Does This Patient Have Ventilator-Associated Pneumonia?

The wards and the post-mortem room show a very striking contrast in their pneumonia statistics . . . —Sir William Osler, 1907¹

A 68-year-old man with a history of congestive heart failure and remote non–small cell lung cancer suddenly develops shortness of breath. His initial chest radiograph shows diffuse infiltrates consistent with pulmonary edema. On arrival at the hospital, respiratory distress and hypoxemia necessitate intubation and ventilatory support. The patient is transferred to the intensive care unit and given diuretics, sedatives, and ulcer prophylaxis administered through a central venous catheter. Anticoagulants are started when a pulmonary embolus is confirmed. Over the ensuing 4 days he slowly improves, the radiographic infiltrates diminish, and his level of ventilatory support is gradually decreased.

On day 5 of admission, he develops tachypnea, tachycardia, hypoxemia, and a new fever. His nurse notes the return of a small amount of thin, beige secretions from his endotracheal tube. His anterior chest is resonant to percussion, but crackles are heard bilaterally over dependent lung fields. The ventilator sounds make cardiac auscultation difficult, but the rhythm seems regular and there are no obvious murmurs. He has mild sacral and peripheral edema. A portable chest radiograph again shows diffuse bilateral opacities. The white blood cell count is 12 000 cells/μL³. The intensive care team members agree that antibiotics should be started pending further evaluation. They wonder, however, whether the combination of fever, hypoxemia, leukocytosis, and a radiographic infiltrate is sufficient to establish a diagnosis of ventilator-associated pneumonia (VAP) and preclude investigation for another condition causing the patient's decompensation.

Ventilator-associated pneumonia is the most common and fatal nosocomial infection of intensive care.^{2 - 3} It affects between 9% and 27% of intubated patients and doubles the risk of dying, compared with similar patients without VAP.^{4 - 8} VAP prolongs the duration of ventilation, length of stay in the intensive care unit, total hospital length of stay, and cost of hospitalization.^{8 - 9} Despite its high incidence, diagnosing VAP is challenging because many conditions common among critically ill patients produce similar clinical signs; these include acute respiratory distress syndrome, thromboembolic disease, alveolar hemorrhage, sepsis, congestive heart failure, and atelectasis. Multiple case series describe the exceedingly poor correlation between the clinical diagnosis of pneumonia and true underlying VAP. More than 50% of patients diagnosed with VAP do not have the disease, while up to one third of those with VAP are not diagnosed.^{10 - 11} Not surprisingly, interobserver diagnostic agreement is consistently poor.^{12 - 13}

The difficulty in diagnosis is reflected in the very different reference standards used in clinical reports. Competing diagnostic criteria include those of the Centers for Disease Control and Prevention's National Healthcare Safety Network (formerly known as the National Nosocomial Infections Surveillance System) ( Article ),¹⁴ the Clinical Pulmonary Infection Score (CPIS) (Table 1),¹⁵ and the clinical criteria of Johanson et al.¹⁶ However, there is little consensus between these criteria. When 4 different criteria were applied to a single cohort of patients, they yielded estimates of VAP prevalence ranging from 4% to 48%.¹⁷

*In patients without underlying pulmonary or cardiac disease (eg, respiratory distress syndrome, bronchopulmonary dysplasia, pulmonary edema, or chronic obstructive pulmonary disease), 1 definitive chest radiograph is acceptable.

Despite the difficulty in making and agreeing about the diagnosis, timely and accurate diagnosis of VAP is critical. Delayed administration of appropriate antibiotics increases mortality.^{18 - 20} Conversely, inappropriate use of antibiotics increases costs, incurs risk of adverse drug reactions, and selects for resistant microbial flora that increase morbidity and mortality.^{21 - 23} Furthermore, an incorrect diagnosis of VAP can engender a false sense of security that might distract a clinician from finding and treating the true cause of a patient's clinical deterioration.

In light of the importance of rapid, accurate diagnosis of VAP paired against the difficulty in making the diagnosis, the medical literature was reviewed to quantify the usefulness of clinical findings to diagnose or exclude bacterial VAP. Unfortunately, there is no clearly accepted gold standard for the diagnosis of VAP.²⁴ This review was limited to patients with histological confirmation of VAP, since it is the most objective measure available. This standard necessarily biases this review to the sickest patients and might limit generalizability. Nonetheless, it is more accurate than bronchoscopic culture-based diagnosis, which yields different rates of pneumonia depending on technique and the quantitative culture threshold chosen.^{25 - 29} Since this review is intended to aid critical care clinicians at the bedside, it integrates clinical observations with the laboratory, radiographic, Gram stain, and culture findings that are typically available to intensive care clinicians evaluating a new pulmonary syndrome. Novel diagnostic techniques, such as triggering receptor expressed on myeloid cells and procalcitonin assays, were not included, since these tests are not yet commonly available and have not been validated against a histological gold standard.^{30 - 31}

The histological hallmark of VAP is multifocal disease favoring dependent lung segments, often at different stages of development and severity, with cultures growing heterogeneous microbial flora.^{29 ,32} From this pattern, pathologists infer that repeated seepage of oral fluids around the endotracheal tube cuff causes multiple microaspirations, bronchiolitis, and VAP.³³ Not all microaspirations progress to clinically manifest disease—localized patches of bronchiolitis may be found at autopsy in previously stable ventilated patients who died from nonpulmonary disease.

The pathophysiology helps explain the clinical findings of VAP. Mechanically ventilated patients who develop pneumonia have a gradual onset of fever, tachypnea, and hypoxemia. Often, the first clue to VAP is a sequence of progressive small increases in the patient's inspired oxygen fraction to maintain adequate oxygenation. These increases in oxygen requirements may mirror an increase in the quantity and purulence of endotracheal secretions. Chest radiographs initially show patchy, diffuse opacities that subsequently consolidate over the ensuing few days.

The multifocal, heterogeneous nature of VAP also explains some of the difficulty in establishing the diagnosis, even with positive lung biopsy or culture results. Biopsy specimens risk missing the area of active disease. Alternatively, when disease is found it might represent an area of clinically silent early bronchiolitis or resolving bronchial pneumonia. Likewise, cultures can miss the area of active disease (yielding false-negative results) or can detect clinically benign areas of bacterial colonization (yielding false-positive results). In addition, the many antibiotics given to critically ill patients may decrease the diagnostic yield of bacterial cultures.

Autopsies of ventilated patients with suspected pneumonia frequently reveal a substantial burden of alternative or coexisting pulmonary diseases that also cause fever, impaired gas exchange, increased secretions, and radiographic opacities. These other conditions include thromboembolic disease, hemorrhage, diffuse alveolar damage, fibrosis, atelectasis, carcinoma, lymphoma, and others.^{10 ,34} The high prevalence of coexisting pulmonary diseases in ventilated patients further complicates attempts to clinically diagnose VAP.

A structured literature search covering the years 1966 through October 31, 2006, was conducted using the PubMed MEDLINE interface and Google Scholar. An initial search strategy was designed to be inclusive by using the intersection of the medical subject and keyword headings ventilator-associated pneumonia, diagnosis, and (histology or pathology or histopathology or biopsy or autopsy). A second, more restrictive, search used the strategy ((ventilator OR ventilation) AND pneumonia) AND ((physical examination/ OR physical exam$) OR (medical history taking) OR (professional competence) OR (((sensitivity and specificity) AND.tw) OR (“sensitivity and specificity”)) OR ((reproducibility of results) OR (observer variation) OR (diagnostic tests, routine) OR (decision support techniques) OR (bayes theorem))). The abstracts of all English-language articles identified were reviewed for pertinence. The full texts of all potentially relevant articles were retrieved. Additional studies were identified from the reference lists of retrieved articles and from review articles located by the search strategy.

Studies were sought that included data on the clinical signs and symptoms and results of routine laboratory tests, chest radiographs, Gram stains, and cultures of pulmonary secretions. Included studies assessed both blind bronchial aspirates and bronchoalveolar lavage (BAL) fluid specimens. Blind bronchial aspirates are typically gathered by nurses, respiratory therapists, or general clinicians via deep suctioning beyond the end of the endotracheal tube. Bronchoalveolar lavage fluid is obtained similarly, but aspiration is performed after instilling saline into the lungs. This can be performed blindly by trained respiratory therapists or generalists (“mini-BAL”), or under bronchoscopic guidance by pulmonologists or thoracic surgeons. Bronchoalveolar lavage is frequently used to supplement clinical evaluation when bedside findings are equivocal. Although culture data are not available until 24 to 72 hours after incubation, culture results are included in this review, since they are the laboratory finding that clinicians most often seek first when considering the diagnosis of VAP. The cultures help despite their delayed availability, since the diagnosis of VAP is often a dynamic process in which the clinical picture is periodically reassessed over a period of 1 to 2 days before assuming the diagnosis.

Included studies assessed a minimum of 25 immunocompetent patients who had undergone mechanical ventilation for at least 24 hours and subsequently had histological examination of pulmonary tissue for VAP by biopsy or autopsy. Studies were excluded if they did not describe patients' clinical findings or if more than 20% of patients had pneumonia on admission to the intensive care unit. Data were systematically abstracted from the candidate studies and assessed for quality using modified criteria developed for The Rational Clinical Examination series.³⁵

Independent studies (quality levels 1-3) enrolled patients for histological evaluation without regard to whether there was a clinical suspicion of pneumonia. Among independent studies, level 1 and 2 studies enroll consecutive patients, while level 3 studies enroll nonconsecutive patients. Nonindependent studies (level 4) included only patients already clinically suspected of having pneumonia. Independent studies inform clinicians as to which findings suggest that they should consider a diagnosis of VAP, while nonindependent studies inform clinicians who already suspect pneumonia how various clinical clues (test results) modify the existing pretest probability of disease. The difficulty with nonindependent studies is that the symptom or sign under assessment was possibly also used to generate the initial pretest suspicion of pneumonia, thereby complicating any assessment of how that test result affected posttest probability. Studies that incorporate this circularity into their design tend to overestimate the sensitivity of key findings, since patients with these findings tend to be overrepresented in the study population relative to the population of mechanically ventilated patients at large. Specificity is either overestimated or underestimated, depending on the accuracy of the clinical suspicion of disease used to select patients for inclusion.

Some studies of VAP report results relative to a gold standard of histology and positive lung culture combined. When findings relative to histology alone were also presented, then histology alone was preferred as the gold standard, since antibiotic administration could affect the culture results.^{36 - 37}

Data from the identified studies were used to calculate sensitivity, specificity, and positive and negative likelihood ratios (LRs). Likelihood ratios use data on sensitivity and specificity of a test to quantify how the pretest odds of a diagnosis should be modified to generate a posttest odds. When multiple independent studies presented data on a single finding, summary LRs with random-effects summary measures were calculated using FAST*PRO version 1.8 (Academic Press, San Diego, Calif).³⁸ The summary measures are conservative in displaying the broad confidence intervals (CIs) around the estimates. For many findings used for VAP, the summary measures expose the paucity of data and apparent lack of diagnostic usefulness better than the single measures alone.

The MEDLINE search strategies identified 585 candidate studies; the review of reference lists and a Google Scholar search yielded 937 articles. Fourteen of these studies met inclusion criteria (Table 2).^{10 ,29 ,32 ,39 - 49} Most rejected articles were excluded because they lacked data on clinical findings. No studies of lung biopsies performed on living patients met the inclusion criteria. Therefore, this study set comprises only patients who underwent full autopsy or limited immediate postmortem lung biopsy. Relative timing of clinical assessment and pulmonary sampling was explicitly stated in 9 of the 14 articles. In each case, clinical criteria were abstracted and clinical suspicion of pneumonia assessed no more than 48 hours prior to histological evaluation. The other studies used the last clinical values measured before death but do not explicitly state the interval between measurement and death.

All of the included studies were either quality level 3 or 4, since none enrolled consecutive ventilated patients. Seven of the 14 studies were from France, where researchers were previously able to take advantage of a law permitting physicians to remove deceased patients' organs for scientific research unless specifically forbidden to do so by the patient before death. The 14 accepted studies included a total of 655 patients.

The prevalence of histological pneumonia in unselected patients who died while receiving mechanical ventilation ranged from 23% to 92% (summary prevalence, 47%; 95% CI, 35%-59%) (Table 2). In contrast, a recently published meta-analysis of 38 prospective cohort studies that assessed for VAP using clinical criteria other than postmortem histology reported a pooled cumulative incidence of VAP of 9.7% (95% CI, 7.0%-12%); the incidence was 17% (95% CI, 5.9%-28%) when limited to medical units alone, compared with 9.1% (95% CI, 5.9%-12%) in combined medical-surgical units.⁸

Data on interobserver variation in the evaluation of clinical signs associated with VAP are limited. A single study compared 2 intensivists' estimates of the CPIS (a clinical prediction rule that assigns points on an ordinal scale for abnormalities in a patient's temperature, blood leukocyte count, oxygenation, chest radiograph, tracheal secretions, and semiquantitative culture of a tracheal aspirate [Table 1]¹⁵ ) in a series of 52 patients.¹² Agreement was fair to moderate for most factors, including seemingly objective quantitative variables such as temperature (κ = 0.3) and oxygenation (κ = 0.4). Agreement was poor (κ = 0.2) on more subjective criteria, such as the quantity of tracheal secretions. Overall, intensivists' estimations of the CPIS score lacked precision (κ = 0.2). The study investigators attributed some of the variation in agreement to missing data during prospective patient evaluations and to subjectivity in the original CPIS definitions. When missing data were excluded from the analysis, the agreement on whether the CPIS suggested VAP increased to the moderate level (κ = 0.5).

Clinical Findings. The clinical features classically associated with VAP, such as fever, abnormally high or low leukocyte counts, and subjective assessment of sputum purulence, lack diagnostic usefulness (Table 3). In patients selected independently of their clinical features, the summary positive LRs for presence of fever, abnormal white blood cell count, and macroscopic purulent secretions were 1.2 (95% CI, 0.76-1.9), 1.3 (95% CI, 0.76-2.4), and 1.3 (95% CI, 0.88-1.8), respectively. This conclusion was also true in the nonindependent studies in which clinicians suspected VAP—all 3 findings still had LRs with CIs that included 1. Hypoxia and lung crepitation were equally nondiagnostic among patients already suspected of having VAP, although they were assessed in only a single study (LR, 1.2; 95% CI, 0.75-2.0 and LR, 1.1; 95% CI, 0.63-1.8, respectively).

Cell Counts on Pulmonary Secretions. Assessment of pulmonary secretions for neutrophils and organisms on Gram stain can contribute useful information (Table 4), particularly if the secretions are obtained by bronchoscopy. The presence of more than 50% neutrophils in BAL fluid marginally increases the likelihood of VAP (LR, 2.0; 95% CI, 1.4-3.0); however, the absence of this finding makes VAP very unlikely. In the single study that evaluated neutrophil percentage in unselected patients, no patient with less than 50% neutrophils in BAL fluid had VAP (LR, 0.09; 95% CI, 0.01-1.4).⁴³ This finding was borne out by 2 studies of patients already suspected of having VAP: both studies found that the LRs for less than 50% neutrophils were low (0.05; 95% CI, 0.01-0.36 and 0.10; 95% CI, 0-2.5).^{47 ,49} Likewise, high cellularity of BAL fluid (>400 000 cells/mL) was highly suggestive of VAP in patients already suspected of having VAP (positive LR, 15; 95% CI, 2.3-103), while its absence was similarly powerful in lowering the likelihood of VAP in a single study (negative LR, 0.11; 95% CI, 0.03-0.40).⁴⁹

Gram Stain of Pulmonary Secretions. Several studies show that the presence of bacteria on Gram stain of pulmonary secretions is suggestive of VAP (Table 4). The positive LRs for this finding increase with progressively more invasive sampling modalities: organisms on Gram stains of blind bronchial aspirates have an LR of 2.1 (95% CI, 0.81-5.5), which increases to 5.3 (95% CI, 1.3-22) with mini-BAL and up to 18 (95% CI, 1.1-302) when organisms are seen on BAL fluid recovered via fiberoptic-guided bronchoscopy. The absence of bacteria on Gram stains of fluid acquired by any of these methods is less definitive but does decrease the likelihood of the diagnosis by about half in both selected and unselected patient series (negative LR range, 0.16-0.60).

Examination of neutrophils recovered from BAL for intracellular organisms is equivocal: the finding adds no information in unselected patients (summary positive and negative LRs are both 1.0). A single small study of patients already suspected of having pneumonia found the identification of intracellular organisms useful but tempered by broad CIs (positive LR, 6.8; 95% CI, 0.43-106); the absence of intracellular organisms added little information (negative LR, 0.66; 95% CI, 0.46-0.96).⁴⁷

Culture of Bronchial Aspirates. Cultures of blind bronchial aspirate specimens that yielded heavy bacterial growth (>10⁵ colony-forming units/mL) were highly suggestive of VAP in the 2 studies that examined this finding (summary LR, 9.6; 95% CI, 2.4-38) (Table 4).^{42 ,45} Likewise, the absence of a positive culture result above this heavy growth threshold reduced the likelihood of VAP being present (summary LR, 0.42; 95% CI, 0.27-0.67).

Culture of BAL Fluid. Quantitative cultures of fiberoptic-guided BAL specimens were less definitive and had LRs with CIs that included 1. The summary LR for BAL culture of more than 10⁴ colony-forming units/mL was only 1.4 (95% CI, 0.76-2.5), while negative cultures had a summary LR of 0.78 (95% CI, 0.51-1.2).

The presence or absence of a new infiltrate on chest radiograph was the only clinical finding other than blind bronchial aspirate cultures that was evaluated in multiple independent studies and found to have both positive and negative LRs with summary CIs that excluded 1. The presence of a new infiltrate marginally increases the likelihood of VAP (summary LR, 1.7; 95% CI, 1.1-2.5), while the absence of a new infiltrate makes pneumonia less likely (summary LR, 0.35; 95% CI, 0.14-0.87).

One study assessed the accuracy of particular radiographic signs in a retrospective review of patients with VAP identified at autopsy.⁴⁰ The sensitivity and specificity of different radiographic features are presented in Table 5. The only signs that appear useful are air bronchograms. A single air bronchogram has a positive LR of 3.8 (95% CI, 0.74-19), but its low sensitivity (17%) suggests that this sign is infrequently found. Multiple air bronchograms are present more frequently (sensitivity, 83%) but appear less helpful (LR, 2.0; 95% CI, 1.3-2.9). The absence of multiple air bronchograms lowers the probability of VAP (negative LR, 0.29; 95% CI, 0.11-0.73).

Combining multiple clinical variables can improve the accuracy of a clinical diagnosis of VAP (Table 6). Four clinical findings have been studied in combination: radiographic infiltrates, fever, peripheral leukocytosis, and purulent sputum on gross inspection.⁴⁵ In a cohort of unselected patients who had pulmonary biopsies performed immediately after death, diagnostic usefulness increased when progressively more findings were combined. VAP was likely only when an infiltrate was accompanied by 2 to 3 additional clinical findings (LR, 2.8; 95% CI, 0.97-7.9). Those with an infiltrate and 0 or 1 additional findings were less likely to have VAP (LR, 0.46; 95% CI, 0.10-2.1 for no additional findings; LR, 0.37; 95% CI, 0.09-1.6 for 1 additional finding). Notably, however, even the absence of all 3 of fever, leukocytosis, and purulent secretions did not definitively exclude VAP: 15% of patients with an infiltrate and none of these clinical signs had histologically confirmed VAP.⁴⁵

Three different studies evaluated the CPIS predictive tool (CPIS criteria described in Table 1; performance evaluation of CPIS presented in Table 7).^{42 ,45 ,48} A CPIS score greater than 6 raises the likelihood of VAP among unselected patients (summary LR, 2.1; 95% CI, 0.92-4.8), while a CPIS score of 6 or less diminishes the likelihood (summary negative LR, 0.38; 95% CI, 0.20-0.74). Among patients already suspected of having VAP, a single study suggested that a CPIS score of 6 or less reduces the likelihood of pneumonia (LR, 0.08; 95% CI, 0.01-0.56).⁴⁸

Most clinical features classically sought to diagnose VAP have limited diagnostic usefulness because other common complications of critical illness generate similar pulmonary syndromes. These sources of false-positive results have been characterized. Meduri et al⁵⁰ were able to establish a diagnosis in 45 of 50 consecutive prospectively evaluated intensive care unit (ICU) patients with fever and pulmonary opacities. Only 42% were confirmed to have VAP. Other patients were ultimately diagnosed with acute respiratory distress syndrome, generalized sepsis, atelectasis, alveolar hemorrhage, pulmonary infarction, and/or congestive heart failure. Most patients had more than 1 diagnosis. Likewise, in a series of 30 ICU patients with respiratory failure and diffuse infiltrates who died and subsequently underwent autopsy, only 60% were found to have pneumonia.³⁴ Other diagnoses occurred with significant overlap, including diffuse alveolar damage (77%), alveolar hemorrhage (23%), carcinoma (10%), emphysema (13%), intravascular lymphoma (7%), pulmonary fibrosis (7%), tuberculosis (7%), and lymphangioleiomyomatosis (3%).

Possible causes of this patient's fever, infiltrates, and tachypnea include acute respiratory distress syndrome, recurrent thromboembolic disease or pulmonary infarction, pulmonary hemorrhage secondary to anticoagulation, congestive heart failure from cardiac strain and excessive fluid administration, hypersensitivity pneumonitis secondary to a medication, atelectasis from imperfect ventilation, metastases from an extrapulmonary malignancy, and pneumonia secondary to mechanical ventilation. The same clinical picture could also result from conditions in combination, such as central line–associated sepsis coincident with atelectasis or persistent pulmonary opacities from the patient's small cell lung cancer. This extensive differential diagnosis is unfortunately typical for a critically ill patient with a prolonged ICU stay, and indeed it is likely that the patient is experiencing more than 1 of these conditions. The prior probability of VAP is about 10% for ventilated patients.⁸

The patient's fever, leukocytosis, and radiographic infiltrate are all nonspecific findings that, when considered individually, do not substantively change the likelihood that VAP is present. The presence of these findings in combination, however, makes pneumonia more likely, yet is still insufficient to definitively establish the diagnosis. Quantitative culture of a blind bronchial aspirate can also help to diagnose VAP when positive, but this test takes 2 days to complete and requires specialized laboratory facilities to quantify the bacterial colony count. To increase diagnostic certainty, the patient's clinicians can consider performing diagnostic bronchoscopy to obtain BAL fluid for Gram stain, cell count, and differential. If cell count analysis of BAL fluid shows less than 50% neutrophils, then VAP is unlikely and investigations for other diagnoses should be pursued. On the other hand, if the result of a Gram stain of BAL fluid is positive for microorganisms, then VAP is likely present.

Critically ill patients are subject to a multitude of pathologic insults that create overlapping systemic and pulmonary signs of inflammation. Consequently, the isolated presence of most clinical features conventionally sought to diagnose VAP, such as fever, leukocytosis, purulent pulmonary secretions, and radiographic infiltrates, do not establish the diagnosis. Consideration of clinical findings in combination, however, can help guide further evaluation when suspicion for VAP arises.

Assessment of a chest radiograph is a practical place to begin. The appearance of a new infiltrate suggests that VAP is possible (summary LR, 1.7; 95% CI, 1.1-2.5) and should prompt a second look at the patient's temperature, sputum purulence, and leukocyte count. When 2 or more of these signs are positive, then VAP becomes more likely (summary LR, 2.8; 95% CI, 0.97-7.9). Taking a 9.7% incidence as the prior probability for VAP among ventilated patients,⁸ the above combination of findings will increase the likelihood of VAP to 23% (95% CI, 9.4%-46% after taking the limits of the LR confidence interval). Expert clinicians tend to also consider impairment of gas exchange as a necessary, albeit nonspecific and insufficient, criterion to diagnose VAP.²⁴ This intuitive requirement was not formally assessed in the independent studies included in this review.

In contrast, the absence of new infiltrates substantively lowers the likelihood of VAP (summary LR, 0.35; 95% CI, 0.14-0.87). The absence of fever, sputum purulence, or an elevated white blood cell count does not add information to that gained from a radiograph without infiltrates. If the pretest probability of VAP is 9.7%,⁸ then routine clinical evaluation including a stable chest radiograph without new infiltrates can lower the probability of VAP to 3.6% (95% CI, 1.5%-8.5%).

Analysis of pulmonary secretions can further refine the likelihood that VAP is present or absent. BAL fluid in which less than 50% of the cells are neutrophils argues against the presence of VAP, with an LR that ranges from 0.05 to 0.10. By contrast, the presence of microorganisms on a Gram stain of pulmonary secretions substantively increases the probability of VAP. The suggestiveness of a positive Gram stain result increases almost exponentially with progressively more invasive diagnostic techniques—the positive LR for a positive Gram stain result on a blind bronchial aspirate is only 2.1 (95% CI, 0.81-5.5), but this increases to 5.3 (95% CI, 1.3-22) with mini-BAL and up to 18 (95% CI, 1.1-302) on BAL fluid gathered via fiberoptic-guided bronchoscopy. Quantitative bacterial cultures of pulmonary secretions can also help diagnose VAP. Growth of more than 10⁵ colony-forming units/mL of bacteria from a blind bronchial aspirate is highly suggestive of VAP (summary LR, 9.6; 95% CI, 2.4-38). A threshold of only 10⁴ colony-forming units/mL on BAL fluid, by contrast, is not definitive. This is likely a reflection of greater potential for contaminant organisms to reach this lower growth threshold. Negative quantitative cultures of both bronchial aspirates and BAL fluid only moderately decrease the probability of VAP.

The finding that analysis of both blind bronchial aspirate and BAL can contribute useful information but that neither is definitive is consistent with a recent randomized clinical trial showing no difference in 28-day mortality regardless of which of these 2 diagnostic strategies was initially used.⁵¹ Patients were selected to be included in that study if they manifested the classic clinical features of VAP analyzed in this review. All patients were given broad-spectrum antibiotics immediately after diagnostic sampling. Approximately 40% of the patients in both groups of the trial, however, were not cured by the end of the study, raising the possibility that their clinical syndrome was not caused by true, histological VAP but by some other etiology. This again bespeaks the imperfect accuracy of clinical signs for VAP. The trial teaches that obtaining blind bronchial aspirates might be a reasonable starting procedure for the majority of patients, but patients who do not improve with antibiotics guided by this strategy merit further diagnostic workup. This could include bronchoscopy to further evaluate whether VAP is present, as well as other specialized studies to assess for noninfectious causes of pulmonary disease.

The findings of this review support the consensus opinions published by the Infectious Diseases Society of America and the American Thoracic Society.²⁴ Their guidelines acknowledge the lack of a definitive gold standard to diagnose VAP and the practical difficulty of making the diagnosis at the bedside. Both societies recommend a combination of clinical signs and quantitative microbiological data to diagnose and manage VAP, while emphasizing the importance of constant reevaluation of the diagnosis over time. Clinicians should suspect VAP when chest radiography reveals the presence of a new or progressive infiltrate and the patient has at least 2 of fever, abnormal white blood cell count, or purulent secretions. As soon as VAP is suspected, clinicians are urged to immediately obtain a lower respiratory tract specimen for microscopy and culture (quantitative or semiquantitative). Because of the uncertainty in diagnosis when a patient is initially examined, antibiotics are typically started unless the pretest probability for VAP is very low. After 48 to 72 hours of antibiotic therapy, clinical response should be assessed by reevaluating the patient's temperature, white blood cell count, oxygenation, and results of chest radiography, pulmonary secretion microscopy, and secretion culture. Absence of improvement should prompt a search for an alternative diagnosis or a reason for therapeutic failure.

The findings of this review are tempered by limitations of the source data. Many of the conclusions were based on a limited number of small studies that differed in both their methods and their findings. Moreover, even a histological gold standard is not without controversy. First, series limited to patients ultimately undergoing autopsy are inherently biased toward the sickest patients. Second, pulmonary biopsies risk missing the isolated patchy lesions that are characteristic of VAP. Third, interobserver agreement between pathologists has been reported to be only moderately good (κ = 0.45).¹⁰ Finally, the interpretation of pulmonary pathology is complicated by delay between the initial clinical development of pneumonia and the subsequent autopsy, during which interim healing and antibiotic exposure can complicate the histological picture.

The best available evidence suggests that clinical examination can be used to alert physicians to the possibility of VAP but that examination alone is insufficient to establish a definitive diagnosis. Examination of pulmonary secretions can help refine physicians' clinical suspicions. The presence of bacteria on Gram stain or bacterial growth above a quantitative threshold favors the diagnosis, while less than 50% neutrophils in a pulmonary specimen makes VAP less likely. The absence of a new infiltrate on chest radiograph also makes disease unlikely. Clinicians caring for ventilated patients with a clinical syndrome consistent with VAP should be ready to consider additional diagnoses and further investigations, particularly when an empirical trial of antibiotics does not lead to improvement within 48 to 72 hours.