Machine learning for early screening of influenza A-associated invasive pulmonary aspergillosis in hospitalized patients: a real-world study.

Background: Influenza A-associated invasive pulmonary aspergillosis (IAPA) is a severe fungal complication with high mortality, while early identification remains difficult because of nonspecific clinical manifestations. This study aimed to develop and validate a machine learning (ML) model for early screening of IAPA in hospitalized influenza A patients. Methods: This retrospective single-center study enrolled 234 hospitalized influenza A patients from January 2023 to December 2025, including 59 patients with IPA. Eligible patients were randomly divided into a training cohort (70%) for model development and a validation cohort (30%) for internal validation. Univariate and multivariate logistic regression analyses were performed to identify independent predictors of IAPA. Five machine learning algorithms, including Logistic Regression (LR), Random Forest (RF), Support Vector Machine (SVM), XGBoost, and LightGBM, were constructed and compared to identify the most clinically applicable model for early screening. Results: Multivariate logistic regression identified seven independent predictors of IAPA, including smoking history, autoimmune disease, fibrinogen level, lymphocyte count, hemoglobin level, cumulative systemic corticosteroid dose, and corticosteroid treatment course of 8-28 days. Among the evaluated algorithms, LightGBM demonstrated the highest sensitivity (0.76) in the validation cohort and was considered the most suitable model for early screening. The LightGBM model achieved an AUC of 0.800 (95% CI, 0.714-0.886), with a specificity of 0.69 and an accuracy of 0.68. Conclusion: LightGBM serves as a robust early-warning tool for identifying influenza A patients at high risk of IAPA. Utilizing routinely available clinical data, this model facilitates bedside risk stratification and early diagnostic intervention.