A Predictive Model for Pulmonary Aspergillosis in ICU Patients: A Multicenter Retrospective Cohort Study.

BACKGROUND

Several predictive models for invasive pulmonary aspergillosis (IPA) based on clinical characteristics have been reported. Nevertheless, the significance of other concurrently detected microorganisms in IPA patients is equally noteworthy. This study aimed to develop a risk prediction model for IPA by integrating clinical and microbiological characteristics.

METHODS

This retrospective study was conducted in adult intensive care units (ICUs) of 17 medical centers in China. Clinical data were collected from patients with severe pneumonia who underwent clinical metagenomics of bronchoalveolar lavage fluid between January 1, 2019, and June 30, 2023. Subsequently, patients were randomly assigned to training and validation cohorts in a 7:3 ratio. In the training cohort, potential influencing factors were identified through univariate analysis, clinical practice, and existing literature, and a risk prediction model was constructed using multivariate logistic regression analysis. The performance of this model was then assessed and validated in the validation cohort.

RESULTS

Out of 1737 patients initially included in the study, 898 were ultimately analyzed, of which 100 (11%) were diagnosed with IPA. The risk prediction model for IPA, incorporating microbiological characteristics, identified six independent risk factors, namely age, immunosuppression, chronic kidney disease, connective tissue disease, liver failure, and cytomegalovirus positivity. The model demonstrated a superior discriminative ability, with area under the curve (AUC) values of 0.791 and 0.792 in the training and validation cohorts, respectively. Sensitivity and specificity reached 73.1% and 74.9%, respectively, and the model demonstrated good calibration.

CONCLUSION

This study developed a novel risk prediction model for IPA incorporating microbiological characteristics based on clinical metagenomics. The model exhibited good discriminative ability and calibration.