住院患者黄疸鉴别诊断数学模型的建立

Computer-aided diagnosis of jaundice

目的通过应用统计学方法建立黄疸鉴别诊断模型。方法回顾性分析北京大学第一医院2001年1月至2010年12月病因明确的黄疸病例,应用Logistic回归的方法得到判断模型公式。结果对肝外梗阻性黄疸和肝内淤胆肝细胞性黄疸判断公式为ln[P/(1-P)]=-1.452-1.774乏力+1.824腹痛+1.763大便颜色变白+1.039皮肤瘙痒-0.003谷草转氨酶(ALT)+0.005碱性磷酸酶(ALP)+0.002γ-谷氨酰转酞酶(GGT)+0.005总胆红素(TBIL)+0.003CA199,其阳性预测值为92.1%,阴性预测值为89.0%,总预测准确率为90.7%,特异性88.9%,敏感性91.6%。对于肝外梗阻性黄疸良恶性的判断模型为ln[P/(1-P)]=-1.345-0.931腹痛+1.284发热+0.001GGT+0.005TBIL,其阳性预测值为82.9%,阴性预测值为59.3%,总预测准确率为80.4%,特异性为76.6%,敏感性为68.0%。结论本研究建立的黄疸鉴别模型对于区分肝外梗阻性黄疸和肝内淤胆肝细胞性黄疸具有较好的敏感性和特异性,对于区分恶性病变具有较好的预测准确率。

Objective To build a diagnosis model of jaundice using statistics methods. Methods The main causes of jaundice cases were analyzed retrospectively and differential diagnosis model was built by using Logistic regression. Results We worked out the formula： In [ P/（ 1 - P） ] = - 1. 452 - 1. 774 fatigue ＋ 1. 824 abdominal pain ＋ 1. 763 pale stool ＋ 1. 039 pruritus - 0. 003 ALT ＋ 0. 005 ALP ＋ 0. 002 GGT ＋ 0. 005 TBIL ＋ 0. 003 CA199,by which,jaundice can be classified obstructive from eholestasis or hepatoeellular. The positive predictive value was 92. 1%, and the negative predictive value was 89. 0%, in which, the accuracy was 90.7%, and the specificity and the sensitivity were 88.9% and 91.6%, respectively. And also the obstructive jaundice can be classified into benign or malignant by formula：In[ P/（ 1 - P） ] = - 1. 345 -0. 931 abdominal pain ＋ 1. 284 fever ＋ 0. 001 GGT ＋ 0. 005 TBIL. The positive predictive value was 82. 9%, while the negative predictive value was 59. 3%. The specificity and the sensitivity were 76. 6% and 68.0%, respectively. Conclusions The logistic fbrmulas had both satisfied with specificity and sensitivity and might be helpful to differential diagnosis of jaundice.