急性胰腺炎并发急性肝衰竭危险因素的列线图风险预测模型构建与评估

Construction and evaluation of a nomogram prediction model of risk factors for acute pancreatitis combined with acute liver failure

目的分析影响急性胰腺炎(AP)患者出现急性肝衰竭(ALF)的相关危险因素,并构建其列线图预测模型。方法将397例急性胰腺炎病人作为研究对象,根据是否合并ALF,分为ALF组和非ALF组,通过单、多因素logistic模型分析其合并ALF的影响因素,将回归分析中具有统计学意义(P<0.05)的变量作为预测因子。利用R语言,构建BMI正常AP合并ALF的列线图和风险预测模型。结果397例AP患者中有42例发生ALF(10.58%),。Logistic多因素回归分析结果显示,BISAP评分、NLR值、谷氨酰转肽酶等是AP合并ALF的独立危险因素。ROC曲线分析结果显示,该模型预测AP合并ALF的曲线下面积为0.959(95%CI:0.937~0.980),灵敏度为97.6%,特异度为82.7%。Bootstrap方法重复抽样1000次验证列线图,显示该列线图预测模型内部验证前后的平均绝对误差为0.034。并且列线图模型的临床净收益显著高于任何单个指标预测结果。结论BISAP评分、NLR值、谷氨酰转肽酶是AP合并ALF的独立危险因素,基于以上因素构建的列线图模型对AP合并ALF发生的早期风险预测准确性较高,可方便临床医生对患者进行早期评估,制定治疗方案提供参考价值。

Objective Analyze the risk factors related to acute liver failure(ALF)in patients with acute pancreatitis(AP)and build their nomogram prediction model.397 patients with acute pancreatitis are taken as the research object.Methods According to whether they combine ALF,they are divided into ALF groups and non-ALF groups.The influencing factors of their combined ALF are analyzed through single and multi-factor logistic models,and the variables of statistical significance(P<0.05)in regression analy-sis are taken as pre-Measure the factor.Use R language to build a line chart and risk prediction model of BMI normal AP combined with ALF.Results ALF occurred in 42 of 397 AP patients(10.58%).Logistic regression analysis showed that BISAP score,NLR value and glutamyl transferase were independent risk factors of AP combined with ALF.ROC curve analysis showed that the area under the curve for predicting AP combined with ALF was 0.959(95%CI:0.937~0.980),the sensitivity was 97.6%,and the specificity was 82.7%.Bootstrap method repeated sampling 1000 validation nomograms,showing that the nomogram prediction model before and after internal validation mean absolute error of 0.034.And the net clinical benefit of nomogram model was significantly higher than that predicted by any single index.Conclusion BISAP score,NLR value and Glutamyl transpeptidase value are independent risk factors for AP combined with ALF.The line chart model based on the above factors has a high accuracy in predicting the early risk of AP with ALF,which can facilitate clinicians to conduct early evaluation of patients and provide reference value for the formulation of treatment plans.