A Clinical Prediction Model for Invasive Pulmonary Fungal Disease in Dialysis Patients: Development, Validation, and Pharmacist-Led Application

Jiao Yang,1,2,* Dongmei Wang,3,* Meng Li,1,* Shuangyuan Li,1,4 Dong Xu,5 Zhongju Chen,6 Jian Yang,7 Yan He1 1Department of Pharmacy, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, People’s Republic of China; 2Department of Pharmacy, The People’s Hospital of Laifeng County, Laifeng, Hubei, People’s Republic of China; 3Department of Transfusion Medicine, General Hospital of Central Theater Command of Chinese People’s Liberation Army, Wuhan, Hubei, People’s Republic of China; 4Department of Pharmacy, Yidu Zhicheng Town Health Center, Yidu, Hubei, People’s Republic of China; 5Department of Infectious Diseases, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, People’s Republic of China; 6Clinical Microbiology Laboratory, Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, Hubei, People’s Republic of China; 7Department of Pharmacy, General Hospital of Central Theater Command of Chinese People’s Liberation Army, Wuhan, Hubei, People’s Republic of China*These authors contributed equally to this workCorrespondence: Yan He, Email [email protected] Jian Yang, Email [email protected]: Patients receiving maintenance hemodialysis (MHD) are particularly vulnerable to fungal infections attributable to immunocompromised status, frequent therapeutic interventions, and environmental exposure. Notably, invasive pulmonary fungal infection (IPFI) carries particularly high morbidity and mortality rates. The non-specific clinical manifestations coupled with inconclusive microbiological findings frequently result in diagnostic delays. Compounded by the narrow therapeutic window, these factors pose significant clinical challenges and adversely affect patient outcomes. Typical misdiagnosis rates for IPFI in MHD patients exceed 50%, with median diagnostic delays of 10– 14 days, contributing to mortality rates up to 40% from delayed antifungal therapy. The aim of this study was to develop and validate a clinical prediction model for identifying IPFI in MHD patients.Methods: This retrospective cohort study included MHD patients with pneumonia admitted between January 2022 and December 2023 for the modeling cohort (IPFI group, n = 51; non-IPFI group, n = 951). An external validation cohort consisted of patients admitted from January to December 2024 (IPFI group, n = 22; non-IPFI group, n = 463). Predictors were screened from 22 clinical variables using least absolute shrinkage and selection operator (LASSO) regression, and a prediction model was constructed using logistic regression. Predictors were standardized (z-score normalization) before LASSO to ensure comparability.Results: A final four-variable model was developed, including highest body temperature, hypothyroidism, pancytopenia, and anti-neutrophil cytoplasmic antibody (ANCA)-associated disease. The model demonstrated strong discrimination, good calibration, and strong performance in the calibration plot. The model showed good internal validity (optimism-corrected AUC = 0.864; calibration slope = 0.97) and maintained robust performance in external validation (AUC = 0.877, 95% CI: 0.800– 0.948; calibration slope = 1.12, 95% CI: 0.828– 1.511). In the LASSO path, pancytopenia showed the highest importance (standard β=0.342), followed by maximum body temperature (standard β=0.289), indicating stronger effect sizes for these variables. A nomogram was subsequently developed based on the model.Conclusion: This prediction model effectively identifies MHD patients with pneumonia at high risk of developing IPFI, demonstrating robust predictive performance. This tool can facilitate early risk stratification by clinical pharmacists in ward decision-support and stewardship consultations, optimizing the prophylactic and targeted use of antifungal agents and promoting rational drug therapy.Keywords: dialysis, fungal infection, fever, pancytopenia, hypothyroidism, ANCA, prediction model, nomogram, pharmaceutical services