老年重症急性胰腺炎早期多指标联合预测模型的建立

Establishment of multiple predictor models of severe acute pancreatitis in elderly patients

目的 探讨老年急性胰腺炎严重程度相关的预后因素,建立老年重症急性胰腺炎（SAP）早期多指标联合预测模型.方法 回顾性分析哈尔滨医科大学附属第一医院胰胆外科2014年1月至2017年5月收治的老年急性胰腺炎并符合纳入标准的146例患者的临床资料,其中轻症急性胰腺炎88例,中重症急性胰腺炎29例,重症急性胰腺炎（SAP）29例.应用单因素分析及有序多分类Logistic回归分析筛选出老年急性胰腺炎严重程度相关的独立预后因素,联合各独立预后因素构建老年SAP非加权预测模型（unwScore）和加权预测模型（wScore）.绘制各独立预后因素及预测模型的受试者工作特征（ROC）曲线,确定截断值、计算ROC曲线下面积、灵敏度、特异度、阳性预测值及阴性预测值,观察各独立预后因素及预测模型的临床预测效果.结果 单因素及有序多分类Logistic回归分析结果显示,降钙素原（PCT）（Z=10.564,P=0.000）、血尿素氮（BUN）（Z=22.231,P=0.003）、血清肌酐（Scr）（Z=14.151,P=0.030）、血钙（Z=34.979,P=0.032）、胸腔积液（χ2=28.463,P=0.015）为老年急性胰腺炎严重程度的独立预后因素,ROC曲线下面积分别为0.908、0.737、0.701、0.753、0.712,灵敏度分别为0.828、0.621、0.552、0.690、0.517,特异度分别为0.915、0.786、0.846、0.966、0.906,阳性预测值分别为70.6％、41.9％、47.1％、83.3％、57.7％,阴性预测值分别为95.5％、89.3％、88.4％、92.6％、88.3％;unwScore和wScore的ROC曲线下面积分别为0.915和0.953,灵敏度分别为0.759和0.931,特异度分别为0.889和0.915,阳性预测值分别为62.9％和73.0％,阴性预测值分别为93.7％和98.2％.结论 PCT、BUN、Scr、血钙、胸腔积液为老年急性胰腺炎严重程度相关的独立预后因素,老年SAP多指标联合预测模型具有良好的临床预测效果,可为老年急性胰腺炎严重程度的早期预测及治疗提供临床参考.

Objective To investigate the prognostic factors related to the severity of acute pancreatitis and to establish the muhiple predictor models of severe acute pancreatitis （SAP） in elderly patients. Methods Clinical data of 146 consecutive elderly patients who met the inclusion criteria between January 2014 and May 2017 at Department of Pancreatic and Biliary Surgery,the First Affiliated Hospital of Harbin Medical University were retrospectively collected and analyzed, wherein 88 cases were mild acute pancreatitis,29 cases were moderately severe acute pancreatitis and 29 cases were SAP. The patients data were subjected to univariate analysis and multiple classified Logistic regression analysis for independent prognostic factors of the severity of acute pancreatitis in elderly patients. Unweighted predictive score （unwScore） and weighted predictive score（wScore）for SAP in elderly patients were established according which the receiver-operating characteristic （ ROC ） curves of independent prognostic factors and predictor models were produced. The cutoff values of independeut prognostic factors and predictor models were determined. The area under the curve, the sensitivity, the specificity, the positive predictive value and the negative predictive value to verify the predictive efficiency of the independent prognostic factors and predictor models were calculated. Results Procalcitonin （PCT） （ Z = 10. 564, P = 0. 000）, blood urea nitrogen （ BUN ） （Z= 22. 231 ,P=0. 003） ,serum creatinine（Scr） （Z= 14. 151 ,P=0. 030） ,serum calcium（Z= 34. 979,P= O. 032） and pleural effusion（χ2 =28. 463,P=0. 015） were independent prognostic factors of the severity of acute pancreatitis by univariate analysis and multiple classified Logistic regression analysis in elderly patients. Respectively, the area under the curve of PCT, BUN,Scr,serum calcium and pleural effusion were 0. 908,0. 737,0. 701,0. 753,0. 712, the sensitivity were 0. 828,0. 621,0. 552,0. 690,0. 517,the specificity were 0. 915, 0. 786, 0. 846, 0. 966, 0. 906, the positive predictive value were 70. 6%, 41.9%, 47. 1%, 83.3% ,57.7%, the negative predictive value were 95.5% ,89. 3% ,88. 4% ,92. 6% ,88.3%. Respectively, the area under the curve of unwScore and wScore were 0. 915 and 0. 953, the sensitivity were 0. 759 and 0. 931, the specificity were 0. 889 and 0. 915, the positive predictive value were 62. 9% and 73.0%, the negative predictive value were 93.7% and 98.2%. Conclusions PCT, BUN, Scr, serum calcium and pleural effusion were independent prognostic factors of the severity of acute pancreatitis in elderly patients. The multiple predictor models of SAP in elderly patients have a good predictive efficiency, which may provide valuable clinical reference for prediction and treatment.