The Development and Validation of a Predictive Nomogram Model for Febrile Non-hemolytic Transfusion Reaction in Patients Receiving Red Blood Cells
Objective A nomogram prediction model for febrile non-hemolytic transfusion reaction(FNHTR)was constructed and validated through regression analysis of red blood cell transfusion patient data,aiming to provide the assistance in clinical prevention of FNHTR.Methods A retrospective analysis was conducted in 61 cases with FNHTR in our hospital from January,2021 to December,2023.A total of 189 cases without FNHTR during the same period were randomly selected as the control group,resulting in a total of 250 cases included in the modeling process.Additionally,a validation study cohort consisting of 91 patients who received suspended red blood cells from other hospitals during the same period and at the same level was developed for further external validation.The independent risk factors leading to FNHTR were identified among 19 influencing factors using single factor and multiple factor logistic regression models.The predictive nomogram model for FNHTR was established using Rstudio software,and its differentiation ability was evaluated by drawing a receiver operating curve(ROC)and calculating area under ROC curve(AUC).Calibration curves and clinical decision curves were utilized to assess model calibration and patient net benefits.Results According to logistic regression analysis,gender,Hb,WBC,ALT,blood transfusion reaction history,and NLR were identified as independent risk factors for FNHTR with respective OR values and 95％CI.These independent risk factors were then incorporated into R program to successfully fit a non-hemolytic fever reaction nomogram prediction model.Hosmer-Lemeshow goodness-of-fit test indicated that the model had a good fit(x2=5.762,P=0.674).AUC of the model was 0.803(95％CI 0.744-0.863),demonstrating a strong differentiation ability.Bootstrap sampling process was performed repeatedly for calibration purposes which showed that the predicted probability overlapped well with actual probability indicating a good accuracy.The clinical decision curve demonstrated a good net benefit for patients utilizing this model.External validation using additional dataset resulted in an AUC value of 0.784(95％CI 0.662-0.907).The calibration curve and clinical decision curve showed that the model also had a good calibration degree and net benefit with the external dataset.Conclusion The nomogram prediction model of FNHTR exhibits an excellent discriminatory capacity and accuracy,capable of indicating the occurrence of FNHTR and minimizing the risk of transfusion adverse reactions.
万方数据
维普
