Predicting Pneumocystis jirovecii Infection in AIDS Patients Using Machine Learning: A Hospital‐Based Study in Northwest Hunan

Objective To develop and validate a machine learning model for predicting Pneumocystis jirovecii infection in patients with acquired immunodeficiency syndrome (AIDS). Methods Clinical data from 651 AIDS patients, including 93 with P. jirovecii coinfection, were retrospectively collected from our hospital information system. Five machine learning algorithms—logistic regression, decision tree (DT), k‐nearest neighbors (KNN), extreme gradient boosting (XGBoost), and support vector machine (SVM)—were used to build prediction models. Models were trained on 70% of the data and tested on the remaining 30%. Performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). SHAP analysis was applied to interpret the contribution of each variable. Results Twenty‐six clinical variables including ethnicity, pneumonia, max temperature, average temperature, initial respiratory rate, max respiratory rate, max pulse, average pulse, hematocrit, percentage of lymphocytes, percentage of monocyte, absolute number of lymphocytes, absolute number of monocyte, hematocrit of platelet, aspartate aminotransferase, albumin, globulin, albumin/globulin ratio, total bilirubin, indirect bilirubin, glomerular filtration rate, uric acid, sodium, fibrinogen, lactate dehydrogenase, and initial CD4‐positive T‐cell count were identified as potentially related to P. jirovecii infection in AIDS patients. The machine learning model demonstrated good predictive performance in the training and test group. Conclusion The developed machine learning model effectively identifies AIDS patients at risk of P. jirovecii infection and may support early risk stratification and clinical decision‐making. External validation is needed to confirm its generalizability.