Predicting Mortality in Intensive Care Unit Patients With Allergic Bronchopulmonary Aspergillosis (ABPA) Using an...

Area of Science:

• Pulmonary Medicine
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Allergic Bronchopulmonary Aspergillosis (ABPA) is a severe lung disease requiring intensive care unit (ICU) admission.
• Predicting in-hospital mortality for ICU ABPA patients is critical for timely clinical interventions.

Purpose of the Study:

• To develop and validate an explainable machine learning model for predicting in-hospital mortality in ICU ABPA patients.
• To identify key clinical factors associated with mortality in this patient cohort.

Main Methods:

• Retrospective analysis of clinical data from 82 ICU ABPA patients.
• Development of a prediction model using an explainable XGBoost algorithm.
• Interpretation of model predictions using SHapley Additive exPlanations (SHAP) and internal validation.

Main Results:

• The XGBoost model achieved high predictive performance (AUC 0.995 training, 0.881 validation).
• Key mortality predictors included BMI, peak procalcitonin, peak eosinophil count, age, asthma history, peak leukocyte count, and lowest platelet count.
• In-hospital mortality rates were 46.3% in the training set and 48% in the validation set.

Conclusions:

• The explainable XGBoost model effectively predicts in-hospital mortality in ICU ABPA patients.
• SHAP analysis provides interpretable insights into mortality predictors.
• External validation and multicenter studies are recommended to improve generalizability and optimize individualized patient care.