Impact of corticosteroid treatment on lung microbiome in patients with severe community-acquired pneumonia

In recent years, certain pneumonia patients have shown an unexpected lack of responsiveness to corticosteroid treatment. In this study, we aim to develop an explainable AI model for personalized treatment recommendations and identify microbiome-based biomarkers that can predict corticosteroid treatment outcomes. Sputum microbiome 16S ribosomal RNA sequencing data from severe community-acquired pneumonia patients admitted to the intensive care unit were collected over three time points (Day 1, 3 and 7). Key microbial taxa, including Alphaproteobacteria and Bacilli, were identified as potential biomarkers. These biomarkers were used to construct a random forest model, which achieved an accuracy of 0.8056 and an F1 score of 0.8571 in predicting treatment recommendations based on their relative abundance. By distinguishing patients who are likely to benefit from corticosteroids and those at risk of adverse effects, this study highlights the potential of microbiome-based biomarkers to enhance and tailor therapeutic strategies for pneumonia patients.