from adults who were receiving PSV (n ⴝ 87) at three university
sites via a combined pressure/flow sensor, which was positioned between the endotracheal tube
and the Y-piece of the ventilator breathing circuit and was directed to the advisory system.
Recommendations from the advisory system for increasing, maintaining, or decreasing PSV were
compared at specific time points to decisions made by physician intensivists at the bedside.
Results: There were no significant differences in the recommendations by the advisory system
(n ⴝ 210) compared to those of the physician intensivists to increase, maintain, or decrease PSV
(p > 0.05). Physician intensivists agreed with 90.5% of all recommendations. The advisory system
was very good at predicting intensivist decisions (r2 ⴝ 0.90; p < 0.05) in setting PSV.
Conclusions: The novel load-and-tolerance strategy of the advisory system provided automatic and
valid recommendations for setting PSV to appropriately unload the respiratory muscles that were as
good as the clinical judgment of physician intensivists.
(CHEST 2008; 133:697–703)
Key words: acute respiratory failure; mechanical ventilation; pressure support ventilation; respiratory monitoring; work of
breathing
Abbreviations: Crs ⫽ respiratory system compliance; f ⫽ spontaneous breathing frequency; Fio2 ⫽ fraction of inspired
oxygen; FIS ⫽ fuzzy logic inference system; IMV ⫽ intermittent mandatory ventilation; PEEP ⫽ positive end expiratory
pressure; Petco2 ⫽ partial pressure end-tidal carbon dioxide; POBn ⫽ real-time measurement of power of breathing;
PSV ⫽ pressure support ventilation; Rrs ⫽ respiratory system resistance; Spo2 ⫽ pulse oximetric oxygen saturation;
V˙e ⫽ minute ventilation; Vt ⫽ tidal volume
on the respiratory muscles of spontaneL oads
ously breathing patients with respiratory fail-
ure who were receiving ventilatory support may be
reflected by the power of breathing (normal in
adults, 4 to 8 J/min1). Power of breathing, the rate
at which work is done per minute, is a better
assessment of respiratory muscle loads than work
www.chestjournal.org
Downloaded From: http://journal.publications.chestnet.org/ on 11/10/2014
of breathing per breath because it is measured
over time, not for an individual breath. Objective
measurement of respiratory muscle loads via
power of breathing may be useful to determine
when loads are appropriate, too high (eg, ⬎ 15
J/min), or too low (eg, totally unloaded at 0
J/min).2
CHEST / 133 / 3 / MARCH, 2008
697
Tolerance of these loads, as reflected by spontaneous breathing frequency (f) and tidal volume
(Vt),3 are routinely assessed when observing patientventilator interactions. As response variables to load,
f and Vt provide information on a patient’s ability to
handle or tolerate some level of load. As the respiratory muscles become loaded from increased elastic
loads (ie, decreased lung and/or chest wall compliance) and/or resistive loads (ie, physiologic airways
resistance and imposed breathing apparatus resistance), a fatiguing breathing pattern may ensue. To
avoid respiratory muscle fatigue, the respiratory
center increases f and minimizes the inspiratory
contraction time of the respiratory muscles, which
decreases Vt. This produces a pattern of rapid and
shallow breathing that minimizes the work of breathing and is, thus, the most energy efficient combination of f and Vt.4 Ostensibly, breathing pattern data
are thought to provide a useful representation of the
patient’s physiologic pulmonary load sensors5–7 and
their ability to tolerate loads, and an inference of the
patient’s reserve or capability of breathing. At times,
however, when applying PSV, a poor relationship
between breathing pattern and respiratory muscle
workload data has been reported.2,8 To provide a
more comprehensive assessment, it may be prudent
to consider monitoring both the f and Vt breathing
pattern and respiratory workload in a complimentary
manner for patients receiving PSV.
When spontaneous breathing of a mechanically
ventilated patient is possible, pressure support ventilation (PSV) may be applied so that respiratory
muscle loads are not too high or too low and they
*From the Departments of Anesthesiology, Physiology, and
Surgery (Drs. Banner, Layon, Peters, and Gabrielli, and Mr.
Bonett), College of Medicine, University of Florida, Gainesville,
FL; Convergent Engineering (Dr. Euliano), Gainesville, FL; the
Division of Pulmonary and Critical Care Medicine (Dr. MacIntyre and Mr. Gentile), Duke University Medical Center,
Durham, NC; and the Department of Medicine (Dr. Bshouty),
University of Manitoba, Winnipeg, MB, Canada.
This research was presented in part at the 2006 annual meeting
of the American College of Chest Physicians.
This research was supported by grants from Respironics, Inc, and
Convergent Engineering. A patent describing the system used in
this study has been submitted.
Dr. Euliano is the President of Convergent Engineering, the
manufacturer of the software used in this study. Authors Banner,
MacIntyre, Layon, Bonett, Gentile, Bshouty, Peters, and Gabrielli have reported to the ACCP that no significant conflicts of
interest exist with any companies/organizations whose products
or services may be discussed in this article.
Manuscript received August 9, 2007; revision accepted December 1, 2007.
Reproduction of this article is prohibited without written permission
from the American College of Chest Physicians (www.chestjournal.
org/misc/reprints.shtml).
Correspondence to: Michael Banner, PhD, University of Florida,
College of Medicine, Department of Anesthesiology, Box 100254,
Gainesville, FL, 32610; e-mail: MBanner@anest.ufl.edu
DOI: 10.1378/chest.07-2011
698
Downloaded From: http://journal.publications.chestnet.org/ on 11/10/2014
appropriately unload the respiratory muscles.3 The
level of inspiratory pressure with PSV is generally
determined clinically by observing the spontaneous
breathing pattern (f and Vt), and the patient’s
respiratory muscles are generally thought to be
appropriately unloaded when a comfortable breathing pattern is present. More sophisticated approaches involve the placement of an esophageal
balloon and titration of the PSV inspiratory pressure
to the measured loads on the respiratory muscles.2
Under some conditions when managing patients
with respiratory failure, respiratory system mechanics and/or neuromuscular capabilities may change,
and expert clinical personnel may not always be
available to make timely, proper bedside clinical
assessments for setting PSV inspiratory pressure
levels. For these conditions, we propose a computerized, ventilator advisory system employing a fuzzy
logic inference system (FIS) to make continuous
real-time recommendations for increasing, maintaining, or decreasing PSV for patients with respiratory
failure based on a load (power of breathing)-andtolerance (f and Vt) strategy (Fig 1). Fuzzy logic is a
process of using probability distributions instead of
simple “yes/no” decisions, as in a simple rule-based
system, to drive the ventilator decision making.9
The purpose of this multisite clinical study was to
validate recommendations for increasing, maintaining, or decreasing PSV from this advisory system
compared to the recommendations of bedside physician intensivists for setting PSV for patients with
respiratory failure. We hypothesized that this advisory system would supply comparable recommendations on PSV treatment compared to those chosen by
experienced clinicians.
Materials and Methods
Institutional review board approvals to conduct the study were
obtained from three university hospital sites (the University of
Florida, Duke University Medical Center, and the University of
Manitoba). A total of 87 adults (Table 1) intubated or tracheostomized (internal diameter, 6.5 to 8.5 mm) and with respiratory
failure from various etiologies were enrolled into the study. Also
listed in Table 1 are the ventilator settings and related data on
enrollment into the study. Pneumonia, pulmonary edema, sepsis,
congestive heart failure, and subarachnoid hemorrhage were
diagnosed in fairly equal numbers of patients from all three sites.
Some patients had direct lung injury from penetrating blunt chest
trauma (n ⫽ 12), and some had COPD (n ⫽ 16). Combining all
sites, 52 patients were treated in medical ICUs and 35 patients
were treated in surgical ICUs. Regarding pulmonary mechanics,
the values for respiratory system compliance (Crs) ranged from
0.01 to 0.09 L/cm H2O; the values for respiratory system
resistance (Rrs) ranged from 5.8 to 17.8 cm H2O/L/s. At the time
of study enrollment, most patients were considered to be in the
resolving phases of respiratory failure and were receiving ventilatory support. All patients were breathing spontaneously, receiving PSV, and were hemodynamically stable, and provided with
Original Research
Figure 1. A clinical example of using the ventilator advisory system is shown. Respiratory muscle load,
reflected by POBn, and tolerance for that load, reflected by f or Vt, are combined in a FIS to formulate
recommendations for increasing, maintaining, or decreasing PSV. In the image, the advisory system
determined that because POBn is too high (13 J/min) and the tolerance for that load is inappropriate,
as reflected by f being too high (37 breaths/min) and Vt being borderline low (5 mL/kg of ideal body
weight), the set level of PSV of 15 cm H2O is insufficient and more PSV is needed. Note in the “FIS
Accuracy” section (bottom), the advisory system (“FIS Suggests”) and “Increase” in PSV, and the
physician intesivists (“Expert Suggests”) concur by agreeing on the “PSV Action” to increase PSV.
MV ⫽ V˙e; PEEPi ⫽ intrinsic PEEP.
appropriate sedation and analgesia as needed. Patients who were
hemodynamically unstable (eg, mean arterial BP acutely changing from 40 to 80 mm Hg, with irregular changes in heart rate
and/or arrhythmias), heavily sedated (eg, high levels of opioids for
analgesia, predisposing the patient to acute changes in spontaneous f and Vt), or had irregular breathing patterns (eg, patients
with closed-head injuries whose spontaneous f and Vt may
acutely increase and decrease over short periods) were excluded
from the study. Patients were ventilated with the same type of
ventilator (model 840; Puritan-Bennett; Pleasanton, CA), which
was set in the PSV mode and positive end-expiratory pressure
(PEEP) mode. For some patients, intermittent mandatory ventilation (IMV), with a Vt of 8 mL/kg of ideal body weight and the
mandatory ventilation rate at 2 breaths/min, was combined with
PSV and PEEP. IMV was used at one site because of the local
practice of combining low IMV rate at 2 breaths/min with PSV
and PEEP. At the other two sites, IMV at 2 breaths/min was
occasionally combined with PSV and PEEP. A constant or square
inspiratory flow waveform was applied with the IMV breaths. For
PSV, the “% rise” setting on the ventilator (ie, the control to
adjust the ventilator flow-rate output during inhalation and, thus,
the rate of pressure rise, which results from the interaction of
ventilator flow rate output and patient inspiratory flow rate
demand10) ranged from 60 to 80%, and the expiratory sensitivity
setting (“Esens”) or PSV breath-termination criteria was 25%.
The levels of PSV, PEEP, fraction of inspiratory oxygen (Fio2),
carbon dioxide elimination, and hemoglobin oxygen saturation
were comparable for patients at all sites at the time of enrollment
into the study (Table 1).
www.chestjournal.org
Downloaded From: http://journal.publications.chestnet.org/ on 11/10/2014
The advisory system under investigation consists of the following two components: a commercially available respiratory monitor (NICO; Respironics; Wallingford, CT); and a laptop computer. Data from a combined pressure/flow/carbon dioxide
sensor, which was positioned between the endotracheal tube and
the Y-piece of the ventilator breathing circuit, were directed to
the respiratory monitor for measurements including f, Vt, minute
ventilation (V˙e), PSV, PEEP, and partial pressure end-tidal
carbon dioxide (Petco2) [Fig 2]. Crs and Rrs were calculated
using the least-squares method during PSV breaths.11 Pulse
oximetric oxygen saturation (Spo2) was also measured from a
finger site with the monitor.
Some of these data were then directed to the laptop computer
containing special software (Convergent Engineering; Gainesville, FL) for operation of an artificial neural network that allows
for the noninvasive (ie, not requiring insertion of an esophageal
balloon catheter), real-time measurement of power of breathing
(POBn), and operation of the FIS to process POBn, f, and Vt
data to recommend an increase, no change, or decrease in PSV.
Obviating the need for an esophageal balloon catheter, greatly
simplifies the measurement of POBn, especially for prolonged
periods such as in patients receiving ventilatory support over
many days. Five predictor variables or input elements are used by
the artificial neural network to calculate POBn (the methodology
describing how these variables are determined has been published elsewhere8), as follows: (1) spontaneous V˙e (not including
IMV breaths) correlates directly with power of breathing; (2)
increased intrinsic PEEP is associated with increased power of
breathing and vice versa; (3) lower inspiratory trigger pressure
CHEST / 133 / 3 / MARCH, 2008
699
96.4 ⫾ 2.6
*Values are given as the mean ⫾ SD, unless otherwise indicated. M ⫽ male; F ⫽ female. All patients were hemodynamically stable prior to and throughout the study.
‡p ⬍ 0.05 (University of Florida vs Duke University Medical Center and University of Manitoba).
†p ⬍ 0.05 (University of Florida vs Duke University Medical Center).
§p ⬍ 0.05 (Duke University Medical Center vs University of Manitoba).
37.4 ⫾ 7
9.7 ⫾ 2.8
0.51 ⫾ 0.21
20 ⫾ 7
0.44 ⫾ 11
6.0 ⫾ 2
12.6 ⫾ 5
17 M/11 F
67 ⫾ 5
61 ⫾ 16
90 ⫾ 24
95.5 ⫾ 3
38.5 ⫾ 8
8.0 ⫾ 2.6
0.46 ⫾ 0.13
19 ⫾ 6
6.5 ⫾ 1.7
12 ⫾ 4.6
66 ⫾ 4
60 ⫾ 16
94 ⫾ 34
10 M/14 F§
0.33 ⫾ 0.06§
96.7 ⫾ 2.5
36.7 ⫾ 11
8.7 ⫾ 3
0.52 ⫾ 0.2
17 ⫾ 7
7⫾4
11 ⫾ 4
69 ⫾ 5
University of Florida
35 patients†
Duke University Medical Center
24 patients
University of Manitoba
28 patients
55 ⫾ 19
83 ⫾ 21
20 M/15 F‡
0.39 ⫾ 0.08†
Spo2, %
Petco2,
mm Hg
V˙e, L/
min
Vt, L
f, breaths/
min
Fio2
PEEP,
cm H2O
PSV,
cm H2O
Patients, No.
Weight,
kg
Sites
Age, yr
Height,
inch
Table 1—Patient Data From All Sites on Enrollment Into the Study*
700
Downloaded From: http://journal.publications.chestnet.org/ on 11/10/2014
Figure 2. Ventilator advisory system (respiratory monitor plus
laptop computer [PC]) employs a load-and-tolerance strategy as
described in Figure 1 by combining POBn (load) with f and Vt
(tolerance for the load) to determine an appropriate level of PSV.
depth (ie, the pressure below the baseline airway pressure just
before the ventilator triggers “ON”) pressures are associated with
increased power of breathing and vice versa; (4) lower inspiratory
flow rise times (ie, how rapidly the inspiratory flow waveform
rises during a PSV breath) are associated with increased power of
breathing and vice versa; and (5) higher respiratory muscle
pressures (based on the equation of motion applied to the
respiratory system; ie, pressure ⫽ [Vt/Crs] ⫹ [inspiratory flow
rate ⫻ Rrs]1,12) are associated with increased power of breathing
and vice versa. POBn has been reported to be highly correlated
(r ⫽ 0.91; p ⬍ 0.05) with invasively measured power of breathing
(using esophageal pressure measurements) and is considered to
be a very good predictor of invasively measured power of
breathing for adults treated with PSV.8 POBn reflects the total
load on the respiratory muscles, which includes elastic loads of
the respiratory system, and resistive loads of the airways, endotracheal tube, and ventilator apparatus. A potential limitation of
POBn is that it cannot differentiate these component loads.
The advisory system attempts to maintain POBn (ie, f or
respiratory rate) and Vt in the center sections or lighter gray
ranges shown in Figure 1 by recommending directional changes
(ie, increase, maintain, or decrease) in the PSV setting. The
ranges used by the advisory system for POBn were derived, in
part, from normal values of power of breathing,1,8 a work of
Original Research
breathing outcome study,2 and our previous clinical experience.8
For adults, the normal range of power of breathing is 4 to 8
J/min.1 Based on measuring the power of breathing in patients
receiving ventilatory support for ⬎ 10 years, it appears that adults
with respiratory failure who are treated with PSV tolerate a
maximum power of breathing up to about 10 to 12 J/min. In a
previous study8 of adults receiving PSV, most demonstrated
a power of breathing in the range of 2 to 8 J/min. Kirton et al2
reported that respiratory muscle workloads maintained in a fairly
normal range using PSV were well tolerated. In our experience,
power of breathing values of ⬎ 12 to 15 J/min in adults are not
well tolerated, requiring higher levels of PSV. Ranges for f and
Vt used by the advisory system are in keeping with those ranges
generally applied to adults who are spontaneously breathing with
PSV.3,13 We agree with most authorities that the generally
established ranges for maintaining f and Vt should be between
about 10 to 25 breaths/min and 6 to 8 mL/kg, respectively, when
applying PSV. It is also reasonable to apply PSV so that inappropriately low f (eg, 4 to 6 breaths/min) or high f (eg, 30 to 40
breaths/min), and low Vt (eg, ⬍ 4 mL/kg) or high Vt (eg, ⬎ 12
mL/kg) do not occur. Wider ranges for POBn, f, and Vt are
indicated in the white and darker gray areas, as shown in Figure
1. The advisory system treats these ranges as cautionary in nature.
The advisory system assesses the probability of the load (ie,
POBn) and tolerance (ie, f and Vt) parameters varying in relation
to one another in all ranges and then makes the most reasonable
decision for increasing, maintaining, or decreasing PSV.
The PSV was set initially by physician intensivists and was
adjusted during the course of treatment, using the traditional
approach of assessing spontaneous breathing pattern data (ie, f
between 15 and 25 breaths/min, Vt between 6 and 8 mL/kg ideal
body weight), the absence of sternocleidomastoid muscle contraction, and the appearance of breathing comfortably. Periodically, the physician intensivist was first asked to evaluate the
patient using the aforementioned traditional approach for setting
PSV. Next, the physician intensivist was informed of the
recommendation of the advisory system to increase, maintain,
or decrease PSV and was asked whether he/she agreed or
disagreed. The actual PSV settings applied to all patients
during the study were always those settings prescribed by the
physician intensivists.
For all sites, the mean (⫾ SD) PEEP and Fio2 settings, as
determined by the physician intensivists, were 6.8 ⫾ 4 cm H2O
and 0.38 ⫾ 0.05, respectively, and were held constant during the
study period (approximately 4 h). There were no significant
differences in PEEP, V˙e, Petco2, and Spo2 prior to beginning
the study comparing all sites (Table 1) or during the study, and no
clinically significant variations in hemodynamic parameters were
observed in all patients throughout the study.
The physician intensivists at each site ordered the level of PSV
to be used for all patients during the study. It was essential to
determine whether there was a correlation between this level of
PSV and the PSV level identified by the advisory system as being
appropriate for treating the patient. When physicians ordered an
increase or decrease in PSV, generally, this was an increase in
PSV by 5 cm H2O or a decease by the same amount from the
current setting. When the advisory system recommended an
increase in PSV, this was interpreted to mean “increase PSV by 5
cm H2O from the current level.” For example, if the PSV was 7
cm H2O, then the recommended PSV setting would be identified
to mean an increase in PSV to 12 cm H2O. Likewise, when a
decrease in PSV was recommended, this was identified to mean
a decrease in PSV by 5 cm H2O from the current setting. With
this approach, all physician-ordered PSV settings were regressed
with the recommended PSV settings identified by the advisory
system.
www.chestjournal.org
Downloaded From: http://journal.publications.chestnet.org/ on 11/10/2014
Data were analyzed using the ␹2 test, analysis of variance,
regression analysis, and ␬ statistic.14 The ␣ level was set at 0.05
for statistical significance.
Results
Approximately 40% of the patients were from the
University of Florida, 30% were from Duke University Medical Center, and 30% were from the University of Manitoba. PSV ranged from 2 to 25 cm
H2O during the study when comparing all sites. The
mean (⫾ SD) level of PSV for all sites was 12 ⫾ 5 cm
H2O. For individual sites, there were no significant
differences in the levels of PSV used. A total of 210
recommendations from all sites was made by the
advisory system to increase, maintain, or decrease
PSV. There was a mean total of 90.5% agreement
between all of the recommendations made by the
advisory system and those made by the attending
physician intensivists, based on their evaluations of
the patients (Table 2). The mean (⫾ SE) strength of
this agreement (␬ statistic) was 0.84 ⫾ 0.03
(p ⬍ 0.001). (A value for the ␬ statistic between 0.81
to 1.0 is considered to be almost perfect agreement.14)
There were no clinically significant differences in the
recommendations made by the advisory system compared to those of the attending physician intensivists to
increase, maintain, or decrease PSV (p ⬎ 0.05) [Table
2]. At individual sites, there were no significant differences in the recommendations made by the advisory
system and the physician intensivists (p ⬎ 0.05). For all
sites, compared to increasing the level of PSV, there
were about three times as many recommendations to
decrease or maintain PSV (p ⬍ 0.05) [Table 2].
Values for POBn from all sites ranged from 2 to 22
J/min during the study. Physician intensivists made
the recommendation to “maintain PSV,” implying
appropriate respiratory muscle loads, 97 times, and
the advisory system agreed 90 times (93% agree-
Table 2—Recommendations Made by the Advisory
System Compared to Recommendations Made by
Physician Intensivists (␹2 Analysis)
Advisory System
Physician Intensivists
Increase PSV
Maintain PSV
Decrease PSV
Total
Increase
PSV
Maintain
PSV
Decrease
PSV
Total
29
3
0
32
1
84
5
90*
1
10
77
88*
31
97†
82†
210
*p ⬍ 0.05 (compared to increasing PSV; ␹2 ⫽ 0.489; degrees of
freedom ⫽ 2; for significance at the 0.05 level, ␹2 should be ⱖ 5.99).
†p ⬍ 0.05 (for all sites compared to increasing PSV).
CHEST / 133 / 3 / MARCH, 2008
701
Figure 3. Relationship between recommendations made by the
ventilatory advisory system on PSV settings, compare to PSV
settings selected by physician intensivists for all sites. The clusters
of symbols at 5, 10, and 15 cm H2O, for example, represent
considerable overlap of many PSV settings at these levels. There
was an excellent and significant correlation between the advisory
system and physician intensivists (r ⫽ 0.95) and the advisory
system as a very good predictor of an appropriate level of PSV as
determined by the physician intensivists (r2 ⫽ 0.90).
ment). Combining patients from all sites under this
condition, the mean POBn was 8.1 ⫾ 2 J/min (ie,
there was no significant differences in POBn between sites).
Regarding the relationship between the PSV level
ordered by physicians and the PSV level identified by
the advisory system, there was an excellent, positive
correlation with r values ranging from 0.94 to 0.96
when comparing all sites. The average r2 value for all
sites was 0.90. In other words, the advisory system
predicted or explained 90% of the variance in the
physician’s recommendations for setting PSV (Fig 3).
Discussion
Valid, clinically appropriate recommendations for
increasing, maintaining, or decreasing PSV to unload
the respiratory muscles were provided by the advisory system. Considering all sites, 90.5% of the time
the advisory system recommended increasing, maintaining, or decreasing PSV in the same manner as
determined by attending critical care physician intensivists who were evaluating the patients at the
bedside (Table 2). There was minimal variability
between different institutions, regardless of whether
patients were treated in a surgical ICU (University of
Florida) or a medical ICU (Duke University Medical
Center and University of Manitoba).
A possible explanation for the 9.5% disagreement
between the advisory system and the physician intensivists is that, at times, physicians may have
decided to either “rest” or “exercise” a patient’s
702
Downloaded From: http://journal.publications.chestnet.org/ on 11/10/2014
respiratory muscles, contrary to advice from the
advisory system, based on their knowledge of the
patient’s actual clinical condition, weaning process,
and overall clinical plan. To rest a patient, a higher
level of PSV could be applied (eg, 25 cm H2O) to
totally unload the respiratory muscles (POBn, 0
J/min). This can be done at times when physician
intensivists determine that it is appropriate to relieve
a patient’s respiratory muscles of the breathing
workload; for example, immediately following intubation for acute respiratory failure. Conversely, to
exercise the respiratory muscles of patients considered to be candidates for weaning, a lower level of
PSV may be applied, allowing a higher workload (eg,
POBn, ⬎ 10 J/min). This may be done on occasions
when physician intensivists determine it appropriate
to “push” a patient for a short time, to assess a
patient’s reserve to breathe.
In one patient (with muscular dystrophy), the
advisory system recommended a decrease in PSV
while the physician intensivist increased PSV. Previously, PSV had been decreased from 14 to 9 cm
H2O, causing Vt to decrease from 0.60 to 0.42 L or
6 mL/kg of ideal body weight (70 kg). Related data at
9 cm H2O were as follows: POBn, 3 J/min; f, 18
breaths/min; Petco2, 40 mm Hg; and Spo2, 97%.
Also, the patient’s respiratory efforts appeared to be
excessive. The advisory system recommended a decrease in PSV because the respiratory muscle load
was low and other parameters were in appropriate
ranges. The physician-ordered increase in PSV may,
in part, have been based on the 30% acute decrease
in Vt and respiratory efforts. As a result of the
weakness associated with the patient’s disease, the
patient may only have been able to generate a small
Vt, despite POBn being in a low range. The physician recommendation to increase PSV to increase Vt
to a more appropriate range was, in part, dictated by
the patient’s overall clinical state and knowledge of
all the patient’s data, which were unknown by the
advisory system.
There were three times as many recommendations
by the advisory system to decrease or maintain PSV
than to increase it. These recommendations were in
agreement with those of the physician intensivists.
The treatment philosophy or method of reasoning
employed by physician intensivists the great majority
of the time was to wean PSV down or at least
maintain the level, rather than to increase PSV and
unload the respiratory muscles further. The loadand-tolerance strategy used in the advisory system
appears to reflect this approach.
The advisory system functions as an open-loop
feedback system. Such systems have been used in
clinical practice before. Belel et al15 employed an
open-loop feedback advisor for ventilating neonates.
Original Research
Parameters from the bedside (ie, Spo2, respiratory
waveforms, heart rate, transcutaneous Po2 and Pco2,
arterial BP, and temperature) and ventilator (ie,
inspiratory time, expiratory time, peak inflation pressure, PEEP, mean airway pressure, and Fio2) were
feedback variables used for formulating the recommendations for setting the ventilator to treat ventilation and oxygenation abnormalities. Clinicians
agreed with the ventilation recommendations 91% of
the time, and with the oxygenation recommendations 94% of the time. Kwok et al16 described an
adaptive neuro-fuzzy inference system, open-loop
Fio2 ventilator advisor that estimates intrapulmonary right-to-left shunt by employing a respiratory
index, which is derived in part from calculating the
alveolar air equation. In turn, clinical advice is given
on the Fio2 needed to attain a target Pao2 for
patients with compromised pulmonary function. Rutledge et al17 described a qualitative and quantitative
ventilator management advisor (VentPlan) to provide
recommendations for setting a ventilator that are
based on a mathematical model of cardiopulmonary
physiology. By applying a patient-specific physiologic
model, the advice for ventilator settings was in
agreement with the physician preferences for setting
the ventilator.
When physician intensivists determined it appropriate to maintain a specific level of PSV, the
implication was that respiratory muscle loads were
optimal at that time. For this clinical condition, the
advisory system agreed with the physician intensivists
more than 9 times in 10. The POBn for all patients at
all sites at that time approximated the upper limit of
normal for the power of breathing in adults.1
The potential clinical utility of this advisory system is
that it is automatic and continuously operational. This
would be important in clinical situations where expert
physician intensivists may not always be available for
patient assessment and to make timely decisions for
setting PSV. Examples of this might include large,
multipatient, ventilatory care facilities/nursing homes,
battlefield military hospitals, mass-casualty treatment
centers, and large ICUs. In such venues with limited
manpower, the advisory system may be able to assist in
setting an appropriate level of PSV.
Although this study was an initial step in validating
the advisory system, a possible limitation of this
study may be that to fully validate the advisory
system a similar study over a wider range of patient
conditions may be needed. Also, such a study could
be performed over the total duration of ventilatory
support up to the point of extubation.
In summary, this clinical study compared recommendations from a computerized, ventilator advisory
www.chestjournal.org
Downloaded From: http://journal.publications.chestnet.org/ on 11/10/2014
system for increasing, maintaining, or decreasing
PSV, and demonstrated a significantly high level of
agreement with physician intensivists for setting PSV
in patients with respiratory failure. Clinical outcome
studies evaluating the novel load-and-tolerance strategy of the advisory system as the means of applying
PSV are needed.
ACKNOWLEDGMENT: The authors thank and acknowledge
the registered respiratory therapists from Cardiopulmonary Services at Shands Hospital at the University of Florida for their
cooperation and technical support.
References
1 Millic-Emilli J: Work of breathing. In: Crystal RG, West JB,
eds. The lung. New York, NY: Raven Press, 1991; 1065–1075
2 Kirton OC, DeHaven CB, Hudson-Civetta J, et al. Reengineering ventilatory support to decrease days and improve
resource utilization. Ann Surg 1996; 224:396 – 404
3 MacIntyre NR. Respiratory function during pressure support
ventilation. Chest 1986; 89:677– 681
4 Otis AB, Fenn WO, Rahn HL. Mechanics of breathing in
man. J Appl Physiol 1950; 2:592– 607
5 Barnes PJ. Neural control of airway smooth muscle. In:
Crystal RG, West JB, eds. The lung. New York, NY: Raven
Press, 1991; 903–916
6 Sant’Ambrogio G, Sant’Ambrogio FB. Reflexes from the
airway, in lung, chest wall and limbs. In: Crystal RG, West JB,
eds. The lung. New York, NY: Raven Press, 1991; 1383–1395
7 Sant’Ambrogio G. Information arising from the tracheobronchial tree of mammals. Physiol Rev 1982; 62:531–569
8 Banner MJ, Euliano NR, Brennan V, et al. Power of breathing determined non-invasively with use of an artificial neural
network in patients with respiratory failure. Crit Care Med
2006; 34:1052–1059
9 MacIntyre NR. Invasive mechanical ventilation in adults:
conference summary. Respir Care 2002; 47:508 –518
10 MacIntyre NR, Ho L. Effects of initial flow rate and breath
termination criteria on pressure support ventilation. Chest
1991; 99:134 –137
11 Volta CA, Marangoni E, Alvisi V, et al. Respiratory mechanics
by least squares fitting in mechanically ventilated patients:
application on flow-limited COPD patients. Intensive Care
Med 2002; 28:48 –52
12 Jolliet P, Tassaux D. Clinical review: patient-ventilator interaction in chronic obstructive pulmonary disease. Crit Care
2006; 10:236 –242
13 Sottiaux TM. Patient ventilator interactions during volume
support ventilation: asynchrony and tidal volume instability; a
report of three cases. Respir Care 2001; 46:255–262
14 Ludbrook J. Statistical techniques for comparing measures
and methods of measurement: a critical review. Clin Exp
Pharmacol Physiol 2002; 29:527–536
15 Belel SY, Taktak AF, Nevill A, et al. An intelligent ventilation
and oxygenation management system in neonatal intensive
care using fuzzy trend template fitting. Physiol Meas 2005;
26:555–570
16 Kwok HF, Linkins DA, Mahfouf M, et al. Adaptive ventilator
Fio2 advisor: use of non-invasive estimates of shunt. Artif
Intell Med 2004; 32:157–159
17 Rutledge GW, Thomsen GE, Farr BR, et al. The design and
implementation of a ventilator: management advisor. Artif
Intell Med 1993; 5:67– 68
CHEST / 133 / 3 / MARCH, 2008
703