| Objective: This study aimed to investigate the risk factors for infection in SLE,and then construct a clinical prediction model to predict the risk of infection in SLE patients,and perform internal and external validation of the model to assess the differentiation,calibration and clinical utility of the model to provide a theoretical basis for clinical identification and prediction of those with a high risk of infection in SLE patients.Methods: According to the inclusion and exclusion criteria,121 SLE patients hospitalized in the Department of Rheumatology and Immunology of the Second Hospital of Lanzhou University were selected as the study subjects(61 cases in the infected group and 60 cases in the non-infected group),27 SLE patients in the inpatient department of the Department of Rheumatology and Immunology of the People’s Hospital of Gansu Province and 29 SLE patients in the inpatient department of the Department of Rheumatology and Immunology of the Second Hospital of Lanzhou University were selected as the external validation group,and the baseline information of the infection group was compared with that of the control group,and LASSO regression analysis was performed to screen the influencing factors,and the screened influencing factors were then subjected to logistic multi-factor regression analysis to obtain the final predictors,and then the nomogram was constructed based on the obtained predictors,and finally the nomogram was evaluated by internal and external validation and differentiation,calibration,and clinical utility.Results: First of all,compared with SLE non-infected patients,SLE co-infected patients were more likely to have respiratory symptoms,pleurisy,pericarditis,and fever,and often had moderate to high disease activity and long-term using high-dose hormone(P< 0.05),while patients’ neutrophil ratio,BUN,creatinine,uric acid,homocysteine,complement C1 q,CRP,PCT,IL-6,ferritin,ANu A positive rate,urine protein level,urine occult blood,UTP,urinary leukocyte count were significantly higher than those of non-infected patients(P < 0.05),while lymphocyte ratio,eosinophil count,hemoglobin,HDL,albumin,Ig A,and complement C3 levels were significantly higher in non-infected patients with SLE than infected patients(P < 0.05).Secondly,the common sites of infection for SLE co-infection were the respiratory tract(34.5%),the urinary tract(26.4%)and the plasma cavity(25.3%).Bacteria were the most common infecting agent(62.5%),followed by mycoplasma pneumoniae(20.8%),fungi(8.3%)and viruses(8.3%).Thirdly,LASSO regression analysis showed that white blood cell count,neutrophil count,basophil count,hemoglobin,albumin,Ig A,IL-6,ferritin,urine protein,urine leukocyte count,SLEDAI score,and severe disease activity were the influencing factors for SLE patients with infection.Fourthly,Logistic multi-factor regression analysis showed that urine protein,urine leukocyte count,severe disease activity,SLEDAI score,and ferritin level were independent risk factors for SLE co-infection(P<0.05),while elevated hemoglobin and albumin levels were independent protective factors for SLE co-infection(P<0.05).Finally,the nomogram was constructed with the above seven predictors and validated internally and externally.After the ROC curve,calibration curve,decision curve and Hosmer-Lemeshow test,the model was considered to have good discrimination,good fit and predictive performance,and clinical utility.Conclusion: Elevated urine protein levels,urine leukocyte count,severe disease activity,SLEDAI score,and ferritin levels were independent risk factors for SLE co-infection,while elevated hemoglobin and albumin levels were independent protective factors for SLE co-infection.The nomogram constructed with urine protein level,urine leukocyte count,disease severity activity,SLEDAI score,ferritin level,hemoglobin level and albumin level has high discrimination and calibration,good fit,clinical utility and good clinical value for predicting the risk of infection in SLE patients. |
PDF Full Text Request