基于3种机器学习方法的神经精神性狼疮影响因素的研究OA

Objective To explore influencing factors of neuropsychiatric systemic lupus erythematosus using LASSO,XGBoost,and RF combined with multivariate Logistic regression.Methods Medical records of 401 systemic lupus erythematosus(SLE)patients hospitalized in the Department of Rheumatology and Immunology in the Second Hospital of Shanxi Medical University from January 2021 to June 2024 were collected.Based on the diagnosis of neuropsychiatric systemic lupus erythematosus(NPSLE),the patients were divided into N-NPSLE group(without NPSLE)and NPSLE group,with the remaining variables serving as explanatory variables.Characteristic variables were selected using LASSO,XGBoost,and RF methods.Common variables from the three methods were as-sessed for multicollinearity,and analyzed by multivariate Logistic regression.Results A total of 18 variables were selected by LASSO,21 variables were selected by XGBoost,and 20 variables were selected by RF.Eight common variables were selected by the three methods,including age,headache,IgA,C3,activated partial thromboplastin time(APTT),anti-ribosomal P protein,and soluble interleukin-2 receptor(sIL-2R),and systemic lupus erythematosus disease activity index(SLEDAI-2k)score.The variance in-flation factor(VIF)values from the multicollinearity test showed that there was no multicollinearity among the variables.Multivariate Logistic regression analysis revealed that age was a protective factor for the incidence of NPSLE in SLE patients(OR=0.936,95％CI 0.904-0.969),while headache(OR=47.153,95％CI 13.065-170.214),anti-ribosomal P protein positive(OR=1.082,95％CI 1.036-1.129),and higher SLEDAI-2k score(OR=5.176,95％CI 3.210-8.347)were risk factors.Conclusion Age is a protective factor for NPSLE,and headache,anti-ribosomal P protein positive,and higher SLEDAI-2k score are risk factors for NPSLE.