Preferred Studies
The microbiologic approach to diagnosis relies on the use of quantitative or semi-quantitative cultures to create thresholds for antibiotic treatment. Bacterial cultures that demonstrate a level of growth above the thresholds described below warrant treatment, while those below it should trigger withholding or discontinuation of antibiotics.
Bronchoscopic Approaches: Bronchoalveolar lavage (BAL) with a cutoff of 10 (4) organisms/mL or protected specimen brush (PSB) with a cutoff of 10 (3) organisms/mL are felt to be the most specific diagnostic tests when performed prior to initiating antibiotics, or prior to changing antibiotics if a patient is already receiving them. In clinically stable patients, antibiotics can be safely discontinued if bacterial growth falls below the thresholds. If cultures are positive, antibiotic therapy should be tailored to target the organism identified. The bronchoscopic approach is favored in patients who are mechanically ventilated, develop their pneumonia late in the hospital stay (>5–7 days), are at risk for unusual pathogens, are failing therapy or suspected of having an alternative diagnosis.
Non-Bronchoscopic Approaches: Qualitative endotracheal aspirates (ETA) have been shown to be quite sensitive in ventilated patients, regularly identify organisms that may be subsequently found by BAL or PSB, and if negative, should result in withholding antibiotics. Quantitative endotracheal aspirates with a cutoff of 10 (6) organisms/mL are often encouraged to reduce antibiotic overuse, but results should be interpreted cautiously as they only have a sensitivity and specificity of about 75% (1). Consideration should be given to withholding antibiotics in a clinically stable patient with a negative quantitative ETA if antibiotics have not been changed in the preceding 72 hours. Many ICUs have begun to perform blinded sampling of lower respiratory tract secretions with suction catheters (blind PSB, blind mini-BAL). These techniques can be performed at all hours by trained respiratory therapists or nurses, provide culture data similar to that of bronchoscopy, and may be safer and less costly than bronchoscopy. In general, non-bronchoscopic techniques are preferred in patients who are not mechanically ventilated. Sputum sampling, while easy to obtain, has not been well studied in NP. However, in patients in whom bronchoscopic or other non-bronchoscopic techniques are not feasible, sputum sampling may be performed to identify potentially resistant organisms and help tailor therapy.
Alternative Options
Clinical Pulmonary Infection Score—Combining Clinical and Microbiologic Approaches
The clinical diagnosis of nosocomial pneumonia (new infiltrate + fever, leukocytosis, or purulent sputum) likely leads to antibiotic overuse, yet pursuing a bronchoscopic diagnosis is invasive, costly, and requires technical expertise. The quantitative ETA, blind PSB, and blind BAL discussed above are examples of some compromises that avoid the need for bronchoscopy, yet add microbiologic data in an attempt to prevent excess antibiotic therapy. Formally combining diagnostic approaches (clinical + microbiologic) may also be useful. One such option is the use of the clinical pulmonary infection score (CPIS), which combines clinical, radiographic, physiological, and microbiologic data into a numerical result. Scores >6 have been shown to correlate well with quantitative BAL (4). More recent studies, however, have suggested a lower specificity which could still result in antibiotic overuse, but this approach remains more accurate than a general clinical approach. Using the CPIS serially at the time NP is suspected and again at 72 hours may be more useful. Patients with an initial low clinical suspicion for pneumonia (CPIS of 6 or less) could have antibiotics safely discontinued at 72 hours if the CPIS remains low (5). Such a strategy may be useful in settings where more sophisticated diagnostic modalities are not available.
Multiple studies of biological markers of infection have attempted to find a non-invasive, rapid, accurate means of determining who needs antibiotics for presumed NP. Unfortunately, the results have largely been disappointing. More recently, measurement of a soluble triggering receptor expressed on myeloid cells (sTREM-1) that is upregulated in the setting of infection has been shown to improve our ability to diagnose NP accurately. Measurement of sTREM-1 was 98% sensitive and 90% specific for the diagnosis of pneumonia in mechanically ventilated patients (6). While promising, more data is needed before this test can be recommended for routine use.