Is Automated Weaning Superior to Manual Spontaneous Breathing Trials?

Evidence in Support of Automated Weaning Modes: Pro

Automated weaning modes first appeared on ventilators more than 2 decades ago. However, few scientific studies supporting manufacturers' claims of improved outcomes accompanied their appearance. An early examination of the evidence at the time conceded that in fact “there is none.”¹ Lack of randomized controlled trials, misunderstanding of mode functionality, and the sometimes overzealous marketing left many reluctant to incorporate potentially beneficial technology and, even worse, encouraged caregivers to continue to employ inferior weaning modalities.² This led clinicians and leaders in respiratory care to call for clear scientific evidence of improved outcomes before the acceptance any new mode and to reevaluate the continued use of traditionally unproven modes, such as synchronized intermittent mandatory ventilation.³ In response, both for academic and market-driven reasons, a considerable amount of research regarding the efficacy of automated weaning modes has since been conducted, reviewed, and evaluated by recognized experts in the field, professional medical organizations, and government regulatory bodies. Their findings have shed light upon the benefits and potential pitfalls of automated modes and offer recommendations as to how they may be best utilized to improve patient care.

Although an in-depth review of the totality of evidence is beyond the scope of this discussion, several recent publications warrant review. Branson⁴ performed a comprehensive appraisal of closed-loop modes, including the evidence for and against their use to facilitate weaning, in 2012. He found insufficient evidence to support the use of several modes for weaning, although improved oxygenation, airway pressures, number of ventilator manipulations, and identification of weaning readiness favored automated modalities compared with traditional modes. The majority of comparative clinical trials have been conducted using Adaptive Support Ventilation (Hamilton Medical) and SmartCare/PS (Table 1). In adult subjects with COPD, ASV reduced weaning time from 3 to 1 d, whereas in a mixed population, SmartCare/PS (Dräger Medical) shortened weaning duration from 5 to 3 d. Similar improvements were seen in the pediatric population, with reductions from 6.7 to 5.1 d. Other studies had mixed results without improvement in weaning time. Nonetheless, no differences in sedation use, ventilator manipulations, or adverse events were recorded, suggesting that at the very least, automated weaning poses no greater risk than and is equivalent to conventional weaning.

The Canadian Wean Earlier and Automatically with New technology trial (the WEAN study) is the largest multi-center, randomized trial conducted in North America to date.⁸ It compared a standardized weaning protocol with automated weaning in a multidisciplinary population and evaluated clinician, respiratory therapist (RT), and nursing compliance with and perceptions of automated weaning. Median time to the first successful breathing trial, extubation, and successful extubation were all significantly shorter (1–3 h, P < .01–.02) in those receiving automated weaning, with fewer tracheostomies and episodes of extended ventilation. Furthermore, less use of sedation and analgesia and fewer days with extreme sedation scoring suggest that automated weaning may facilitate sedation management. Clinician and RT protocol acceptance scores were lower in the automated weaning group, attributed to the reluctance to relinquish ventilatory control and resistance to the implementation of a new technology.

Finally, a large systematic review and meta-analysis by Rose et al⁹ compared conventional and automated weaning in children and adults. The Cochrane Collaboration quality standards were applied to randomized controlled trials from electronic databases, conference proceedings, and original articles, and 16 studies utilizing protocolized weaning as a comparator were identified. Overall, SmartCare/PS reduced weaning by a geometric mean of 30% in mixed/medical, but not in surgical, ICU populations. There was no difference in weaning duration with any other modes; however, reductions in time to extubation, prolonged intubations, tracheostomies, and ICU stay all favored automated weaning.

As with any promising new technology, there is a need for continued evaluation and development. Limited data exist in the pediatric population, and further research is needed before widespread acceptance. Other populations, including neurological, surgical, and ARDS/refractory hypoxemia also deserve additional attention, since many will require prolonged weaning. However, the evidence to date supports the conclusion that automated weaning modalities are safe and may provide important benefits beyond simply speeding the weaning process.

Automated Weaning Will Improve Best Practice

Evidence-based medicine and quality improvement are the methodologies for translating knowledge into best practice. Evidence-based medicine focuses on “doing the right things” based on available evidence from research, whereas quality improvement focuses on “doing things right” through systems and processes.¹⁰ Successfully implemented protocols and guidelines that embrace evidence-based medicine and quality improvement concepts have clearly been shown to reduce practice variability, improve outcomes, decrease costs, improve safety, and augment education.^11–13 Although weaning protocols have been widely endorsed and adopted as standard of care in many ICUs,^14–17 adherence to guidelines and protocols continues to be a major hurdle even under the scrutiny of research.¹⁸ This is due to the fact that no matter how explicitly designed or enthusiastically endorsed, the success of any non-automated protocol is largely dependent on human interaction.

Failure to follow a weaning protocol may in part be due to lack of knowledge. Physician or ICU staff may be responsible for a patient being weaned with little or no formal training on the weaning protocol in place. The emergence of telemedicine and the Electronic ICU has resulted in an increase in complex patients managed remotely and at smaller facilities where experience with ventilated patients may be limited. Even experienced personnel require periodic retraining to maintain proficiency.¹⁹ Process of care and ICU structure recommendations call for the use of standardized protocols by well-trained multidisciplinary teams.²⁰ Implementation of computerized protocols for automated weaning will improve best care, at least in part, by reducing errors caused by inexperience and lack of knowledge.

Successful weaning during manual spontaneous breathing trials (SBTs) depends upon the practitioner's ability to recognize physiologic readiness of the patient for trial initiation, time the trial with other interventions such as a sedation holiday, and appropriately recognize, in a timely fashion, when the patient has either passed or failed the SBT. Recognition of readiness can be difficult even for experienced clinicians. A study by Stroetz and Hubmayr²¹ revealed that in subjects who had failed previous weaning attempts, the ability of physicians to predict successful weaning was only 50%, with physician bias toward continuation of unneeded ventilatory support. In contrast, closed-loop algorithms follow very specific rules in which parameters are adjusted based upon individual patient feedback and are comparable with those selected by experienced clinicians.^22,23 They assist the practitioner in doing “the right thing” with the additional benefit of providing non-distractible, non-fatigable attention to detail beyond human capability.

Well-established objective clinical criteria provide guidance as to when a patient may be ready to undergo a weaning trial. The use of combined sedation and weaning protocols has been shown to shorten the weaning process^16,24; however, its success is totally dependent upon the coordination of the “sedation holiday” and return of spontaneous breathing with the availability of a practitioner to evaluate and make appropriate changes in a timely fashion. Delays in adjusting ventilator parameters during weaning may cause patient-ventilator asynchrony, reinstitution of sedation, and loss of the window of opportunity for successful extubation.²⁵

Real world circumstances routinely prevent even the best-intentioned practitioner from doing “the right thing at the right time.” Increased work load, family interactions, emergencies, patient admissions, transports, and other distractions are just a few examples of things that may interfere with the practitioner's bedside presence when needed. The advantage of automated weaning modalities is their ability to provide full support before weaning, recognize spontaneous efforts almost immediately, and continuously assess and adjust settings up or down accordingly based on individualized patient feedback, without vulnerability to outside distractions.

Automated Weaning Will Reduce Work Load

Staffing shortages and the increasing number of patients requiring mechanical ventilation may be the most compelling argument for automated weaning. It has been demonstrated that work load has a direct effect on patient outcomes.^26,27 Current respiratory therapist workforce shortages already exist, as highlighted in a 2006 survey of 30 hospitals.²⁸ They reported an average staffing level of 10.8 beds/respiratory therapist, exceeding the accepted preferred ratio of 9.4, with a perceived need for an additional 1.3 respiratory therapists in current staffing levels. The ratio of respiratory therapists to ventilated patients was not reported; however, some states have set the maximum ratio at 1:4.²⁹ To meet this requirement, some hospitals have adopted unpopular mandatory overtime policies. Of the 30 hospitals surveyed, 7 utilized mandatory overtime on at least a monthly basis. Since automated weaning modes have been shown to decrease the number of ventilator manipulations, their use should translate into a reduction in work load for already overburdened staff. Future projections of the need for mechanical ventilation estimate an annual growth rate of 5% due mainly to the aging baby boomer population.³⁰ Those requiring prolonged mechanical ventilation consume two thirds of total hospital resources devoted to mechanical ventilation. Their numbers are expected to double by 2020, with projected costs reaching $60 billion.³⁰ The average additional daily ICU cost of patients requiring mechanical ventilation is $1,500.³¹ Delayed extubation increases the incidence of complications with increases in hospital costs of $29,057/incidence.³² Although the price of adding automated weaning software may be several thousand dollars per ventilator, a reduction of only 1–2 ventilator days in just a few patients would easily offset the additional expense.

Last, there is the threat of a mass casualty disaster. Significant financial resources and logistical preparation have been expended in planning for various scenarios in which personnel and supply resources will be stretched to their limits.³³ Although ventilators have been purchased and stockpiled for future mass casualty emergencies, the availability of personnel with sufficient skills to manage them is of concern,^33,34 reinforcing the vital role for respiratory therapists in disaster planning. One possible solution would be to have “respiratory extenders” oversee basic ventilator operation under the supervision of a therapist,³⁵ liberating skilled RT practitioners to provide valuable respiratory services to the more critically ill. Automated weaning modalities would enhance this concept. By having at least a portion of the current ventilator inventory in use equipped with automated modalities, the need for manual manipulations would be reduced, and the potential for errors from inadequately trained personnel would be minimized.

Automated weaning modalities are superior to manual spontaneous trials in their ability to recognize changes in breathing patterns and adjust settings quickly and without human intervention. Their utilization should not be viewed as a relinquishment of control but as a means to improve patient care through the standardization of ventilator management. The growing number of patients requiring mechanical ventilation combined with a diminishing workforce makes the use of automated weaning modalities more a medical and financial necessity than an expensive and unnecessary luxury.

Is Automated Weaning Superior to Manual Spontaneous Trials? No

An undeniable rationale (if not current justification) exists for “closing the loop” on ventilator support and its withdrawal (weaning). There are worrisome pressures in modern medical practice, as we face the likelihood of increased demand for mechanical ventilation in our aging population and confront staffing issues,^19,36,37 the mandate to lower costs,³⁸ the widespread failure to implement best practices, and the need for closer patient surveillance so as to intervene in a timely fashion. Well-designed care protocols, if carefully followed by trained health-care professionals, speed the process of liberation from invasive ventilation.^39,40 One might logically ask why not automate protocolized weaning? The short answer is that, as presently implemented, automation does not incorporate all of the key variables that determine success or failure.

Present day automated algorithms rely primarily on assessing respiratory mechanics, patterns of breathing, and gas exchange, each measure of which has questionable reliability when assessed over brief periods and none of which is currently integrated with unmonitored data coming from the highly influential cardio-hemodynamic axis. The drivers of automated weaning are not fine tuned to recognize trends in response, which may be critical in predicting the patient's trajectory toward the ultimate outcome. Furthermore, current systems work only for well-selected and “tee-ed up” populations. Implicit in the question asked in print by Hess and MacIntyre (“… Why are we still weaning ?”⁴¹) is the concern that graded withdrawal of machine support itself may not be necessary or wise. If so, we clearly do not need to automate that process.

There are 5 essential steps to ventilator independence: preparation, evaluation, withdrawal of ventilator support, extraction of the endotracheal tube, and peri-extubation care. Attention to the considerations outlined in Table 2, particularly cardiovascular status, fluid balance, secretion clearance, and appropriate mental status, prepares the patient optimally for disconnection of the machine.^42–44 Some data convincingly argue that once the patient has optimally been prepared, he or she should be kept fully supported until evaluated for extubation potential under the scrutiny of a trained professional by an abrupt “sink or swim” spontaneous breathing trial in which a T-piece or minimal pressure support is utilized.^45,46 Depending on performance during the trial, the patient is either left connected and fully supported or extubated expeditiously. Whatever the wisdom of gradual withdrawal of machine support, the goal of weaning is not only to discontinue invasive ventilation but also to remove the endotracheal tube.⁴⁴ Evaluation and management of the upper airway both before and after completion of the spontaneous breathing trial comprises the peri-extubation care component of the sequence. No computerized system can assess or manage those crucial pieces of the process. Automated weaning, as currently implemented, only addresses the withdrawal of support from the standpoint of ventilation power and lung performance, with graded interventions gauged primarily by patient breathing pattern response.^6,19,47,48 Faith is placed implicitly in oxygenation efficiency and the tidal volume and frequency amplitudes to signal reserve or intolerance, whatever the cause of trouble might be: ventilation power, hemodynamics, or oxygenation.⁴⁸ But the inconvenient truth is that the respiratory pattern has never been shown to reliably index hemodynamic or oxygenation distress. Newer systems have begun to integrate gas exchange data, which clearly is a conceptual step forward. Unfortunately, to this point, evidence is very limited regarding their incremental value and performance merits.⁹

It is obvious that automated weaning does little for the preparation phase or for the peri-extubation phases of the weaning process. The causes of postextubation distress include retention of airway secretions, upper airway dysfunction, positive fluid imbalance, and inadequate cardiovascular reserve: congestive failure, arrhythmia, or coronary ischemia.^42–44 Respiratory distress and central vascular congestion can be precipitated by spontaneous ventilatory effort.⁴⁹ These problems of secretion retention and cardiac dysfunction are often best revealed considerably after extubation has been accomplished, cough effectiveness is tested, and net intrathoracic pressures fall.⁵⁰ Moreover, because separation of the patient from the ventilator may fail because of neuro/psychiatric and sedation issues as well as for respiratory or cardiac reasons, the patient should ideally be evaluated for alertness, command response, and cough effectiveness before acting on any “go ahead” prompt from the automated system, which cannot assess these important characteristics.

On the respiratory side, imbalance of capability with demand may be provoked by postextubation secretion retention, dynamic hyperinflation, bronchospasm, upper airway obstruction, or atelectasis. Each may contribute to gradually developing ventilatory inadequacy and or hypoxemia after extubation. The lungs are challenged on the cardiovascular side as fluid shifts into the central compartment in the transition from positive to negative intrathoracic pressure, and the left ventricle may be excessively afterloaded by strong spontaneous efforts. In fact, worsening lung edema, blood translocation from the periphery to the central compartment, diastolic dysfunction, coronary ischemia, and arrhythmias are frequent causes for the need to re-intubate.^43,51 Automated weaning before extubation might help ease the changeover or avoid abrupt transitions, but comprehensive patient assessment would need to include all variables relevant to postextubation stability, which the attentive bedside caregiver (but not a computer-aided ventilator) can assess and integrate effectively. Although neuropsychiatric distress and sedation issues generally present themselves before attempting extubation, they can arise at any point during the weaning process and may frustrate automation of the process.

Complex mechanisms and critical illnesses often have more than one key element. Perhaps for this reason, the ability of breathing pattern variables alone to predict extubation outcome and monitor “weanability” is weak.^51–53 In fact, integral components of most automated weaning algorithms, such as the rapid shallow breathing index, although as good or better than any other weanability indicator, do not predict well for the most complicated patients.^51,52,54 Unmeasured variables, such as the tidal swings and esophageal pressure, outperform the rapid shallow breathing index in head-to-head comparisons.⁵⁵ This advantage is particularly significant when monotonic trends of the esophageal pressure are detected; by comparison, the rapid shallow breathing index tends to plateau to its final value at a very early stage in the failed trial. Moreover, breathing pattern variability, an innate characteristic of health,⁵⁶ is not taken into full account by automated programs. Breath-to-breath variation and complexity of the breathing pattern, key observations not incorporated in today's automated algorithms, have been correlated with ventilation reserve and improved gas exchange.^57–59 Conversely, lack of variation, unaccounted for in current algorithms, bodes poorly for weanability or even survival.⁶⁰ Variability tends to decline as the work load/capacity ratio increases and is influenced by the mode and level of support.⁶¹

The existing literature suggests that the mode of mechanical ventilation used in the weaning attempt is less important than previously thought.^1,3,16 Without question, however, the modes now available for graded withdrawal of ventilation are much more flexible than in prior years. These include neurally adjusted ventilatory assist,^61,62 proportional assist ventilation,^5,63 and complex modes that vary the level of pressure assistance and frequency, depending primarily on targeted values for tidal volume and breathing rate^4,6 (Table 3).

Although there has been progress, we still have a long way to go before automated modes can be relied upon to automatically withdraw ventilator support at an appropriate rate. Most of the automated weaning modes now deployed for clinical use impose important restrictions on the type of patient subjected to them. It is assumed for example, that the patient is hemo-dynamically stable, is well-oxygenated with limited concentrations of inspired oxygen, has been adjusted to the appropriate level of sedation (to “even” the baseline for the automated start), does not have exacerbated illness, does not have unstable respiratory drive or bizarre spontaneous breathing pattern, does not have fever or acid/base disturbance, etc.^22,64 What's left in the pool of patients to automate are usually not our most troublesome problems, perhaps a reason why evaluations of routine surgical patients do not show a noticeable advantage for weaning automation.⁶⁴

In automated weaning, arbitrarily defined zones of respiratory comfort are identified for tidal volume and frequency.^63,65 However, many ventilated patients have diseases or chronic conditions that violate these assumptions.⁶⁶ For example, a patient with severe kyphoscoliosis may have a high baseline frequency/tidal volume ratio perfectly compatible with adequate compensation. The same holds true for some patients with chronic neuromuscular disorders.⁶⁶

Whether automated weaning proves superior to manual surveillance and adjustment depends on the answers to 4 important questions (Table 4). These include the population in question, the level of staffing and surveillance, the implementation of the automated weaning process to include gas exchange and nonrespiratory (eg, cardiovascular and hemodynamic) variables, and fidelity of the sensor information that drives the adjustments. When the conditions are right and/or the unit is suboptimally staffed, automated weaning appears to hold some advantage over non-automated weaning.^8,9,64,67 However, when they are not, current systems do not seem to do any better than a well-tuned and well-resourced environment, particularly those that effectively apply protocols.⁶⁸ Patients who do not depend on the machine for reasons of mechanical insufficiency and excessive work load, such as many who undergo surgical care, are not advantaged by automated weaning. Co-management decisions apart from ventilator care, such as those related to levels of sedation, have influenced protocol efficacy in published studies.^69,70 Indeed, the results of comparative weaning trials depend heavily on how they are conducted. As a classic example, published comparisons of pressure support, synchronized intermittent mandatory ventilation, and T-piece weaning have come to differing conclusions,^45,46,71 which to some extent can be traced to how each technique was implemented. The same is true in published comparisons of automated methodology with standard conventional methodology.^6,47,67 Ideally, funding for comparative trials of any technology would be free of potential conflicts of interest. Unfortunately (but understandably) the majority of the published work favoring commercially available modes of automation has been industry-funded and -supported.

Factors at work that determine the incremental value of automated weaning include the quality of the inputs, the delay until adjustments are implemented, the presence of erratic patterns of ventilation demand, the nature of the patients under study, the skills of practitioners, and the surveillance and protocols available to guide non-automated weaning decisions. There is some concern that automated modes will lead to a lack of valuable interactions that now help to synergize and integrate the efforts of and communications among therapists, nurses, and physicians. Subtle forms of patient-ventilator asynchrony, a factor that influences duration of ventilation,⁷² may not be addressed when the algorithm is not geared specifically to detect them.

Clearly, there are populations where automated weaning (at least as presently configured) would be unsafe, because computer-driven systems do not integrate all of the important variables that can influence patient outcome. As already noted, most studies exclude potentially problematic patients, such as those with neurological impairment and hemodynamic concerns, and little to no data exist for these cohorts. A shortcoming of most automated programs is that there may be failure of appropriate machine response to agitation, secretions, bronchospasm, etc. Automated programs may tend to interrupt expression of an informative trend that otherwise would be detected and promptly addressed by appropriate intervention. As currently implemented, automated algorithms are relatively insensitive to such trends and can neither sense nor integrate the key bedside signs of tolerance or developing problems, such as physical appearance, physical examination, electrocardiogram abnormalities, etc.