RESUMEN
BACKGROUND: In the absence of lung biopsy, there are various algorithms for the diagnosis of invasive pulmonary aspergillosis (IPA) in critically ill patients that rely on clinical signs, underlying conditions, radiological features and mycology. The aim of the present study was to compare four diagnostic algorithms in their ability to differentiate between probable IPA (i.e., requiring treatment) and colonisation. METHODS: For this diagnostic accuracy study, we included a mixed ICU population with a positive Aspergillus culture from respiratory secretions and applied four different diagnostic algorithms to them. We compared agreement among the four algorithms. In a subgroup of patients with lung tissue histopathology available, we determined the sensitivity and specificity of the single algorithms. RESULTS: A total number of 684 critically ill patients (69% medical/31% surgical) were included between 2005 and 2020. Overall, 79% (n = 543) of patients fulfilled the criteria for probable IPA according to at least one diagnostic algorithm. Only 4% of patients (n = 29) fulfilled the criteria for probable IPA according to all four algorithms. Agreement among the four diagnostic criteria was low (Cohen's kappa 0.07-0.29). From 85 patients with histopathological examination of lung tissue, 40% (n = 34) had confirmed IPA. The new EORTC/MSGERC ICU working group criteria had high specificity (0.59 [0.41-0.75]) and sensitivity (0.73 [0.59-0.85]). CONCLUSIONS: In a cohort of mixed ICU patients, the agreement among four algorithms for the diagnosis of IPA was low. Although improved by the latest diagnostic criteria, the discrimination of invasive fungal infection from Aspergillus colonisation in critically ill patients remains challenging and requires further optimization.