Predicting Eventual Success or Failure to Wean in Patients Receiving Long-term Mechanical Ventilation

Predicting Eventual Success or Failure to Wean in Patients Receiving Long-term Mechanical Ventilation* Eric H. Gluck, MD, FCCP; and Linda Corgian, RN, PhD

Purpose: To determine if the ultimate ability of a long-term ventilated patient to wean can be predicted at the time of his or her admission to a long-term ventilator unit. Design: Two-phased prospective study. Setting: Long-term ventilator facility, university-affiliated. Subjects: Adults ventilated for an average of 3 weeks, who did not have sepsis and who did not have chest tubes or progressive neurologic impairment. Interoentions: On admission to the long-term ventilator unit, historic factors, ventilator history, and the foUowing laboratory and metabolic tests were obtained: electrolytes, serum calcium, magnesium, and phosphorus, WBC, hemoglobin, albumin, total protein, transferrin, oxygen consumption, carbon dioxide production, respiratory quotient, and dead space/tidal volume. The patients were then placed in a weaning protocol utilizing increasing duration of pressure support ventilation during the day with complete rest at night. Forty-two days after enrollment in the study, representing three times the duration of the weaning protocol, the patients who successfully weaned were compared to those who remained ventilator dependent (n=20). Patients who died or were transferred to another institution were excluded from this phase of the study, because we were trying to develop parameters that would be predictive of successful weaning. A parameter was considered to be predictive, and retained for the scoring system, if it produced at most 15% false-positives and false-negatives. A score of 0 was then assigned to the threshold value that produced no false-positives; 2 to the threshold value that produced no false-negatives and I to the intermediate values. The scoring system was then applied to a new prospective group of patients (n=72). Measurements and main results: Of aU the parameters evaluated, only the foUowing satisfied the false-positive and false-negative requirements; static compliance, airway resistance, dead space to tidal volume ratio, PaC0 2 , and frequency/tidal volume. Applying these, in the scoring system, to the initial group of patients, demonstrated that a score greater than 3 was associated with failure to wean; a score less than 3 was associated with successful weaning, and a score of 3 was not predictive. Using these thresholds, the data were applied to the new prospective group of patients, which again demonstrated that a score of greater than 3 was associated with failure to wean in aU cases. A score less than 3 was again associated with successful weaning but there were two false-positives. The sensitivity, specificity, and positive predictive and negative predictive values for the scoring system were 1.0, 0.91, 0.83, and 1.0, respectively. None of the individual parameters included in the scoring system demonstrated equivalent statistical results. AU but two of the patients who died prior to finishing the weaning period had weaning scores, which suggested that they would not be successfully weaned. Conclusions: Parameters that are generaUy available, when combined into a scoring system, can predict at the time of admission to a long-term ventilator unit, in most cases, whether a patient will eventuaUy wean. The scoring system resulted in no false-negatives and an acceptable number of false-positives. None of the individual parameters were as reliable as the scoring system as a whole. (CHEST 1996; 110:1018-24) Key words: long-term mechanical ventilation; protocols; weaning Abbreviations: CPAP=continuous positive airway pressure; RSB=rapid shallow breathing index

*From the Vencor Hospital, Northlake, Ill, and North Chicago VA Medical Center, North Chicago, Ill; Finch University of Health Sciences!fhe Chicago Medicii! School (Dr. Gluck); and RushPresbyterian-St. Lul
To handle the increasing load of patients requiring long-term ventilator support, institutions devoted solely to the care of these patients are cropping up all over the United States. By grouping patients with similar needs together, the costs of caring for these patients can be significantly reduced without the loss Clinical Investigations in Critical Care

Table 2-Initial Parameters Evaluated

Table !-Demographics of the Patient Population

Parameters

Characteristic Etiology of respiratory failure COPD Pulmonary fibrosis SIP ARDS* SIP CPR* Neurologic disease Other Unknown Age, yr Duration of ventilation prior to transfer, d Tracheotomy prior to entrance into study, d

58% 10%

8% 10%

8% 3% 3% 67.4±14.9 23.9±8.9 12.4±6.4

*SIP=status post; CPR=cardiopulmonary resuscitation.

of quality of care. While the cost of caring for these patients is reduced considerably under these new circumstances, it is by no means inexpensive care. Medical expenditures in the millions of dollars will be spent during the next fiscal year at these institutions. Further means of reducing the cost for caring of these patients should be sought. One means of reducing cost would be to allocate the services available at these institutions to the patients who are most likely going to benefit from these services. To do this, we must be able to determine who can eventually become ventilator independent (totally or partially) at or near the time of transfer to the long-term care facility. Our study was designed to determine if there exists a group of parameters that can predict, with appropriate sensitivity and specificity, those patients who, although currently receiving long-term ventilator support, will eventually wean. In this study we evaluated numerous historic, laboratory, and physiologic parameters on an initial group of patients to determine which, if any, of these parameters might be predictive of success or failure in weaning these patients. Having established threshold values for the appropriate parameters, we then developed a scoring system that utilized them and applied this system to a prospective group of patients. It was believed that no one single parameter would be predictive in all circumstances because of the various causes of long-term ventilator dependence that we had encountered in our patient population. The data suggest that we were able to satis£)' the goals of the study. Reallocation of available services may be possible in long-term ventilator units. MATERIALS AND METHODS

All patients who entered Vencor Hospital, Chicago, or the North Chicago VA Medical Center who had been ventilated for a minimum of 3 weeks were eligible to be studied as long as they did not exhibit signs of sepsis, which included the following: a temperature greater than 38°C or less than 35.5°C, a WBC count of greater than 12,000 mm 3 , or a documented source of infection. Patients with progressive neurologic impairment or expected survival ofless than 2 months were also excluded. Pulmonary function tests were not available on any of the transferred patients. In view of the

Historic factors Duration of mechanical ventilation Initial episode of respiratory failure Documented COPD Pulmonary fibrosis Cancer Congestive heart failure Renal failure Depressed mental status Initial ventilator settings Mode Tidal volume Respiratory rate Inspiratory flow I:E ratio Positive end-expiratory pressure Initial ventilator measurements Peak inspiratory pressure Plateau pressure Minute ventilation Dead space/tidal volume Maximum inspiratory force Rapid shallow breathing index at 1 min and 5 min Laboratory data Serum sodium Serum potassium Serum chloride Serum calcium Serum phosphorus Serum magnesium WBC count Hemoglobin Serum bicarbonate Albumin Total protein Transferrin Metabolic study Oxygen consumption Carbon dioxide production Respiratory quotient

Table 3-Weaning Protocol Protocol

Time

CPAP plus pressure support

10 10 10 10

CPAP plus pressure support

8 8 5 5 5

CPAP plus pressure support

Flow-by

for for for for for for for for for for for for for for

2h 4h 8h 12 h 8h 12 h 4h 8h 12 h 4h 8h 12 h 16 h 24 h

change in clinical status resulting in the need for long-term ventilatory support, we assumed that this information would have been difficult to evaluate. The Investigational Review Boards for human CHEST/110/4/0CTOBER, 1996

1019

Table 4-Scoring System Score Parameter*

0

l

2

HSB(f!Vt) VoNT Compl(st) Hesistance PaC02

<120 <0.64 >36 <9

120-180 0.64-0.74 32-36 9-17 >64

>180 >0.74 <32 >17

*VoNT=dead space to tidal volume ratio; compl(st)=static lung compliance. research approved the study and waived the requirement that we obtain informed consent prior to studying the patients. The study was divided into two parts, the "training patients" and the prospective patients. All patients in this study had tracheotomies performed prior to being admitted to Vencor. The demographics of the patient population are depicted in Table l. Training Patients

The first 20 patients entered into the study served as the training group to determine the discriminating parameters and their threshold values from the list of parameters depicted in Table 2. Because this population has not been studied extensively as yet, it was believed that standard values for many of the weaning parameters might need to be modified. These potentially useful parameters were obtained by reviewing two textbooks of critical care medicine and pulmonary medicine that had chapters relating to the weaning of patients from mechanical ventilation 1 ·2 The 14-day weaning protocol depicted in Table 3 served as a basis for the gradual removal of ventilatory support for all patients. The data were collected on the day of hospital admission. The ventilator parameters were those chosen by the transferring physicians and modified only to normalize oxygenation and acid-base status. Tidal volumes were chosen based on lean body weight using 7 to 10 mllkg as a guideline. Peak inspiratory pressure, minute ventilation, and respiratory rate were obtained from the ventilator indicators. Plateau pressure was obtained by switching the patient to a square wave inspiratory pattern combined with a 0.5-s inspiratory pause. The resultant plateau pressure was then read off the ventilator manometer. Compliance was calculated by dividing the tidal volume delivered during the plateau maneuver by the plateau pressure minus any positive end-expiratory pressure that the patient was receiving. No attempt was made to correct for the distention of the ventilator circuit. Because all patients were using the same type of disposable ventilator circuits, we believed this correction was

I• Weaned

N

~ Not W eaned

I

unnecessary. Airway resistance was calculated by dividing the difference between the peak and plateau pressure by the inspiratory flow, again using the square wave inspiratory flow pattern. Three measurements for the compliance and resistance were made and the average value was recorded. The maximal inspiratory force was measured by using a standard airway occlusion technique under the vigorous coaching of a respiratmy therapist. The patient was then removed from all ventilator support and the frequency/tidal volume was measured at 1 min and at 5 min. At the completion of 5 min off ventilator support, an arterial blood gas determination was obtained. A metabolic study was done during the first night after hospital admission while the patient was asleep. Dead space/tidal volume was measured using the data obtained from the metabolic cart and an arterial blood gas. The laboratory data that are depicted in Table 2were obtained on the first day of hospitalization and were measured using standard automated techniques. All electrolyte imbalances present at the time of hospital admission were corrected. Nutritional parameters were measured and the patients were fed according to the data obtained from the metabolic cart studies. The weaning protocol depicted in Table 3 was used to wean all patients in the study. In this manner, we could ensure that all the patients were treated identically after they were evaluated and entered into the study. We are not implying that this is the only way to wean patients who are ventilator dependent for prolonged periods of time. Patients were considered to be weaned if they could tolerate 48 consecutive hours without any pressure or flow suppmt from a mechanical ventilator. Patients were considered to be unweanable if they did not meet the above criteria after 42 days of weaning trials (ie, 3 times the duration oftheweaningprotocol). The 42-day limit was initially chosen based on our prior experience with patients in this unit. This limit was eventually validated during this study as no patients who were still ventilator-dependent weaned after this time period. Subjects in the training group of patients who died during the training period prior to the 42-day limit were excluded from the study because the purpose of the training group of patients was to identifY those parameters that were predictive of success in weaning. After obtaining data for 20 patients, each parameter was evaluated with respect to its ability to predict success or failure and threshold values were obtained. For parameters with dichotomous outcomes (yes or no), a threshold value of 65% was chosen. The binomial theorem for dichotomous outcomes predicts that the likelihood of a result being present or absent 65% of the time due to chance alone is approximately 10%. 3 Continuous variables were considered acceptable if they resulted in, a t most, 15% false-positives or false-negatives. Using the same theorem as above, the likelihood of a false-positive or false-negative result occurring by chance alone would be less than 5%. All other parameters were excluded from consideration. The parameters that met these criteria were used to create the weaning scoring system depicted in Table 4. The creation of a weaning scoring system is similar to the premise behind the acute physiology and chronic health evaluation (APACHE) II Severity of Illness scoring system 4 The scores for each of the patients were then calculated. Scores of 4 to 6 were associated with failure to wean. Scores of 0 to 2 were associated with successful weans and a score of 3 was indeterminate (Fig 1). Prospective Study

0

2

4

5

6

Weaning Score

FIGURE l. Score vs weanability in the training group. 1020

The above protocol was then applied to a prospective group of patients (n= 72). Each of these patients was evaluated and a weaning score was determined on admission to the institution. Patients were excluded from analysis based on the above-mentioned criteria. The clinicians who were caring for the patient were blinded to this information. Once the patient met the criteria for ventilator independence or, after 42 days, whichever came first, the outcome was compared with the prediction from the scoring system. StanClinical Investigations in Critical Care

Table 5-Scores and Outcomes

dard formulas were used to calculate sensitivity, specificity, positive predictive value, and negative predictive value. Patients who failed to wean were followed up for at least 6 months. None of these patients were successfully weaned during this period of follow-up.

No. of Patients With a Given Weaning Score Group

RESULTS

Training Group of Patients Twenty patients served as the training group for the study. Twenty-four patients were excluded based on the above criteria. None of the historic data reached the threshold limit of being present or absent 65% of the time. None of the serum chemistry values obtained were shown to be predictive, although serum albumin showed a strong trend in this direction. The parameters that met the threshold criteria were as follows: rapid shallow breathing index (RSB) measured 5 min after the beginning of a trial of continuous positive airway pressure (CPAP), dead space/tidal volume ratio, static compliance, airway resistance, and arterial Pco 2 measured after the CPAP trial. The scoring system depicted in Table 4 was created by assigning a score of 0 to the value of the parameter below which there were no false-positives and a score of 2 to the value of the parameter above which there were no false-negatives. Intermediate values were assigned a score of l. The cutoff values for each of the parameters are again depicted in Table 4. A weaning score greater than 3 was associated with failure to wean and had no false-negatives, ie, a negative predictive value of 100%. A weaning score less than 3 was associated with a successful wean with no false-positives , apositive predictive value of 100%. A score of 3 was nondiscriminatory. Seventy percent of the patients entered in the study could be correctly categorized without any false-negatives. The mean score for the patients who weaned was 1.8±0.249, and for those who failed to wean it was 4.25±0.335. These values were shown to be statistically distinct using the Student's t test (p value <0.01). All the patients in this group who weaned were completely liberated from ventilator support. Prospective Patient Group Seventy-two patients were evaluated on admission to the long-term ventilator unit. Nine patients died of sepsis prior to the onset of weaning. Seven patients were transferred to another institution. Three patients died due to ventilator or other mechanical failure. Fifteen additional patients started weaning but died prior to reaching the 42-day limit of the study. Eleven of these patients died from sepsis, two died due to cardiopulmonary arrest, and two died of acute neurologic catastrophe. The scores of these patients and the outcomes are reviewed in Table 5. Thirty-eight patients finished the 42-day time period in the prospective group of patients: 14 patients weaned and 24 patients did not. Of the 14 patients who weaned, 1

1

0

Died prior to the beginning of weaning (9) Died due to mechanical failure (3) Transferred to another site ( 7) Died prior to completing the 42-day weaning period (15)

2

3

1 2

5 6 2

2

3

2

2

2

4

1 1 1 5 2

2

1

4

1

required partial ventilator support from 10 PM to 6 AM consisting of flowby; the rest of the patients were completely liberated from ventilator support. A weaning score of greater than 3 was associated with failure to wean in all cases. A score of 3 or less was associated with successful weaning in 14 cases and 6 falsepositives (Fig 2). A score of less than 3 demonstrated 10 successful weans and 2 false-positives. Thus, the sensitivity of the scoring system was 100% while the specificity BS value was 77%, using 3 as the threshold value. Using a score of less than 3 as the prediction for success and a score of greater than 3 as a prediction of failure, the sensitivity and specificity become 100% and 91%, respectively, with a value of 3 not being discriminatory. The positive predictive value of a score of less than 3 was therefore 83% and the negative predictive value for a score greater than 3 is 100%. These calculations, for each of the individual parameters of the scoring system, are shown in Table 6. Of the individual parameters, only lung compliance approached the scoring system as a whole, with respect to sensitivity and negative predictive value. The mean score for the prospective patients who weaned, the prospective patients who failed to wean, the patients in both the training and prospective group who weaned, and the patients in both the training group and prospective group who failed to wean are, respectively, 1.8±0.25, 4.25±0.34, 1.41±0.23, and 4.31±0.25. The differences among these groups was statistically significant

N u

10 9

m 8 b

e

7 6

5 0

f

p t

4 3 2

1 0

0

3 4 2 Weaning Score

5

6

FIGURE 2. Weaning score vs success or failure in prospective patients.

CHEST I 110 I 4 I OCTOBER, 1996

1021

Table 6-Prospective Patient Data* Index

Sensitivity

Specificity

PPV

1'\PV

Nondescriminatory

l.OO

0.25 0.73 0.66 0.30 0.77 0.91

0.60 0.40 0.66 0.50 0.66 0.83

NC 0.73 0.89 1\'C

16 12 9 17 0 6

RSB VoNr C(st) Raw Scoring system Modified scoring system

0.40

l.OO l.OO l.OO l.OO

l.OO l.OO

*VDNT=dead space to tidal volume ratio; C(st)=static lung compliance; PPV=positive predictive value; NPV=negative predictive value; 1\'C=NPV could not be calculated.

using the Student's t test (p value <0.01). None of the patients who failed to wean during this study became ventilator-independent during a follow-up period that averaged approximately 6 months. Two of the successfully weaned patients developed respiratory failure and required renewed ventilator support, 1 after 8 days of spontaneous ventilation and 1 after 25 days. Excluding only the patients who were transferred to another institution, there were 3 patients who did not reach the 42-day time limit who had a weaning score less than 3. One of these patients died from a mechanical failure, one died from sepsis, and one died from a neurologic catastrophe. DISCUSSION

The purpose of this study was to develop a means of predicting the likelihood of a successful wean in patients who are admitted to a long-term ventilator unit after having received prolonged mechanical ventilation. In this way, an appropriate matching of available services to those patients who are likely to benefit from those services could be made. Because the etiologies of respiratory failure in this patient population are varied, we reasoned that a single parameter would not be predictive in a sufficient number of cases. The optimal characteristics of such a scoring system should include, among others, measurements that are made easily at the bedside without incurring a large expense. The test or tests should be applicable soon after hospital admission, after an initial period of stabilization and correction. The tests should be reproducible. The parameters selected should be interpreted easily without subjective bias. False-negatives must be kept to an absolute minimum. The tests should discriminate and be applicable to the vast majority of patients likely to be in the system. In this study, we allowed the patient population to naturally select the parameters that would meet some, if not all, of the above requirements. This is not the first attempt to establish a scoring system for patients receiving mechanical ventilation, but as best as we can determine, it is the first attempt to apply such a system to this patient population and for this intended purpose. Morganroth et al 5 did a retrospective evaluation of 11 patients who had un1022

dergone prolonged mechanical ventilation. They proposed an adverse factor score that would help delineate those patients who were capable of weaning. This score consisted of 21 variables measured every 3 days while the patient was receiving ventilatory support. To this, a ventilator score that included six additional variables was added. The initial scores did not predict future weaning success or failure, but those patients who gradually weaned showed a significant decline in these scores. The study was never followed by a prospective evaluation in a new subset of patients. Most recent studies that evaluate weaning parameters do so individually, and most suffer from the fact that the data are obtained solely from the training group with no prospective follow-up for corroboration. 6-9 One notable exception to this was the study by Yang and Tobin. 10 Their study demonstrated that the commonly applied indices of weaning, when applied individually, were not predictive of success or failure of extubation. Their index of RSB and an additional index that included measurements of compliance, resistance, oxygenation, and respiratory muscle strength did demonstrate the ability to predict those patients who would tolerate extubation. Their study differs from this one in two important aspects. They applied their indices to patients who were ready to be considered for extubation and their patient population had been mechanically ventilated for a shorter period. Their ultimate goal was also different than ours. Herein we wanted to be able to predict as soon as possible which patients had the potential for ventilator independence so that this information could be used to establish more realistic patient care ratios. The success of this scoring system rests in its ability to determine, shortly after admission to a long-term ventilator unit, those patients with the highest likelihood of weaning. In addition, it does so with no falsenegatives and without sacrificing sensitivity. Excluding those patients who received a weaning score of3, 93% of the patients had their eventual outcome at the end of the protocol predicted correctly. However, even when those with a score of 3 are included in the successful group, 84% of the patients were appropriately classified at the time of hospital admission. None of the patients who were misclassified would have been Clinical Investigations in Critical Care

placed in the predicted "not able to wean group," rather the scoring system has a slight tendency to overestimate the ability of a patient to wean. The tests utilized are readily available and easily performed at the bedside using standard equipment. The tests are easy to interpret and are reproducible. A potential difficulty lies in the patient population used in this study. While we believe this to be representative of patient populations at other long-term care units around the country, different f1nancial constraints and clinical practices could result in a more or less acute patient population at a given center. As an example of this phenomenon, the study by Gracey et al 11 demonstrated a significantly improved outcome when compared to the patients described herein. This can be accounted for by the highly selective admission criteria to which their patients were subjected. Many of the patients described herein would not have been accepted for admission to their unit. Recent investigations have been concerned with the mode of ventilation that is most likely to be successful in weaning patients from mechanical ventilation. Two studies in particular reached different conclusions as to the most effective weaning mode, T-piece in one and pressure support in the other.l 2 •13 Neither of these studies employed a patient population similar to the one in our study. At the present time, it is impossible to discern whether the results of our study would change significantly if a different weaning protocol were employed. The successful parameters were self-selected by our patient population during their application to the training population. Therefore, no subjective bias was introduced with respect to the ultimate selection of parameters. By then applying these parameters to the prospective group of patients, we provide validation of these parameters. We tried to ensure that all patients were given ample opportunity to wean by allowing them 6 weeks of weaning after admission to our unit. Although a 42-day limit to weaning is somewhat arbitrary, the fact that no one weaned after this time period, despite continued effort to do so, validates this to some degree. Similar mean scores for both groups of the training patients and the prospective patients suggest that the population characteristics did not change during the study. It is interesting to note that the vast majority of the patients who did not survive to the end of the weaning period had scores that would not have predicted weaning success. The scoring system may have some utility as a prognostic agent in addition to its use in a weaning protocol. This, however, would require additional study to validate. It is interesting to evaluate the parameters that were self-selected by this study as being predictive of success or failure at weaning. The parameters themselves de-

fine different aspects of lung function and physiology. With various etiologies oflung impairment as potential causes of long-term ventilator dependency, it is not surprising that a representative group of weaning parameters such as these were necessary to achieve the necessary sensitivity, positive predictive values, and negative predictive values. Combining these parameters into a weighted scoring system prevents one of the parameters from assuming a more influential role. The anticipated final phase of this study will be to apply this weaning scoring system to another prospective group of patients and then, based on their predicted ability to wean, simulate a reallocation of the available services. Appropriate fail-safe interventions can be developed to reduce the likelihood that a patient would remain, inappropriately, in a nonweaning area of the unit. We have estimated that reallocation of services could reduce the cost of caring for these patients by more than $200 per day in hospital costs alone. In addition, there would be cost savings in other areas as well, such as physician charges. It is estimated that there are about 9,000 long-term ventilated patients in the United States at the present time. Even if only 30% of these patients were delegated to the nonweaning group, a savings of more than $400,000 per day could be achieved. In summary, we have demonstrated that a group of physiologic parameters with appropriate threshold values, determined on an initial group of ventilatordependent patients, could be applied to a prospective group of patients, and used to predict success or failure of eventual weaning with reasonable sensitivity, positive predictive values, and negative predictive values. The parameters, self-selected by the initial group of patients, evaluate different aspects oflung function, a necessary condition when dealing with respiratory failure from various causes. The parameters are inexpensive to obtain and easily measured at the bedside. This scoring system could then be used to reallocate available services to those patients who are most likely to benefit from them. REFERENCES

1 Parrillo J. Current therapy in critical care medicine. 2nd ed. Hamilton, Ontario: BC Decker, 1991; 204 2 Fishman A. Pulmonary diseases and disorders. 2nd ed. New York: McGraw Hill, 1992; 2383 3 Mattson DE. Statistics-difficult concepts, understandable explanations, revised Ed. Oak Park, Ill: Bolchazy-Carducei, 1986; 6870, 151-53 4 KnausWA, DraperEA, WagnerDP,etal. APACHE II: a severity of disease classification system. Crit Care Med 1985; 13:818-29 5 Morganroth M, Morganroth J, Nett L, eta!. Criteria for weaning from prolonged mechanical ventilation. Arch Intern Med 1984; 144:1012-16 CHEST I 110 I 4 I OCTOBER, 1996

1023

6 Pierson D. Weaning from mechanical ventilation. Respir Care 1983; 28:646-60 7 Tahvanainen J, Salmenpera M, Nikki P. Extubation criteria after weaning from intermittent mandatory ventilation and continuous positive airway pressure. Crit Care Med 1983; 11:702-07 8 Fiastro JE, Habib MP, Shon BY, et al. Comparison of standard weaning parameters and the mechanical work of breathing in mechanically ventilated patients. Chest 1988; 94:232-38 9 Hubmayr RD, Loosbrock LM, Gillespie DJ, et al. Oxygen uptake during weaning from mechanical ventilation. Chest 1988; 94:1148-55 10 Yang K, Tobin M. A prospective study of indexes predicting the

1024

outcome of trials of weaning from mechanical ventilation. N Engl J Med 1991; 324:1445-50 ll Gracey DR, Naessens JM, Viggiano RW, et al. Outcome of patients cared for in a ventilator dependent unit in a general hospital. Chest 1995; 107:494-99 12 Esteban A, Frutos F, Tobin M, et al. A comparison of four methods ofweaning patients from mechanical ventilation. N Engl J Med 1995; 332:345-50 13 Brochard L, Rauss A, Benito S, et al. Comparison of three methods of gradual withdrawal from ventilatory support during weaning from mechanical ventilation. Am JRespir Crit Care Med 1994; 150:896-903

Clinical Investigations in Critical Care