Development And Validation Of A Nomogram Model For Predicting Delayed Hemorrhage After Pancreaticoduodenectomy

Objective: To develop a diagnostic predictive nomogram model for delayed hemorrhage after pancreaticoduodenectomy,we analyze the independent risk factors for delayed hemorrhage after pancreaticoduodenectomy,which provides a convenient and reliable tool for individualized prediction of delayed bleeding after PD,aiming to identify high-risk patients early and reduce the incidence of delayed bleeding after PD surgery through active and effective interventions.Methods: This study uses the retrospective cohort study method.According to the inclusion and exclusion criteria,we collected the clinical data of patients who underwent standard pancreaticoduodenectomy in the General Hospital of Ningxia Medical University hospital from June 2012 to June 2022.The patients were divided into a postoperative delayed hemorrhage group(more than 24 hours)and no bleeding group according to the time of bleeding after PD surgery.Through studying the literature and combining with the actual situation of the medical center,some relevant observation indicators were selected as potential risk factors.Independent risk factors for delayed hemorrhage after PD were found by univariate logistic regression(P<0.05)and binary multivariate logistic regression analysis,and a diagnostic prediction model for delayed hemorrhage after PD was constructed.The predictive ability and clinical practicability of the model were objectively evaluated through the degree of discrimination,calibration,and clinical decision curve.The verification of the predictive model is carried out using the internal bootstrap repeat sampling method(Bootstrap1000 times).Finally,the visualization of the predictive model is displayed in the form of a nomogram.Results: 1.According to the inclusion and exclusion criteria,476 patients with PD surgery were finally included,including 63(13.24%)patients with delayed bleeding after PD surgery.2.By univariate and multivariate logistic regression analysis,male(OR=3.49,95% CI:1.68-7.24),preoperative total bilirubin > 171 μ mol/L(OR=2.72,95% CI: 1.45-5.12),and albumin on the first postoperative day were found <30 g/L(OR=3.83,95% CI: 1.72-8.53),Clinically relevant postoperative pancreatic fistula(OR=4.52,95% CI: 2.33-8.78),biliary fistula(OR=3.03,95% CI: 1.24-7.42),and intra-abdominal infection(OR=3.92,95% CI:2.03-7.56)were independent risk factors for delayed hemorrhage after PD.3.The six independent risk factors were selected to develop a predictive model for delayed hemorrhage after PD surgery.The Area Under the ROC Curve(AUC)of the predictive model in the whole cohort is 0.857(95% CI: 0.810-0.905),which indicates that the predictive model has good discrimination ability.The optimal cut-off value of the predictive model is 0.162,the sensitivity is 74.6%,and the specificity is 81.4%.The calibration curve of the prediction model shows that the actual prediction curve is close to the ideal curve,and the P value of the Hosmer-Lemeshow goodness-of-fit test is 0.985,which can be considered to be well calibrated.The clinical decision curve results of the model suggest that the net benefit rate of patients is > 0 within most of the probability thresholds,suggesting that the model has certain clinical application value.After internal validation,the recalculated AUC is 0.844,and it is considered that the model still has good discriminative ability.Through the calibration curve after resampling,it is found that it is more consistent with the ideal curve,indicating that the calibration degree of the model is better.4.This predictive model has good discrimination,calibration,and clinical practicability,so the model is displayed in the form of the nomogram.Conclusion: The independent risk factors for delayed hemorrhage after pancreaticoduodenectomy were male,preoperative total bilirubin >171 μ mol/L,albumin < 30g/L on the first day after surgery,Clinically Relevant Postoperative Pancreatic Fistula(CR-POPF),biliary fistula,and abdominal infection.Based on the independent factors,a prediction model was constructed,which has good predictive ability and clinical practicability after internal validation.The model can predict the occurrence of delayed hemorrhage after PD surgery,which has good clinical application value.