Ventilator-associated pneumonia in intubated children: comparison of different diagnostic methods.

Objectives: To compare different methods for diagnosis of ventilator-associated pneumonia in intubated children.

Design: Prospective epidemiologic study.

Setting: Pediatric intensive care unit of a tertiary care university hospital.

Patients: All consecutive pediatric intensive care unit patients <18 yrs of age with suspected ventilator-associated pneumonia.

Interventions: For all patients, the following diagnostic methods were compared: 1) clinical data using Centers for Disease Control criteria; 2) blind protected bronchoalveolar lavage, evaluating quantitative cultures, bacterial index of >5, Gram stain, and presence of intracellular bacteria; and 3) nonquantitative cultures of endotracheal secretions. The reference standard used was clinical judgment of three independent experts (Delphi method) who retrospectively established by consensus the presence of ventilator-associated pneumonia based on clinical, microbiological, and radiologic data. Concordance between each diagnostic method and the reference standard was evaluated by concordance percentage and kappa score. Validity was evaluated using sensitivity, specificity, positive predictive value, negative predictive value, and global value.

Results: A total of 30 patients were included in the study. According to the reference standard, ventilator-associated pneumonia occurred in 10 of 30 patients (33%). Bacterial index of >5 in bronchoalveolar secretions showed the best concordance compared with the reference standard (concordance, 83%; kappa, 0.61). Bacterial index of >5 also showed the best validity (sensitivity, 78%; specificity, 86%; positive predictive value, 70%; negative predictive value, 90%; global value, 90%). Intracellular bacteria and Gram stain from bronchoalveolar secretions were very specific (95% and 81%, respectively) but not sensitive (30% and 50%, respectively). Clinical criteria and endotracheal cultures were very sensitive (100% and 90%, respectively) but poorly specific (15% and 40%, respectively).

Conclusion: Our data show that the most reliable diagnostic method for ventilator-associated pneumonia is a bacterial index of >5, using blind protected bronchoalveolar lavage. Further studies should evaluate the validity of all these methods according to the gold standard (autopsy).