Development and multicenter validation of a predictive nomogram for infection risk in patients with multiple myeloma

Abstract

To explore the key influencing factors of the risk of nosocomial infection in patients with multiple myeloma (MM), to construct a prediction model to effectively assess the risk of infection, and to provide a basis for individualized infection prevention and control. 245 patients with MM admitted to the First Affiliated Hospital of Guangzhou University of Traditional Chinese Medicine from 2005 to 2024 were retrospectively included, and the patients’ gender, age, BMPC%, time of initial diagnosis, primary symptom at the time of diagnosis, underlying disease, clinical staging (DS staging and ISS staging), treatment regimen, results of bacterial and fungal cultures, and laboratory indexes (e.g., HGB, Rdw%, platelet corpuscular pressure PCT, uric acid, creatinine, β2-MG, CRP, and M protein content). Potential risk factors were screened by univariate and multivariate logistic regression analyses, and logistic regression models were constructed in R language and verified for predictive efficacy with external data. ISS stage III, lower HGB, lower ALB, higher CRP, higher M-protein content, lower plateletcrit (PCT) and higher uric acid (UA) were independent predictors of infection (all P < 0.05), and the predictive model performed well, with a model AUC = 0.988, which was significantly better than that of a single indicator; the model was validated by external data in a cohort with AUC = 0.839, which validated the model’s predictive efficacy, verifying the generalization ability of the model. The prediction model integrating ISS staging, HGB, ALB, CRP, Monoclonal Protein content, PCT, and UA has excellent discriminatory and calibration abilities, and may provide a reliable tool for risk stratification of infection in MM patients.