METHODS

Section:

ChooseTop of pageAbstractMETHODS <<RESULTSDISCUSSIONReferencesCITING ARTICLES

Detailed methods are available in an online supplement. The medical ICU is a closed 14-bed unit staffed by 2 attendings and 10 M.D. trainees. The physicians work in two teams of six, each team caring for half the patients. All attend structured daily bedside rounds lasting approximately 3 hours. Presentations on rounds are based on a printed template covering each major physiologic system, which is completed by the house staff daily (see online supplement). Most physicians remain in the ICU for the workday, and three house officers stay overnight. The nurse (all registered nurses) to patient ratio is 1:2, plus one to two additional senior nurses and one to two respiratory therapists. The staff is experienced with numerous protocols guiding nonventilatory care and has used ventilator protocols in patients enrolled in the Acute Respiratory Distress Syndrome Network studies (19).

All patients requiring invasive ventilation for 24 hours or more from April 2000 to July 2001 were eligible. Previous participants, those enrolled in other studies that controlled weaning (such as the ARDS Network studies) or transferred from other facilities already intubated, were excluded. The institutional review board waived the requirement for informed consent.

Patients were assigned to a study group by their hospital numbers (odd versus even). The protocol wean (PW) was based on one previously shown to reduce duration of ventilation (3). This consisted of a daily screen for readiness, followed by measurement of the ratio of respiratory frequency to mean Vt (f/Vt) in patients passing the screen. If f/Vt was 105 or less, a spontaneous breathing trial (SBT) was then attempted on continuous positive airway pressure (CPAP), with 5 cm H₂O pressure support added if the endotracheal tube was 7 mm or less. Physicians were told whether the SBT was tolerated for 1 hour. If not, the patient was rested on mechanical ventilation until the next morning (Table 1)

The ICU fellow documented the reason for intubation and presence of chronic respiratory disease on case report forms. A study coordinator contemporaneously reviewed and reinforced staff compliance.

Data on weaning methods were not collected contemporaneously on patients in the UC arm. After completion of the study, a random sample of 50 patient charts in this group was reviewed to extract data on weaning methods. Because measurement of f/Vt was not routinely performed or documented in the UC group, these data are unavailable.

Data are presented as mean ± SD, median (interquartile range), or proportions, as appropriate. The primary outcome was duration of ventilation, defined as the time from intubation to the beginning of the SBT that ended with successful discontinuation of mechanical ventilation. Patients were considered to have successfully weaned if they were able to breathe unassisted for 48 hours. Compliance with the protocol was evaluated by review of the weaning documentation form used by the therapists and nurses and chart review. Analyses were based on intention to treat, with χ² tests or Fisher's exact tests and t tests or Wilcoxon-rank sum tests, as appropriate. Multivariable linear regression evaluated the sensitivity of results to imbalances in patient baseline characteristics; p values of more than 0.05 were considered nonsignificant. The sample size provided 82% power (assuming two-sided type I error = 0.05) to detect a 1-day difference in mechanical ventilation duration. Computations were performed using STATA 7.0 (StataCorp LP, College Station, TX).

RESULTS

Section:

ChooseTop of pageAbstractMETHODSRESULTS <<DISCUSSIONReferencesCITING ARTICLES

During the study period, there were 749 admissions to this ICU, of which 356 (47.5%) required invasive mechanical ventilation for at least 24 hours and hence were eligible to participate. Among the excluded patients, 320 were not intubated or ventilated, and 73 were extubated or died in less than 24 hours. Fifty-three patients who were intubated and ventilated for more than 24 hours were excluded for the reasons shown in Figure 1

Figure 1. Patient-flow diagram. MV = mechanical ventilation; PW = protocol-directed MV discontinuation; UC = usual care (traditional, physician-directed MV discontinuation).

Download Figure | Download Powerpoint

[More] [Minimize]

Compliance with the protocol was high among the 154 patients assigned to PW. Screening was performed on 743 of 863 patient-days (86.1%) of mechanical ventilation, with 130 patients (84.4%) screened every day when they were on mechanical ventilation. Of the 406 patient-days on which all the screening criteria were passed, 343 (84.5%) were followed by a 1-minute chronic obstructive pulmonary disease trial. Among the 223 patient-days on which f/Vt was 105 or less, patients were advanced to a SBT in 88.3% of cases. Ten patients were withdrawn from the PW by their treating physicians, with three (30%) of them dying before ICU discharge.

In the intention-to-treat analyses, there was no difference between groups (PW vs. UC) in the number of patients who successfully discontinued mechanical ventilation before ICU discharge (115 [74.7%] vs. 109 [75.2%], p = 0.92) or their duration of mechanical ventilation (median [interquartile range]: 60.4 hours [28.6–167.0 hours] vs. 68.0 hours [27.1–169.3 hours], p = 0.61; Figure 2)

Figure 2. Duration of mechanical ventilation, by treatment group, among patients who discontinued mechanical ventilation before ICU discharge. The horizontal line is the median. The box denotes 25–75th percentile, and whiskers denote 10% and 90% confidence limits. Protocol group, n = 115; traditional group, n = 109; p = 0.61 by Wilcoxon rank sum.

Download Figure | Download Powerpoint

[More] [Minimize]

Among the 50 patients in the UC group whose charts were randomly selected for review, 34 (68%) were weaned, and 1 was discharged on mechanical ventilation. Fourteen of these 34 patients (41%) were successfully weaned using intermittent T-piece trials. Fourteen were weaned using pressure support, four (12%) by both pressure support and T-piece trials (changing mode after failure to progress), and two patients were weaned using intermittent mandatory ventilation. Sixty-five percent of the successfully weaned patients were extubated after their first SBT (either CPAP or T-piece), 15% after their second, and the remainder required three or more SBTs before successful weaning. Mean ± SD pulse oximetric measurement of arterial oxygen saturation at the initiation of the SBT was 97 ± 2.1%. Six patients, including three with an endotracheal tube larger than 7 mm, had their SBT on low levels of pressure support immediately preceding extubation. Four patients underwent a SBT while on vasopressors. For comparison, among the patients in the PW group who were weaned, 54% required only one SBT, 19.5% required two, and the remainder required three or more. Mean oxygen saturation at the initiation of the SBT was 97 ± 2.4%.

DISCUSSION

Section:

ChooseTop of pageAbstractMETHODSRESULTSDISCUSSION <<ReferencesCITING ARTICLES

We could not document any improvement in clinical outcomes with the use of a nursing/respiratory therapy-driven protocol for discontinuing mechanical ventilation. The protocol we used was similar to that previously shown to speed weaning in medical and coronary ICU patients (3). Our protocol was more aggressive in a few respects. First, patients could be advanced to a f/Vt determination even if they were still on vasopressors. Second, we did not require a physician order to begin a SBT. Finally, we required only 1 hour of spontaneous breathing before discontinuing mechanical ventilation, based on a report that as little as 30 minutes of a SBT may be sufficient to identify patients likely to breathe unassisted (20).

A protocol will be ineffective if it is not followed. Poor compliance with study procedures has been noted in a previous investigation in which a protocol shown to be effective in shortening mechanical ventilation duration in one setting was not effective elsewhere (14). In this investigation, we documented rates of compliance with the study procedures in the intervention group, which compare favorably with similar studies. In other studies of weaning protocols, compliance with the screening step ranged from 60 (5) to 95% (3). The compliance with the SBT step in eligible patients also varied widely in other studies from 10–81% (14). In this study, we cannot distinguish weaning steps that were omitted because of simple oversight from steps that were omitted because of clinical judgment (e.g., a patient undergoing a procedure). Moreover, in this study, only 10 (3.3%) enrolled patients were withdrawn by treating physicians. Thus, we do not believe that inadequate compliance to study procedures explains the difference between this study and others addressing the same question.

A protocol may not be needed if it merely codifies a set of behaviors that are already in use. It is possible that the house staff and attendings had already incorporated some strategies that have been shown to speed weaning into their regular practice. Although we did not prospectively record physicians' approaches to discontinuing mechanical ventilation in the UC group, our chart review found that patients in UC were weaned by a variety of methods. Therefore, it is unlikely that the use of once-daily T-piece trials was an essential element of expeditious weaning. We also did not find evidence of a notably aggressive style of weaning in the UC group, as few of these patients were weaned while on vasopressors or were extubated directly from low levels of pressure support, and arterial oxygen saturation was well above minimum acceptable levels at the beginning of SBTs.

What factors then may have allowed physicians to function as well as the protocol? We believe one simple but important factor is the amount of attention that can be provided to assess patients' readiness to breathe unassisted. Unstable, critically ill patients demand physician attention, and thus, it is nearly unavoidable that stable patients (e.g., those needing only mechanical ventilation discontinuation) fall to a lower priority. When physician availability is a limiting factor in weaning, a protocol can fill that gap. However, when more physicians are available, there may be little chance for a weaning protocol to improve care. To help us further explore this possibility, the authors of previous clinical trials that evaluated protocols for weaning generously provided information to allow us to compare physician staffing levels across studies. We considered the number of ICU beds, the number of physicians assigned to the ICU, and the average number of hours they were present in the ICU each day (during the day and overnight) to calculate the number of physician-hours per bed per day in each of the studies. Compared with approximately 9.5 physician-hours/bed/day in this study, physician staffing was considerably less in the three previous randomized clinical trials: 3.5 physician-hours/bed/day (personal communication, E.W. Ely, M.D., M.P.H.) (3), 4.0 physician-hours/bed/day (personal communication, M.H. Kollef, M.D.) (5), and 4.7 physician-hours/bed/day (personal communication, G. Marelich, M.D.) (6). Although the optimal level of physician staffing for critically ill patients is not known, we speculate that the twofold to threefold increased levels of physician staffing in our ICU may have allowed more timely discontinuation of mechanical ventilation in patients ready to breathe unassisted, rendering the protocol unnecessary. This explanation is supported by results of recent studies reporting an inverse relationship between workload and patient outcomes in ICU and other healthcare settings (16, 21–23).

Another potential explanation is our use of a printed rounding template that covers each of the physiologic systems and then becomes the house staff progress note (reprinted in the online supplement). Similar templates were not in use during other randomized, controlled trials of weaning protocols (3, 5, 6) (personal communications, E. Haponik, M.D.; M.H. Kollef, M.D.; and S. Murin, M.D.). Although the template does not specifically address weaning, it may have helped prompt the team to address ventilator issues each day. In support of this explanation, the use of a simple checklist on rounds was recently shown to reduce the length of stay in a surgical/oncologic ICU (17).

Our findings might suggest that protocol-directed weaning would reduce needs for costly physician staff time. Protocols generally address focused and specific problems that are common, formulaic, and relatively straightforward. However, adequate numbers of skilled and sentient staff can promptly recognize and treat multiple subtle or complicated issues that may not be suitable for management by protocol. Adequate nursing staff, like protocols, have been shown to improve ICU outcomes, including discontinuation of mechanical ventilation and mortality (16, 21–24). In this study, the observed hospital mortality rates of 36.4% and 33.1% in PW and UC groups compare favorably with the mortality rates predicted by their Simplified Acute Physiology Score II severity of illness score on ICU admission (50.5%, p = 0.02, and 48.8%, p = 0.008, respectively; p values for observed vs. predicted in each group). Intensivist-led or closed ICUs have been shown to improve ICU mortality (15, 16). For these reasons, we do not believe that ICU protocols can substitute for qualified physician and nursing staff. Greater staffing levels are expensive. However, costs of insufficient staff, although hidden to hospital administrators, may be greater.

Our study has several potential limitations. As in other similar unblinded studies (3, 5, 6, 9), it is possible that some physicians, nurses, or respiratory therapists may have changed their UC ventilator management practice because they knew the study was underway (Hawthorne effect). If they reverted to less salutary practice after the trial period ended, the effectiveness of the protocol during a more typical period might have been obscured.

Our allocation process by even or odd medical record number was only quasirandomized and therefore potentially subject to selection bias. Medical record numbers are assigned administratively by a computer in order of a patient's first contact with our medical system, without regard to age, sex, race, diagnoses, ICU admission, etc. This system allowed patients to be assigned to a group immediately on admission without intervention by the investigators and contributed to there being few patients who were inadvertently overlooked (just 4 of 749 admissions, 0.5%). The small number of subjects (six) who were excluded by their treating physicians also suggests that there was no bias to the treatment assignments.

There was a clinically unimportant but significantly longer interval between the initiation of the final SBT and extubation in the patients weaned by protocol. We believe we understand why this occurred. In PW patients who successfully completed 1 hour of a SBT, nurses were often reluctant to interrupt attending physician rounds (generally from 8–11 a.m.) for an order to discontinue mechanical ventilation. Physicians on rounds, if interrupted, were also likely to postpone a decision if that patient had not yet been reviewed by the team. In contrast, physicians were likely to review and promptly act on the results of a SBT that they had initiated in patients assigned to UC.

Finally, our data on the UC group are limited by the retrospective chart review. It is difficult to infer the factors motivating physicians in that group, and the influence of staffing levels, the rounding template, or other factors remains speculative and subject to prospective confirmation.