摘要：目的 构建可早期预测穿孔性阑尾炎患儿住院时间延长风险的列线图模型。方法 本研究为病例对照研究，回顾性收集2019年1月至2022年12月昆明市儿童医院普通外科收治、行腹腔镜阑尾切除术的穿孔性阑尾炎患儿临床资料，共818例；按7 ∶ 3比例随机划分为建模组575例、验证组243例。以本组患儿住院时间第75百分位数(≥11 d)为截断值，将病例分为住院时间正常组、住院时间延长组。单因素分析两组临床、手术相关指标差异，将有统计学意义变量纳入多因素Logistic向前逐步回归，筛选住院时间延长独立危险因素；基于回归结果采用R 3.6.3构建列线图预测模型。通过ROC曲线评价模型区分效能，Bootstrap重复抽样法完成内部验证，采用验证组数据集开展外部验证。结果 建模组单因素分析显示：发病时间、术前B超提示腹腔积液、术前C反应蛋白(C-reactive protein，CRP)、术前抗感染治疗、术中腹腔冲洗、术中留置腹腔引流管、术中出血量、手术时间在两组间差异具有统计学意义(P＜0.05)；多因素Logistic回归剔除无统计学意义变量(术前抗感染治疗、术中留置引流管、术中出血量，P＞0.05)，最终确定发病时间、术前B超腹腔积液、术前CRP、术中腹腔冲洗、手术时间为住院时间延长独立影响因素(P＜0.05)。建模组ROC曲线下面积=0.915，95%CI：0.889~0.942；验证组AUC=0.926，95%CI：0.892~0.961；模型区分度良好。Bootstrap内部验证平均绝对误差0.021，预测校准曲线与理想曲线贴合度高，模型预测准确度可靠。结论 发病时间、术前腹腔积液、术前CRP水平、术中腹腔冲洗、手术时长与穿孔性阑尾炎患儿住院时间延长独立相关；基于上述指标构建的列线图模型预测效能、校准度优异，可于入院早期识别住院延长高危患儿，对临床针对性干预有指导意义。

Abstract: Objective To establish an early nomogram model for predicting prolonged length of stay in children with perforated appendicitis (PA). Methods The relevant clinical data were retrospectively reviewed for 818 PA children undergoing laparoscopic appendectomy at the Department of General Surgery, Kunming Children's Hospital from January 2019 to December 2022.All patients were randomized at a ratio of 7∶3 into a derivation cohort (n=575) and a validation cohort (n=243).The 75th percentile of hospitalization duration (≥11 days) was set as a cut-off value for assigning them into two sub-groups of normal hospitalization and prolonged length of stay.Univariate analysis was performed for comparing the clinical and surgical parameters between two subgroups.Parameters with statistical significance were included in multivariate forward stepwise Logistic regression to screen independent risk factors for prolonged length of stay.R 3.6.3 software was utilized for constructing a nomogram based upon regression outputs.Receiver operating characteristic (ROC) curve was plotted for assessing discrimination capacity; Bootstrap resampling for internal validation and external validation conducted with the independent validation cohort. Results Univariate analysis of derivation cohort demonstrated significant intergroup differences (P < 0.05) in time from symptom onset to admission, preoperative ultrasonographic peritoneal effusion, preoperative C-reactive protein (CRP), preoperative antibiotic therapy, intraoperative peritoneal lavage, intraoperative abdominal drainage placing, intraoperative volume of blood loss and operative duration.Multivariate Logistic regression eliminated three non-significant confounding factors of preoperative antibiotic therapy, intraoperative drainage tubing and intraoperative volume of blood loss (P>0.05).Five independent risk factors were identified as time from symptom onset, preoperative peritoneal effusion on ultrasound, preoperative CRP, intraoperative peritoneal lavage and operative duration (P < 0.05).The area under ROC curve (AUC) of derivation cohort was 0.915(95%CI: 0.889-0.942) while AUC of validation cohort reached 0.926(95%CI: 0.892-0.961), indicating satisfactory discrimination performance.Internal Bootstrap validation presented a mean absolute error of 0.021.A favorable fitting between predicted calibration curve and ideal curve verified reliable predictive accuracy of the model. Conclusions Time from symptom onset, preoperative peritoneal effusion, preoperative CRP, intraoperative peritoneal lavage and operative duration are independently correlated with prolonged length of stay among PA children.The nomogram constructed by these routinely available parameters exhibits excellent discrimination and calibration.Thus it enables early identification of high-risk children and facilitates targeted clinical interventions.