摘要
目的探讨急性胆源性胰腺炎(ABP)发生的危险因素,构建早期预测模型。方法回顾2015年1月至2019年12月在安徽医科大学第二附属医院住院治疗的急性胰腺炎患者病例资料,最终纳入334例患者,随机分成建模组234例(其中胆源组96例,非胆源组138例),验证组100例。采集患者基本信息、临床资料及入院24 h内血液检查结果。采用单因素分析及二元logistic回归正向LR法筛出ABP的独立预测因素,联合各独立预测因素建立ABP加权预测模型(wScore)及非加权预测模型(unwScore),构制以上独立预测因素及预测模型的受试者工作特征曲线(ROC),并在验证组中对非加权预测模型加以验证,分析预测效能。结果建模组中,通过单因素及二元logistic回归正向LR法分析显示,性别、年龄、丙氨酸转氨酶(ALT)、天冬氨酸转氨酶(AST)、碱性磷酸酶(ALP)是ABP的独立预测因素。对性别、年龄、ALT、AST、ALP进行二元logistic回归分析,建立wScore预测模型,此模型预测急性胰腺炎的胆道病因AUC为0.945,灵敏度为91.67%,特异度为86.23%,阳性预测值82.24%,阴性预测值为93.70%。为便于临床使用,通过ROC曲线计算出上述预测因素的截断值,建立了unwScore评分模型,该模型预测ABP截断值为2分,ROC曲线下面积(AUC)为0.932(95%CI 0.891~0.960),灵敏度为75.00%,特异度为95.65%,阳性预测值92.31%,阴性预测值为84.61%,Hosmer-Lemeshow检验(P=0.176)。在验证组中,AUC为0.931(95%CI 0.871~0.969),Hosmer-Lemeshow检验(P=0.761),结果提示该模型对ABP有较好的预测价值。结论性别、年龄、ALT、AST、ALP是急性胆源性胰腺炎的独立预测因素;基于以上因素建立的加权预测模型的敏感性更高,非加权预测模型特异度高,2种模型均具有较好的临床预测效果。
Objective To explore the risk factors of acute biliary pancreatitis(ABP)and construct an early prediction model.Methods Reviewing the medical records of patients with acute pancreatitis who were hospitalized in the Second Affiliated Hospital of Anhui Medical University from January 2015 to December 2019,334 patients were finally included and randomly divided into modeling group 234 cases(including biliary group 96 cases,and 138 cases were in the non-biliary group),100 cases in the verification group.Collected basic patient information,clinical data and blood test results within24 hours of admission.Single factor analysis and binary logistic regression forward LR method were used to screen out the independent predictive factors of ABP,and the independent predictive factors were combined to establish the ABP weighted predictive model(wScore)and unweighted predictive model(unwScore).Constructed the receiver operating characteristic curve(ROC)of the above independent predictors and predictive models,and verified the unweighted predictive model in the verification group,and analyzed the predictive performance.Results In the modeling group,through single factor and binary logistic regression analysis with forward LR method,gender,age,alanine aminotransferase(ALT),aspartate aminotransferase(AST),alkaline phosphatase(ALP)were independent predictors of ABP.Binary Logistic regression analysis was performed on gender,age,ALT,AST,and ALP,and a w Score prediction model was established.This model predicted the biliary cause of acute pancreatitis with AUC of 0.945,sensitivity of 91.67%,specificity of 86.23%,and positive predictive value 82.24%,the negative predictive value was 93.70%.For the convenience of clinical use,the cut-off value of the above predictive factors was calculated through the ROC curve,and the unw Score scoring model was established.The model predicted the cut-off value of ABP is 2 points,and the area under the ROC curve(AUC)was 0.932(95%CI 0.891-0.960),the sensitivity was 75.00%,the specificity was 95.65%,th
作者
余旺
章礼久
宋莎莎
路燕
YU Wang;ZHANG Li-jiu;SONG Sha-sha;LU Yan(Department of Gastroenterology,the Second Affiliated Hospital of Anhui Medical University,Hefei 230601,China)
出处
《中国实用内科杂志》
CAS
CSCD
北大核心
2021年第7期613-618,共6页
Chinese Journal of Practical Internal Medicine
基金
安徽省自然科学基金(1708085QH192)。
