目的 探究经 T 管瘘道行肝内外胆管取出残留结石术后，影响患者出现胆道并发症的相关因素，并构建其预测模型。方法 回顾性分析 2018 年 1 月至2020 年 7 月期间于本院行经 T 管瘘道肝内外胆管取石术的患者 194 例，根据患者术后胆道并发症胆道出血、胆道感染、腹痛和呕吐等的发生情况，分为无并发症组（n = 132）与并发症组（n = 62），收集患者的相关资料，并以单因素和多因素logistic 回归分析影响患者发生相关并发症的独立影响因素，并以此构建风险预测模型。另选取 2020 年 8 月至2022 年 8 月 期间于本院行经 T 管瘘道肝内外胆管取石术的患者 92 例作为外部验证。以受试者工作特性曲线（ROC） 评估风险预测模型的预测价值，采用 Hosmer-Lemeshow 检验判断模型的拟合优度。结果 194 例接受经 T 管瘘道肝内外胆管取石术的患者中，有 132 例未出现并发症，有 62 例出现并发症，其中腹痛、呕吐、胆道感染 57 例，胆道出血 5 例。单因素分析结果显示，患者的术前白蛋白、手术时间、胆汁性状、术中出血量、胆汁多重耐药菌感染在两组患者之间差异有统计学意义（P < 0.05）；患者的年龄、性别、体质量指数、病程、 术前总胆红素、术前白细胞、合并高血压、合并糖尿病、合并乙型肝炎、手术医师经验、胆总管直径在两组 患者之间差异无统计学意义（P > 0.05）。多因素分析结果显示，患者的术前白蛋白、手术时间、胆汁性状、 术中出血量、胆汁多重耐药菌感染均为影响患者术后出现并发症的独立影响因素（P < 0.05）；风险预测模 型：Logit（P）=8.564-0.438×（术前白蛋白）+0.092×（手术时间）+2.038×（胆汁性状）+1.253×（术中出血量） -1.179×（多重耐药菌感染），风险预警模型 ROC 曲线下面积（AUC）为 0.890（95%CI：0.875 ~ 0.951），敏感度 为 89.8%，特异度为 78.2%，Hosmer-Lemeshow 拟合检验显示，χ² = 10.109，P = 0.57，模型拟合效度好，预测价值高。模型外部验证敏感度 85.2%，特异度 84.6%，总准确率 84.8%。结论 患者的术前白蛋白、手术时间、胆汁性状、术中出血量、胆汁多重耐药菌感染均为影响患者术后出现并发症的独立影响因素，以此构建风险预测模型具有一定的预测效能。 

Objective To explore the relevant factors affecting the occurrence of biliary complications in patients after removing residual stones from intrahepatic and extrahepatic bile ducts through T­tube fistula，and to construct a predictive model. Method A retrospective analysis was conducted on 194 patients who underwent intrahepatic and extrahepatic bile duct stone removal through T ­tube fistula in our hospital from January 2018 to July 2020. Based on the incidence of postoperative biliary complications such as biliary bleeding， abdominal pain， and vomiting， the patients were divided into a non­complication group （n = 132） and a complication group （n = 62）. Relevant data of the patients were collected， and independent influencing factors for the occurrence of related complications were analyzed using univariate and multivariate logistic regression analysis. Another 92 patients who underwent intrahepatic and extrahepatic bile duct stone removal through T­tube fistula in our hospital from August 2020 to August 2022 were selected as external validation. The predictive value of the risk prediction model was evaluated by the receiver operating characteristic curve （ROC）， and the Goodness of fit of the model was judged by the Hosmer Lemeshow test. Results Among the 194 patients who underwent intrahepatic and extrahepatic bile duct stone removal through T­tube fistula，132 did not experience complications， while 62 experienced complications， including 57 cases of abdominal pain， vomiting， biliary tract infection， and 5 cases of biliary bleeding. The results of univariate analysis showed significant differences between the two groups of patients in preoperative albumin， surgical time， bile properties， intraoperative bleeding volume， and bile multidrug­resistant bacterial infection （P<0.05）； No significant difference was noted in age，sex，BMI，course of disease，preoperative total bilirubin， preoperative white blood cells， hypertension， diabetes， hepatitis B，surgeon experience and bile duct diameter in the two groups （P > 0.05）. The results of multivariate analysis showed that preoperative albumin， surgical time， bile properties， intraoperative bleeding， and infection with multidrug­resistant bile bacteria were all independent influencing factors for postoperative complications in patients（P<0.05）；Risk prediction model：Logit （P）=8.564­0.438 ×（Preoperative albumin）+0.092 ×（Surgical time）+2.038 ×（Bile traits）+1.253 ×（intraoperative bleeding volume）1.179 ×（Multi drug resistant bacterial infection），the area under the ROC curve （AUC） of the risk warning model was 0.890（95% CI： 0.875 ~ 0.951），the sensitivity was 89.8%，and the specificity was78.2%. The Hosmer Lime show fitting test showed that，χ² = 10.109，P = 0.57，the model had good fitting validity and high predictive value. The sensitivity of external validation of the model was 85.2%， specificity was 84.6%， and overall accuracy was 84.8%. Conclusion Preoperative albumin， surgical time， bile properties， intraoperative bleeding， and infection with multidrug­resistant bile bacteria were independent influencing factors for postoperative complications in patients. Therefore，constructing a risk prediction model based on these factors has certain predictive power.