Ventilator-associated pneumonia rates: The wrong quality measure

Legislators, payers and quality-of-care advocates across the nation are considering requiring hospitals to report ventilator-associated pneumonia (VAP) rates as a way to benchmark quality of care and penalize poor performers.


Legislators, payers and quality-of-care advocates across the nation are considering requiring hospitals to report ventilator-associated pneumonia (VAP) rates as a way to benchmark quality of care and penalize poor performers. A handful of states, including Illinois, Pennsylvania and South Carolina, have already passed legislation requiring hospitals to report their rates of VAP, and other states are actively considering adding VAP to existing laws that mandate the reporting of other health care-associated infections.

In a similar vein, the Centers for Medicare and Medicaid Services has announced a new policy of withholding payments to hospitals for costs incurred in treating preventable complications of medical care. The current list of uncompensated complications doesn't include VAP, but given the cost and morbidity of this disorder, observers suspect that it might eventually be added. In addition, more than 3,100 U.S. hospitals are participating in the Institute for Healthcare Improvement's 5 Million Lives Campaign, which includes a bundle of measures to prevent VAP. The campaign advocates VAP surveillance to assess the impact of these preventive measures.

The unspoken premise underlying these well-intentioned initiatives is that it is possible to diagnose VAP efficiently and accurately. In contrast to most other quality indicators, however, VAP surveillance is highly labor intensive yet inherently inaccurate. It cannot yield concrete, reproducible results appropriate for benchmarking or quality assessment.

Inaccurate diagnosis

The clinical diagnosis of VAP is notoriously imprecise. Autopsy series for patients who received mechanical ventilation attest to a striking inaccuracy in physicians' diagnoses of VAP. One-third of patients given a clinical diagnosis of VAP have no evidence of pneumonia at autopsy, while one-quarter of mechanically ventilated patients who die without a clinical diagnosis of VAP do.

Physicians' ability to diagnose VAP is poor because many pulmonary complications of intensive care present with similar clinical signs. VAP is usually diagnosed when a patient has some combination of fever, abnormal leukocyte count, increased or purulent pulmonary secretions and a new radiographic infiltrate. Each of these findings, however, is a nonspecific sign common to many other conditions that complicate critical care. Intensive evaluations of ventilated patients with fever, pulmonary infiltrates, purulent sputum or some combination of the three reveal that only 30% to 40% have VAP. The rest have one or more of the following conditions:

  • Atelectasis,
  • Pulmonary edema,
  • Thromboembolic disease,
  • Acute respiratory distress syndrome,
  • Alveolar hemorrhage,
  • Hypersensitivity pneumonitis, and
  • Pulmonary contusion.

Not infrequently, two or more disorders—for example, bloodstream infection and pulmonary edema—can develop in a patient simultaneously and together can mimic VAP perfectly.

Clinicians might believe that an increase in pulmonary secretions or a positive pulmonary culture is highly specific for VAP, but such assumptions are not supported by the literature. Abundant or purulent pulmonary secretions are frequently present in intubated patients, regardless of their underlying diagnosis, because of the disruption of physiologic mechanisms for clearing the fluids that are constantly produced by the sinuses, oropharynx and lungs. Similarly, positive cultures of respiratory secretions are only moderately specific for VAP. The mouths of intubated patients become colonized with pathologic organisms within a week of hospital admission. These bacteria can contaminate respiratory cultures and mislead clinicians into suspecting parenchymal infection. Even quantitative cultures of protected-specimen brush specimens obtained by bronchoscopy overdiagnose VAP about 30% of the time, while missing up to 50% of patients with histologically confirmed disease.

Guidelines published jointly by the American Thoracic Society and the Infectious Diseases Society of America acknowledge the imperfections of both clinical and bacteriologic diagnostic strategies for VAP. Reliance on clinical signs leads to overdiagnosis, and an invasive strategy involving bronchoalveolar lavage leads to underdiagnosis. A recent randomized trial comparing diagnosis based on endotracheal aspirate culture with diagnosis based on bronchoalveolar lavage culture showed no difference in clinical outcomes.

Clinicians typically compensate for the uncertainty of diagnosing VAP by empirical use of antibiotics when signs and symptoms consistent with the condition are present. The uncertainty surrounding diagnosis of VAP is more problematic, however, when performing surveillance for public reporting or to determine compensation. Different interpretations of nonspecific and ambiguous clinical signs can yield very different rates of VAP.

The CDC has published a formal definition of nosocomial pneumonia (see Table), designed for anonymous epidemiologic surveillance by a core set of designated hospitals rather than for universal, high-stakes public reporting. The CDC definition, however, uses the same nonspecific criteria that clinicians grapple with at the bedside. In addition, the CDC definition permits subjective interpretation of key points. For example, surveyors are asked to assess for a “change in character of sputum” or “increased ventilation demand.” Precise criteria for what constitutes a meaningful change in respiratory secretions or ventilation requirements are left to the discretion of the observer. Likewise, the CDC criteria do not specify minimum durations for observed changes, the appropriate way of handling differences of opinion among observers or the necessary qualifications of the observer. The subjectivity and complexity of the definition not only invite inaccuracy but also make surveillance expensive and time-consuming to implement.

An unreliable quality measure

The difficulty in accurately diagnosing VAP and the subjective nature of the CDC criteria make VAP an unreliable basis for internal quality control, interhospital benchmarking of quality of care or withholding hospital compensation. Measures selected for quality assessment or benchmarking ought to yield consistent results regardless of where or in whom they are applied. They should also closely reflect processes of care that hospitals can modify to improve outcomes. The current definition of VAP does not meet these standards. The rates that are calculated on the basis of the existing criteria risk being confounded by interobserver variation and the presence of other conditions in the intensive care unit that can mimic VAP.

Furthermore, if VAP rates are used in determining hospitals' compensation or are factored into their public reputations, the rate of diagnosis may decline simply because observers shift their interpretations of criteria such as “change in secretion character” or “worsening gas exchange.” Even well-intentioned observers might report substantial decreases in VAP rates that are more a reflection of haziness in the surveillance definition than true improvements in the quality of care. Such changes in interpretation might not be deliberate—they would certainly be almost impossible to detect.

The subjectivity and lack of specificity of the current definition make comparison of different hospitals' VAP rates not only uninterpretable but also potentially harmful. Mandatory reporting of VAP rates could paradoxically hurt patients, for example, if a hospital were lulled into complacency by comparatively low VAP rates that in fact reflected a narrow interpretation of surveillance criteria rather than excellence in clinical care. Conversely, other hospitals would risk being penalized for applying the definition more broadly, thereby inflating their VAP rates relative to those of peer institutions. In quality reporting, as in clinical care, the first principle ought to be primum non nocere.

Toward a solution

Currently, we have no obvious alternative quality measure for ventilated patients that is objectively measurable, is indicative of serious complications and can reliably reflect quality of care. We need to develop new outcome measures that will meet these standards.

Ideally, any new measure or measures should reflect the broader array of pulmonary complications that can befall ventilated patients in addition to pneumonia, such as pulmonary embolism, atelectasis, the acute respiratory distress syndrome and pulmonary edema. Recording all of these events will give a more thorough picture of the quality of an intensive care unit's pulmonary care and will emphasize the provision of the best possible comprehensive care, rather than solely the prevention of pneumonia. New measures should also be relatively straightforward for hospitals to collect. Until new outcome measures are developed, we recommend an interim strategy of tracking evidence-based process-of-care measures, such as daily cessation of sedation and appropriate patient positioning, because these practices have been shown to reduce intensive care length of stay.

The concept of benchmarking outcomes to inspire improvements in care, reward best practices and inform consumer choice is laudable. The limitations of the current surveillance definition of VAP, however, preclude its use for this purpose. We need to develop objective new quality measures for ventilated patients that are easier to collect and that more reliably reflect the outcomes clinicians and patients care about: serious, preventable complications of mechanical ventilation.