Should Ventilator-Associated Events become a Quality Indicator for ICUs?

Limitations of VAP as a Quality Metric for Ventilated Patients

VAP has long been the de facto means for hospitals and regulators to track quality of care for ventilated patients. Advocates point to VAP's frequency, morbidity, and high cost to support its use as a quality marker and argue that eliminating VAP would save tens of thousands of lives per year.² In practice, however, VAP has proven to be a poor means to measure and drive quality improvement for ventilated patients. Weaknesses of VAP as a quality metric are summarized in Table 1. The fundamental problem is that VAP diagnosis is subjective and nonspecific. Typical diagnostic criteria include signs such as “new or progressive infiltrates,” “increased secretions,” and “worsening oxygenation.” These signs are all subjective; hence, reasonable people are prone to disagree. Indeed, a wealth of studies attests to high rates of disagreement between different surveyors evaluating the same subjects for VAP.^3–6 Furthermore, there is no consensus on how best to combine these and other criteria to identify VAP. One recent study identified 6 different possible definitions for VAP and demonstrated that they each identify different sets of subjects and generate different VAP rates.⁷ Finally, VAP's subjectivity is compounded by low specificity. Many conditions are associated with poor oxygenation, radiographic infiltrates, and changes in secretions, including pulmonary edema, ARDS, thromboembolic disease, pulmonary contusions, and others.⁸ Different background rates of these conditions in an ICU can have a substantial impact on the observed VAP rate.⁹

These weaknesses of traditional VAP definitions make it difficult or impossible to compare VAP rates between institutions or even to compare VAP rates within a single institution over time. One cannot know whether lower VAP rates in one institution or in one period versus another period are due to stricter interpretations of subjective clinical signs, differences in the prevalence of common mimickers of VAP, or true differences in the frequency of bona fide pneumonias.

A further limitation of VAP definitions is their relatively low attributable mortality rate. Early investigations noted that subjects with VAP were much more likely to die than subjects without VAP and that the net mortality rate among VAP subjects could be as high as 50%. These analyses were confounded, however, by failure to take into account the baseline expected mortality rate for patients who develop VAP, since patients with VAP tend to be among the sickest and most vulnerable of critically ill patients. Subsequent analyses that attempted to match subjects with VAP to similar subjects without VAP, to adjust for important comorbidities, and/or to employ novel statistical strategies to overcome these sources of bias generated much lower estimates of the attributable mortality of VAP.^10–12 Furthermore, if VAP is only responsible for a small fraction of ICU mortality, this begs the question of whether focusing quality improvement efforts on VAP detection and prevention alone is the best way to improve outcomes for ventilated populations.¹³ What about all of the other common and morbid conditions that can develop in ventilated patients, such as ARDS, pulmonary edema, severe sepsis, thromboembolic disease, and others? Focusing on VAP alone devalues the importance of these events and the necessity of designing prevention programs to address all of these potential complications of mechanical ventilation, not just pneumonia.

Finally, focusing on VAP alone can generate a misleading sense of improvement when none may be present. The weaknesses of VAP definitions allow for the possibility that lower VAP rates may be due to changes in surveillance rather than due to true improvements in care.¹⁴ This is a major risk when evaluating quality improvement programs and prevention strategies in before-after studies or open label trials. VAP surveyors and/or investigators may subconsciously apply VAP criteria more strictly in the intervention period or study arm, leading to fewer observed VAPs despite no true change in disease incidence.¹⁵

Difficulty distinguishing bona fide pneumonias from colonization of the oropharynx or endotracheal tube alone further complicates the interpretation of VAP prevention studies and initiatives. This is particularly the case for interventions designed to decrease microbial burden in the oropharynx and/or endotracheal tube, such as oral chlorhexidine and silver coating of endotracheal tubes. These interventions can lower the frequency of positive endotracheal aspirates, which in turn leads to fewer perceived pneumonias, but neither intervention has been shown to decrease mean duration of mechanical ventilation, ICU stay, or mortality rates.^16–18 Failure to impact these outcomes begs the question of whether these interventions prevented true pneumonias or if they simply decreased the frequency of false positive pneumonias. Indeed, oral care with chlorhexidine may even be harmful to patients: 2 meta-analyses of oral care with chlorhexidine found higher mortality rates among patients randomized to chlorhexidine.^18,19

The Genesis of VAE Definitions

VAE definitions were developed by a working group of professional society representatives convened by the Centers for Disease Control and Prevention (CDC) in 2011 and 2012. Their intent was to overcome many of the limitations of traditional VAP definitions, specifically to make surveillance more objective, to broaden the focus of surveillance beyond pneumonia alone, and to allow for the possibility of automating surveillance using electronic health data. VAE surveillance is designed to identify patients with respiratory deterioration after a period of stability or improvement. This is defined as at least 2 days of increased levels of ventilator support after at least 2 days of stable or decreasing ventilator support. Ventilator support is measured using patients' daily minimum PEEP and daily minimum F_IO2. A sustained increase in the daily minimum PEEP of ≥3 cm H₂O or a sustained increase in the daily minimum F_IO2 of ≥20 points qualifies (Fig. 1). The definitions use patients' daily minimum ventilator settings in order to get patients' best values each day and to avoid noise attributable to transient events, such as suctioning the endotracheal tube, bronchoscopies, repositioning patients, transportation, and/or procedures.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Two examples demonstrating patterns of daily ventilator settings in patients with ventilator-associated events. Patient A has a significant increase in daily minimum PEEP lasting for ≥2 d following ≥2 d of stable or decreasing daily minimum PEEP. Patient B has a significant increase in daily minimum F_IO2 lasting for ≥2 d following ≥2 d of stable or decreasing daily minimum F_IO2.

VAE definitions follow from the recognition that there is no perfect or even good definition for pneumonia available using current routine diagnostic tools. All existing definitions have many subjective components, and none has been shown to be more accurate than another.^20,21 All yield many false positives and false negatives relative to histological evaluation at autopsy. Rather than try to come up with a new VAP definition then, the working group suggested shifting the focus of surveillance away from pneumonia to complications of mechanical ventilation instead. In doing so, the working group was simultaneously able to sidestep the inherent limitations of current VAP diagnostics, broaden the focus of surveillance (and hence prevention) to include additional morbid complications of mechanical ventilation, and simplify the surveillance definition, since the intent is only to identify episodes of nosocomial respiratory deterioration, not pneumonia.

The VAE definition set includes subcriteria for infection-related ventilator-associated complications (IVACs) and possible ventilator-associated pneumonia (PVAP). IVAC criteria aim to identify the subset of VAEs that may be infection-related, as indicated by concurrent abnormal temperature or white blood cell counts and the initiation of ≥4 days of new antibiotics (Fig. 2). PVAP aims to identify the subset of IVACs that may be pneumonias by identifying patients with concurrent pulmonary specimens with moderate or heavy amounts of organism growth or with less growth but moderate or many neutrophils on Gram stain (an objective marker of purulence). An example of PVAP is shown in Figure 3. Although the VAE definition set does include PVAP, there is no reason to believe that PVAP is any more (or less) accurate than any prior pneumonia definition. The sole benefit of PVAP is to provide a familiar entity to clinicians within the framework of VAE surveillance. Indeed, CDC explicitly notes that whereas VAE and IVAC are potentially suitable for public reporting and benchmarking, PVAP is not suitable, given that (1) there is no reason to think it overcomes the limited accuracy of other VAP definitions, and (2) hospitals vary widely in their practices around the acquisition and processing of pulmonary specimens, and hence PVAP rates are inherently unsuitable for comparisons between institutions.¹

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

An example of a patient with an infection-related ventilator-associated complication (IVAC). This patient has a fever, leukocytosis, and ≥4 d of new antibiotics occurring within 2 d of a ventilator-associated event. Vanco = vancomycin; pip-tazo = piperacillin/tazobactam; WBC = white blood cells.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

An example of a patient with a possible ventilator-associated pneumonia (PVAP). This patient meets all of the criteria for an infection-related ventilator-associated complication (IVAC) (increase in ventilator settings, abnormal temperature or white blood cell count, and ≥4 d of new antibiotics) but also has a concurrent pulmonary specimen with evidence of purulence (4+ neutrophils) on Gram stain and a positive quantitative culture for a pathogenic organism. The combination of IVAC with concurrent evidence of purulence and/or a positive quantitative or semiquantitative culture meets the criteria for PVAP. WBC = white blood cell; EA/BAL = endotracheal aspirate/bronchoalveolar lavage; Polys = polymorphonuclear leukocytes; Epis = epithelial cells.

Data Supporting the Potential Value of VAEs as a Quality Indicator

There is no consensus on what constitutes a suitable quality indicator, but arguably 3 key criteria are 1) clinical importance, 2) capacity to support objective and efficient surveillance, and 3) preventability. An ideal metric should identify a common clinical condition associated with high morbidity or mortality. The metric should be amenable to efficient surveillance in order to avoid burdening hospitals with excessive demands for data collection and analysis. Surveillance efficiency includes the capacity to define the metric using objective clinical data that are readily available in existing electronic data systems and amenable to automated analysis. Most importantly, a good metric should help identify potential lapses in care that can inform tangible strategies to improve care. Better care should then lead to lower event rates and better outcomes for patients. Studies to date suggest that VAE definitions may fulfill many of these criteria.

Clinical Importance

VAEs appear to be clinically important events. They affect 5–10% of patients and have consistently been associated with longer duration of mechanical ventilation, longer hospital lengths of stay, more antibiotic usage, and higher mortality rates.^22–32 Multiple studies suggest that patients who develop VAEs are approximately twice as likely to die compared with similar patients who do not develop VAEs.^22–32

Surveillance Efficiency

VAE definitions were purposefully designed to be objective and suitable for electronic surveillance. VAE avoids the subjectivity inherent in VAP definitions, since it is based solely on trends in 2 data elements: daily minimum PEEP and daily minimum F_IO2. These are both concrete measures that all observers ought to be able to agree upon (in contrast to classic VAP criteria that invite debates as to whether an infiltrate is new or whether the volume and consistency of patients' secretions have changed from one day to another, etc). An increasing number of hospitals now document ventilator settings in their electronic health record systems or in independent databases maintained by respiratory therapy departments. Some hospitals are also able to automatically download ventilator settings directly from their ventilators into their electronic health records. The key data elements for VAE detection are therefore increasingly available electronically, making full automation of VAE surveillance increasingly possible. Multiple hospitals have now published their experience successfully automating VAE case detection.^{27–29,31,33}

Workers from the Netherlands have pointed out that one can define daily minimum PEEP and F_IO2 in different ways and that this in turn can have a large effect on case detection.²⁸ Options include the lowest values derived from minute-to-minute ventilator settings, the lowest values derived from paper charts with hourly documentation, or the 10th percentile. The CDC subsequently clarified that a setting can only qualify as the daily minimum value if the patient is maintained at that setting for at least 60 min.

Capacity to Guide Improvements in Care (Preventability)

VAE criteria appear to identify potential opportunities to improve care and outcomes for patients. This is not readily apparent at first blush insofar as VAE is a novel surveillance concept rather than a longstanding clinical concept. Clinicians do not have an immediate intuitive feel for what conditions trigger VAEs and hence how best to prevent them. An expanding body of literature, however, fills this gap. At least 4 studies have evaluated consecutive subjects with VAEs to try to identify the clinical conditions that led to increases in ventilator support.^22,25,28,30 Findings from these studies are similar. The majority of VAEs are caused by pneumonia, fluid overload, atelectasis, and/or ARDS. The relative frequency of these conditions varies between studies, but generally pneumonia accounts for about 20–40% of VAEs, fluid overload for 30–40%, atelectasis for 10–20%, and ARDS for 10–20%.^22,25,28,30

Knowing which conditions typically trigger VAEs helps to inform the development of strategies to prevent VAEs. Clinicians can intuit prevention strategies based on their prior experience with these 4 conditions as well as draw upon an increasing number of studies that identify specific risk factors for VAEs and demonstrate strategies that can decrease the frequency of VAEs.

Case-control studies have identified benzodiazepines, opioids, fluid balance, and high tidal volumes as risk factors for VAEs.^34,35 Interventional studies have helped to affirm that improvements in practice can lower both VAE rates and mean duration of mechanical ventilation. Muscedere and colleagues^26,36 assessed the impact of a 2-y educational program to increase the adoption of clinical practice guidelines for ventilated subjects in 11 ICUs in Canada and the United States. They achieved relatively modest improvements in the frequency of guideline concordant practices, yet the investigators were still able to detect a significant decrease in VAEs from 13.6% of subjects sampled at baseline to 9.7% of subjects sampled at the end of the study (P = .05).²⁶ On multivariable analysis, spontaneous awakening trials and spontaneous breathing trials were potentially protective against VAEs.

The CDC Prevention Epicenters Wake Up and Breathe Collaborative further affirmed the responsiveness of VAEs to less sedation and earlier extubation.³⁷ The collaborative included 12 ICUs affiliated with 7 hospitals. These ICUs collectively worked to increase the frequency of coordinated daily spontaneous awakening trials and spontaneous breathing trials in eligible subjects using an opt-out protocol for nurses and respiratory therapists. Significant increases in the frequency of spontaneous awakening trials, spontaneous breathing trials, and the percentage of spontaneous breathing trials performed off sedation were associated with significant reductions in VAEs and IVACs. Collaborative members were able to decrease their collective VAE rate from 9.7 to 5.2 events/100 episodes (adjusted odds ratio 0.63, 95% CI 0.42–0.97) and their IVAC rate from 3.5 to 0.52 events/100 episodes (adjusted odds ratio 0.35, 95% CI 0.17–0.71). Importantly, the decrease in VAE rates was also accompanied by significant decreases in mean duration of mechanical ventilation (2.4 fewer days, 95% CI 1.7–3.1), ICU stay (3.0 fewer days, 95% CI 1.6–4.3), and hospital stay (6.3 fewer days, 95% CI 4.0–8.6).

Finally, Mekontso Dessap et al³⁸ have demonstrated that restrictive fluid management can decrease VAE rates by as much as 50%. These investigators randomized subjects ready for weaning from mechanical ventilation to daily B-type natriuretic peptide measures versus usual care. B-type natriuretic peptide is a biomarker for hypervolemia. Subjects with high B-type natriuretic peptide levels were managed using a depletive fluid strategy (less fluid in, more fluid out). Daily B-type natriuretic peptide level measurements and depletive fluid management were associated with a 50% decrease in VAEs (8.6% of subjects vs 17.8% of subjects, P = .02) and significantly more ventilator-free days.

VAE Surveillance: A Patient Safety Opportunity

All told, the adoption of VAE as a quality metric carries substantial promise to improve outcomes for ventilated populations.^39,40 VAE surveillance brings to light a broad group of patients suffering morbid complications of mechanical ventilation beyond those identified through VAP surveillance alone. VAE analyses suggest promising strategies to improve care for ventilated patients.⁴¹ The potential interventions likely to prevent VAEs are highly aligned with emerging best practices in critical care.^42–44 Finally, VAE surveillance is objective and amenable to automation, making it a more rigorous and efficient means to track the impact of care improvement programs compared with VAP surveillance.

Concerns About the Potential Value of VAEs as a Quality Indicator

Observers have raised important concerns about the potential suitability of VAE as a quality indicator. Critics cite the following issues: (1) the positive predictive value of VAEs for pneumonia is low, (2) the sensitivity of VAE for pneumonia is low, (3) VAE surveillance is susceptible to gaming, and (4) VAEs are probably just markers of severe illness rather than complications per se, and most are not preventable.

The Positive Predictive Value of VAEs for Pneumonia Is Low

A number of investigators have noted that the positive predictive value of VAEs for traditionally defined VAP is low. Most authors have reported that only about a quarter to a third of VAEs meet traditional surveillance definitions for VAP.^22,26,28 Some investigators have even found that as few as 7% of VAEs meet VAP criteria.²⁹ The fact that only a fraction of VAEs qualify as pneumonias, however, should not be a surprise, given that VAE definitions were specifically designed to broaden the purview of surveillance beyond pneumonia alone. The low positive predictive value then is expected and is arguably a strength of VAE insofar as it indicates that VAE surveillance is bringing to light morbid complications of mechanical ventilation beyond just pneumonia, such as fluid overload and ARDS.

The Sensitivity of VAE Definitions for VAP Is Low

Many studies have also pointed out that VAE surveillance misses many patients with traditionally defined VAPs. Indeed, in most series, VAE surveillance misses about 65–75% of traditionally defined VAPs.^26,29,45,46 These false-negative cases merit contemplation.

First, it is worth recalling that VAE criteria were born out of the recognition that traditional VAP criteria are too subjective and inaccurate to produce meaningful benchmarks. Clinical audits suggest that up to 75% of patients treated for VAP most likely do not have pneumonia.⁴⁷ Autopsy series also find that 40–60% of subjects who meet clinical criteria for VAP do not have histological evidence of pneumonia.²¹ VAP is therefore a questionable standard against which to measure the accuracy of VAE definitions.

Second, by definition, a VAP that does not meet criteria for VAE either did not have a period of stability or improvement before the pneumonia or did not require a sustained increase in ventilator settings equal to or greater than VAE thresholds (3 cm of PEEP or 20 points of F_IO2). In one case series in a trauma ICU, 39% of pneumonias missed by VAE were due to the lack of a period of stability or improvement.⁴⁶ Many of these patients had pulmonary contusions and/or rib fractures, a population at high risk for pneumonia but also a population in which the diagnosis of pneumonia is particularly difficult, given the overlap in clinical signs between contusion, hemorrhage, ARDS, and pneumonia. Almost all of the remaining pneumonias missed by VAE criteria were patients who did not require sustained increases in ventilator settings. One wonders about the clinical importance of pneumonias that require little or no extra respiratory support. Some of these may have been mild pneumonias that improved very rapidly. Others may not have been pneumonias at all but perhaps alternative diagnoses or bacterial colonization alone.

Third, it is important to reiterate that VAE is a surveillance concept and not a clinical diagnosis. Surveillance is governed by a different set of values compared with clinical diagnosis. Clinical care, particularly in intensive care settings, places a very high value on sensitivity, given the potentially catastrophic consequences of missed or delayed diagnoses. Surveillance, by contrast, favors objectivity, reproducibility, and morbidity, since surveillance is only meaningful if it allows for comparisons over time and if it identifies severe events worth preventing. Focusing on more severe events decreases subjectivity, since severe events tend to be less ambiguous and more clearly meaningful. For example, workers trying to prevent surgical site infections tend to focus on deep and organ space infections rather than on superficial infections.⁴⁸ Similarly, surveillance efforts to support better sepsis care tend to focus on patients with severe sepsis and septic shock rather than patients with systemic inflammatory response syndrome or sepsis alone. In the same vein, VAE criteria focus surveillance on the subset of particularly severe and unambiguous pneumonias that require sustained increases in ventilator support. At the end of the day, however, VAE analyses by their nature are designed to identify care improvement opportunities that can be applied to the care of all patients in the ICU. Hence changes in care learned from analyses of triggers for severe pneumonias should also help to prevent milder pneumonias, even if they would never have triggered VAE criteria.

VAE Surveillance Is Susceptible to Gaming

Two papers have suggested strategies for gaming VAE surveillance. One group suggested managing patients with high ventilator settings throughout their time on mechanical ventilation, since this will preclude the need to increase ventilator settings in response to new clinical events.⁴⁹ The problem with this strategy is that it is clinically dangerous, since sustained high PEEP and/or hyperoxemia are both independent risk factors for harm.⁵⁰ Another group suggested alternately raising and lowering the PEEP by 1 cm H₂O each day. This will eliminate the possibility of a stable baseline and therefore preclude any VAEs.²⁹ Both strategies, however, could arguably be construed as fraud. If VAE ever does become a quality indicator, then presumably health departments, accrediting agencies, and/or payers will review hospitals' primary ventilator data periodically to assure compliance. Hospitals that appear to be deliberately adjusting ventilator settings solely to avoid VAE detection could face penalties.

Nonetheless, these two suggestions on avoiding VAEs underscore the fact that VAE surveillance is not immune to gaming and bias. Clinicians do have the capacity to prevent VAE detection by manipulating ventilator settings. Traditional VAP surveillance was also susceptible to missing cases depending on clinician behavior (eg, opting not to order chest radiographs or culture specimens). The big difference in bias risk between VAE and VAP surveillance is that VAP surveillance was subject to bias from both clinicians and surveyors. VAE surveillance, by contrast, eliminates infection preventionists' scope for subjective decision making, since the VAE definitions are strictly mathematical.

Most VAEs Are Not Preventable

Commentators have suggested that VAEs are merely surveillance markers for severe disease rather than complications per se.^2,49,51 Boyer et al³⁰ evaluated all VAEs detected in their institution over a 1-y period and tried to determine whether each event was preventable. They estimated that only 37% of VAEs were potentially preventable. Adjudicating preventability, however, is notoriously difficult. Although the immediate event that precipitated the VAE may not have appeared to be preventable, it is impossible to know what could have happened if the patient had been managed differently from their first day of mechanical ventilation. Perhaps with less sedation and earlier mobilization, the patient may have been extubated long before the supposedly inevitable event. Preventability is better assessed through prospective intervention studies, and, as outlined above, at least 3 studies thus far have demonstrated successful decreases in VAE rates.^26,37,38

The VAE Algorithm Is Confusing

Since the initial development of a VAE algorithm in 2013, the National Health Safety Network has made several modifications to the algorithm aimed at improving VAE surveillance. The new VAE algorithm has three definitional tiers: ventilator-associated conditions (VAC), IVAC, and PVAP (Figs. 1⇑–3). The first tier of the algorithm, VAC, only takes into account changes in patients' ventilator settings. The 2 subsequent tiers of the VAE algorithm, IVAC and PVAP, require that patients must first meet the initial VAC definition. However, there is a clear conceptual flaw with all 3 tiers insofar as they are all only apparent ≥48 h after respiratory deterioration first occurs. Thus, although the VAE algorithm is not intended for diagnostic use by clinicians, its after-the-fact nature is so divorced from clinical logic that it is clearly not useful to practicing ICU physicians within the normative time course of care.

VAE Surveillance Is Insensitive for Identifying VAP

An ideal ICU quality indicator should have both high sensitivity and high specificity and thus must be compared with some accepted standard of measurement. For VAE, the accepted standard of comparison is arguably VAP. Many studies have noted, however, that VAE surveillance misses many pneumonias. Klein Klouwenberg et al,²⁸ for example, found that the VAE algorithm detected 33% of VAPs at most. Likewise, most VAEs were caused by volume overload and infections but not necessarily VAP. Indeed, the positive predictive value of a VAE for VAP was only 25% (95% CI 18–33%). Klein Klouwenberg et al²⁸ further reported that, even in cases that were positive for VAC or IVAC as well as clinically defined VAP, these events did not necessarily occur at the same time. Lilly et al,²⁹ investigating 8,408 mechanically ventilated subjects, also demonstrated that VAE/VAC detected less than a third of VAP cases, with a sensitivity of 32.5% and a positive predictive value of 7%. A third study in trauma subjects requiring >48 h of ventilation similarly reported that IVAC criteria had low accuracy in identifying VAP in high-risk trauma subjects.⁵² The sensitivity and positive predictive values of IVAC for VAP identification were 28.1 and 58.6%, respectively, meaning that the algorithm failed to detect about three quarters of subjects with VAP.

The VAE algorithm detects a broad spectrum of conditions and complications associated with changes in ventilator settings that correspond to the VAC criteria, the most common of which are VAP, ARDS, pulmonary edema, and atelectasis.²² In a retrospective study by Hayashi et al,²⁵ the authors assessed the etiology of VAE, finding that among 143 subjects with VAE, 30.7% had microbiological evidence of VAP. Another 31%, however, had no microbiological evidence of VAP. Other etiologies that could have explained VAE included: atelectasis in 16.3% of subjects, acute pulmonary edema in 11.8%, ARDS in 6.5%, pleural effusion in 3%, pulmonary embolism in 2%, and abdominal distention in 1.3%. Lilly et al²⁹ found that ARDS was associated with 73% of VAE cases and that acute kidney injury was associated with 27%. Given that most VAEs are not VAPs, the definition will cause confusion among physicians and will be difficult to implement for the purpose of significant clinical surveillance.

VAE Surveillance Is Not a Comprehensive Assessment of a Physiological System

An ICU quality indicator should be universally applicable and easy to use for clinicians, researchers, and administrators for characterizing patients in terms of severity of illness. The Acute Physiology and Chronic Health Evaluation (APACHE) along with the Risk, Injury, Failure, Loss, and End-stage kidney classification (RIFLE) are the most frequently used general severity-of-illness and renal failure severity scores in adult ICU patients. These score systems seek to take into account the entire range of relevant and available clinical data to predict patient outcomes.^53,54 The applicability and universality of both the APACHE and RIFLE systems have been well tested and demonstrated by comprehensive multi-center large sample studies.^55,56 However, there is still debate regarding the rationale for using VAE in the ICU, and further multi-center prospective validation studies are needed. Furthermore, VAE is not a scoring system for an organ or functional system in the body. As previously mentioned, the first qualifying tier of VAE is entirely based on changes in ventilator settings (specifically PEEP and F_IO2), and these changes do not necessarily represent changes in patients' pulmonary status. An increased APACHE or RIFLE score is associated with increased mortality, whereas a diagnosis of VAE might not always be associated with increased mortality because it can sometimes be indicative of what could actually be improved care, such as raising PEEP in the setting of atelectasis.

VAE Surveillance Is Susceptible to Abuse and Misuse

Since the first qualifying tier of the VAE algorithm, VAC, concerns changes in the ventilator settings of F_IO2 and PEEP, the VAE algorithm is susceptible to manipulation. Lilly et al,²⁹ for example, reported that VAE/VAC rates can be reduced 93% by easy manipulation of ventilator management protocols. Thus, the new VAE algorithm can easily be gamed to suggest improved ICU quality of care. If VAE surveillance is subsequently linked to public reporting, pay-for-performance, or data-driven incentive policies, physicians will be inclined to manipulate patients' ventilator strategies to minimize the incidence of VAE.

Furthermore, a lack of changes in ventilator settings is not equivalent to stable patient health, and vice versa. Changes in ventilator settings also have a close relationship to the therapeutic approach of the caregivers. Due to the diversity of detailed ventilator management strategies, different ICU teams may have different F_IO2 and PEEP protocols for a patient on mechanical ventilation. The VAE algorithm is in truth a reflection of caregiver tendencies rather patient condition. This is one of the most important reasons why such a ventilator setting-oriented algorithm cannot be used as a health-care quality indicator in the ICU.

VAEs Can Sometimes Be Triggered by Improvements in Care

The VAE algorithm focuses on increases in ventilator settings during mechanical ventilation. However, when a patient is obese or suffering from fluid overload, clinicians generally make significant PEEP and F_IO2 changes in order to improve care. Also, the more aggressive the attempts are to liberate the patient from the ventilator, the more likely failed extubation becomes, consequently leading to a higher likelihood of increased PEEP and F_IO2 and thus a higher incidence of VAE. Therefore, if VAE becomes a universally adopted form of ICU quality surveillance, clinicians might be less inclined to extubate mechanically ventilated patients. Paradoxically, this could lead to a situation in which VAP incidence increases due to prolonged ventilatory times and VAE incidence decreases because of fewer failed extubations.⁵⁷

Con Position Concluding Remarks

The primary aim of the new VAE algorithm was to create a more objective, efficient, and reliable measure of complications associated with mechanical ventilation compared with the current clinician-dependent assessment of VAP occurrence.^1,4 However, VAE surveillance does not perform as well as VAP surveillance at detecting VAP and has several undesirable characteristics that could make its implementation as an ICU quality indicator difficult. The VAE algorithm has a low sensitivity and positive predictive value for identifying VAP, and there are difficulties in standardizing the implementation of VAE surveillance due to easily exploitable loopholes in its definition. The VAE algorithm needs significant modifications and improvement in order to replace VAP as a surveillance method and quality indicator in the ICU. The preliminary data on the topic must be confirmed in additional large prospective studies.