探讨肝内胆管结石肝切除术后胆漏的相关危险因素，建立预测模型并验证。

回顾性分析2019年1月至2024年12月绵阳市中心医院收治的244例肝内胆管结石肝切除术患者临床资料。其中男162例，女82例，年龄25~84岁，中位年龄60岁。本研究已通过绵阳市中心医院伦理委员会审批，为非干预性回顾性研究，已免除知情同意。根据术后是否发生胆漏分为胆漏组和非胆漏组。将患者围手术期临床指标进行Logistic单因素和多因素回归分析，构建胆漏预测模型。采用ROC AUC评估模型预测效能。采用Hosmer-Lemeshow检验和校准曲线评估模型的准确性。绘制决策曲线分析（DCA）用于评估模型的临床实用价值。

30例患者术后发生胆漏，发生率为12.3%（30/244），其中A级胆漏占60.0%（18/30），B级占33.3%（10/30），C级占6.7%（2/30）。Logistic多因素回归分析显示，既往上腹部手术史、肝硬化、肝门阻断时间、非左外叶切除及非解剖性肝切除为术后胆漏的独立危险因素（OR=3.983，3.661，1.070，5.390，3.724；P<0.05）。根据上述5个因素构建预测模型，公式为：Logit(P)=-7.597+1.382×既往上腹部手术史+1.298×肝硬化+0.068×肝门阻断时间+1.684×非左外叶切除+1.315×非解剖性肝切除。预测模型ROC分析显示，该模型AUC为0.862，C-index=0.862，模型诊断的灵敏度为0.808，特异度为0.833，模型预测结果稳定可靠。校准曲线和Hosmer-Lemeshow检验显示该预测模型具备较好的准确性（χ²=2.662，P=0.954）。运用Bootstrap自助抽样法对1 000例样本进行内部验证，结果显示模型仍保持良好的区分度。DCA显示，当阈值概率处于0.03~0.81时，应用该模型对患者进行评估可获得净收益，表明该模型具有一定的临床应用价值。

既往上腹部手术史、肝硬化、肝门阻断时间、非左外叶切除及非解剖性肝切除为肝内胆管结石肝切除术后胆漏独立危险因素。基于这些因素构建的预测模型具有良好的区分度与校准度，可有效评估患者术后胆漏风险。

To investigate the risk factors associated with bile leakage after hepatectomy for intrahepatic bile duct stones, and to develop and validate a prediction model.

The clinical data of 244 patients who underwent hepatectomy for intrahepatic bile duct stones at Mianyang Central Hospital from January 2019 to December 2024 were retrospectively analyzed. There were 162 males and 82 females, aged 25-84 years, with a median age of 60 years. This study was approved by the Ethics Committee of Mianyang Central Hospital. As a non-interventional retrospective study, the requirement for informed consent was waived. Patients were divided into a bile leakage group and a non-bile leakage group according to whether bile leakage occurred after surgery. Perioperative clinical indicators were analyzed using univariate and multivariate logistic regression, and a prediction model for bile leakage was constructed. The predictive performance of the model was evaluated using the receiver operating characteristic curve and area under the curve (ROC AUC). The accuracy of the model was assessed using the Hosmer-Lemeshow test and calibration curve. Decision curve analysis (DCA) was performed to evaluate the clinical utility of the model.

Postoperative bile leakage occurred in 30 patients, with an incidence of 12.3% (30/244). Among them, grade A bile leakage accounted for 60.0% (18/30), grade B for 33.3% (10/30), and grade C for 6.7% (2/30). Multivariate logistic regression analysis showed that a history of previous upper abdominal surgery, liver cirrhosis, hepatic inflow occlusion time, non-left lateral sectionectomy, and non-anatomical hepatectomy were independent risk factors for postoperative bile leakage (OR = 3.983, 3.661, 1.070, 5.390, and 3.724, respectively; P < 0.05). A prediction model was established based on these five factors, with the following formula: Logit (P) =-7.597 + 1.382×history of previous upper abdominal surgery + 1.298×liver cirrhosis + 0.068×hepatic inflow occlusion time+1.684×non-left lateral sectionectomy + 1.315×non-anatomical hepatectomy. ROC analysis of the prediction model showed that the model had an AUC of 0.862 and a C-index of 0.862. The diagnostic sensitivity and specificity of the model were 0.808 and 0.833, respectively, indicating that its predictive results were stable and reliable. The calibration curve and Hosmer-Lemeshow test showed good accuracy of the prediction model (χ²=2.662, P = 0.954). Internal validation was performed using the bootstrap resampling method with 1 000 samples, and the results showed that the model maintained good discrimination. DCA showed that when the threshold probability ranged from 0.03 to 0.81, applying this model to patient assessment yielded a net benefit, indicating that the model has certain clinical application value.

A history of previous upper abdominal surgery, liver cirrhosis, hepatic inflow occlusion time, non-left lateral sectionectomy, and non-anatomical hepatectomy are independent risk factors for bile leakage after hepatectomy for intrahepatic bile duct stones. The prediction model constructed based on these factors has good discrimination and calibration, and can effectively assess the risk of postoperative bile leakage.