A diagnostic model for discrimination between Pneumocystis jirovecii pneumonia and colonization based on multiple parameters

Distinguishing Pneumocystis jirovecii pneumonia (PJP) from colonization (PJC) is crucial due to overlapping symptoms but different treatments. This study aims to evaluate whether peripheral blood parameters can serve as a non-invasive tool for distinguishing PJP from PJC. We retrospectively enrolled 174 patients with PJP and 61 with PJC from the First Affiliated Hospital of Sun Yat-sen University (April 2022-March 2024). Peripheral blood parameters were analyzed and compared between groups. Normally distributed variables were assessed using Student's t-test, while non-parametric data were analyzed with the Wilcoxon rank-sum test. A diagnostic model was subsequently developed based on significant hematological indicators. Utilizing a significance threshold of p < 0.05, red blood cell (RBC) and lymphocyte percentage (Lym%), while excluding neutrophil percentage (Neu%), procalcitonin (PCT), and lactic dehydrogenase (LDH) were used to build a random forest diagnostic model. The optimal XGBoost model achieved an AUC of 0.9991 internally and 0.787 in external validation. A web-based tool was developed to assist diagnosis. The findings of this study offer an effective tool for clinical practice, enabling physicians to accurately diagnose and differentiate between PJP and PJC, guiding appropriate treatment for patients.