Prediction of pulmonary aspergillosis in patients with ventilator-associated pneumonia

Background Predictors of ICU-acquired pulmonary aspergillosis (IPA) are not well-established in critically ill patients with ventilator-associated pneumonia (VAP), making IPA commonly misdiagnosed and anti-fungal therapy delayed. We aimed to develop a clinical score for prediction of IPA among patients with VAP. Methods Mechanically ventilated patients who developed VAP in 4 ICUs in Bretagne, Western France, were included. The score was constructed in a learning cohort, based on predictors of IPA in logistic regression model, and validated in a validation cohort. Results Among 1636 mechanically ventilated patients, 215 developed VAP but only 39 developed IPA (4 possible and 35 probable/putative) (18%). Most cases (31/39) were documented through a positive broncho-alveolar sample culture. Independent predictors of IPA were immunodepression (including onco-hematological disorder, immunomodulatory treatment, solid organ transplant, neutropenia < 0.5G/L and high-dose steroids ≥ 1 mg/kg/day of prednisolone equivalent) ( p = 0.001; score = 1 point) and lymphocyte count at admission < 0.8 G/L ( p = 0.019; score = 1 point). Operational values of the predictive score in the learning/validation cohort were 50%/52% sensitivity and 90%/87% specificity, respectively, for high PiPa score (score = 2) and 94%/91% sensitivity and 44%/46% specificity, respectively, for moderate PiPa score (score = 1). Finally, the AUC for the prediction of IPA was 0.783 in the learning cohort and 0.770 in the validation cohort. Conclusions We evaluated a clinical score with good predictive value which may help to predict IPA in patient with VAP. External validation will be needed to confirm our preliminary findings.