A Nomogram Model for Individualized Prediction of the Risk of Respiratory Tract Infection within Six Months after Diagnosis in Patients with Primary Immune Thrombocytopenia

Abstract

The risk factors of upper respiratory tract infection (URI) within 6 months after diagnosis in patients with idiopathic thrombocytopenic purpura (ITP) were analyzed, and the nomogram model was established and verified, with 242 and 50 ITP patients as the training and validation set, respectively. The patients were followed up for six months after the diagnosis of ITP. The clinical data of patients were collected, and the risk factors of URI in ITP patients within six months after diagnosis were analyzed using univariable, followed by multivariable logistic regression. Among the 242 ITP patients in the training set, 52 cases (21.49%) had URI, including 24 cases of viral infection, 11 cases of Mycoplasma pneumoniae infection, and 17 cases of bacterial infection. Logistic regression analysis showed that advanced age, use of glucocorticoid, smoking history, platelet count, serum CRP level, and lymphocyte subsets CD4+ and CD8+ were all risk factors for ITP patients to develop symptoms within six months after diagnosis ( $P<0.05$ ). Using the above five indicators, a nomogram prediction model was built for URI occurrence in patients with ITP within half a year after diagnosis, and the results showed an AUC, a sensitivity, and a specificity of 0.936 (95% CI: 0.878-0.983), 0.942, and 0.865, respectively. The nomogram model was internally verified by the bootstrap method for 500 self-sampling times, and the prediction of the calibration curve was in high consistency with the real results. External validation of the nomogram model resulted in an AUC, a sensitivity, and a specificity of 0.890 (95% CI: 0.757-0.975), 0.949, and 0.727, respectively. The nomogram model of URI in ITP patients within half a year after diagnosis based on logistic regression analysis has good discrimination and prediction accuracy. This provides important guidance value for individualized prediction of URI in ITP patients.