Abstract: Background　Peptic ulcer bleeding （PUB） is a common cause of hospitalization in the gastroenterology department. The length of stay （LOS） is a key factor in assessing disease severity，diagnosis and treatment as well as management quality，yet the studies on LOS of PUB are rare. Objective　To investigate the risk factors of prolonged LOS in PUB patients，and based on this，to develop a predictive scoring model，providing a reference for bettering clinical intervention and management of PUB. Methods　A total of 485 discharged PUB inpatients were selected from Department of Gastroenterology，Wuxi People's Hospital from 2018 to 2019，and two thirds of them were randomly assigned to the training set，and the rest to the validation set. Prolonged LOS was defined as more than the 75th percentile of LOS for all participants （over eight days）. Multivariate Logistic regression analysis based on a backward elimination was carried out to determine the risk factors for prolonged LOS. Then the risk factors were scored according to the regression coefficient，and used to develop a predictive scoring model. The ROC with corresponding AUC of the predicitve scoring model for predicting the prolonged LOS in the training set and the validation set was analyzed，respectively，Results　Of the cases in the training set（n=323），75 had prolonged LOS，and 248 had normal LOS. Of the cases in the validation set（n=162），34 had prolonged LOS，and the other 128 had normal LOS. Multivariate Logistic regression analysis based on a backward elimination showed that age≥ 60 years〔OR=2.810，95%CI（1.567，5.040）〕，increased WBC and/or NE〔OR=2.491，95%CI（1.310，4.738）〕，ForrestⅠA-ⅡC peptic ulcers〔OR=1.960，95%CI（1.098，3.498）〕，blood transfusion 〔OR=3.367，95%CI（1.664，6.812）〕 and renal dysfunction 〔OR=4.809，95%CI（1.456，15.884）〕were associated with increased risk of prolonged LOS （P<0.05）. The AUC of the predictive scoring model for predicting the prolonged LOS for patients in the training set was 0.77〔95%CI（0.71，0.84）〕，and that for those in the validation set was 0.76〔95%CI（0.66，0.86）〕；the optimal cut-off value was 2.45 for both. Conclusion　Age≥60 years，increased WBC and/or NE，ForrestⅠA-ⅡC peptic ulcers，blood transfusion and renal dysfunction were independent risk factors for prolonged LOS of PUB. Based on these risk factors，an easy-to-use predictive scoring model with a good discriminative ability was developed，which has the potential to optimize the evaluation and management of PUB hospitalization.

Key words: Peptic ulcer hemorrhage, Length of stay, Risk factors, Prediction score model

摘要： 背景　消化性溃疡出血（PUB）是消化内科住院患者常见的病因之一，而住院时间在评估疾病严重程度、疾病诊疗和管理质量中起着重要作用，但PUB患者住院时间延长相关的研究报道较少见。目的　分析PUB患者住院时间延长的危险因素，并制定相应的预测评分模型，为临床干预及管理提供参考。方法　选取2018—2019年从无锡市人民医院消化内科出院的PUB住院患者485例，采用随机化方法进行分组，其中2/3的病例纳入训练集，1/3的病例纳入测试集。以住院时间的第75百分位数为分界点（>8 d），将全部患者分为超长组和正常组。采用回退法多因素Logistic回归分析探究PUB患者住院时间延长的危险因素，并依据回归系数对危险因素进行赋分，制定预测评分模型。分别绘制预测评分模型在训练集和测试集中预测PUB患者住院时间延长的受试者工作特征（ROC）曲线，并计算ROC曲线下面积（AUC）。结果　485例患者中训练集323例，其中超长组75例，正常组248例；测试集162例，其中超长组34例，正常组128例。回退法多因素Logistic回归分析结果显示，年龄≥60岁〔OR=2.810，95%CI（1.567，5.040）〕、白细胞计数（WBC）和/或中性粒细胞分数（NE）升高〔OR=2.491，95%CI（1.310，4.738）〕、Forrest Ⅲ级以上溃疡〔OR=1.960，95%CI（1.098，3.498）〕、输血〔OR=3.367，95%CI（1.664，6.812）〕、合并肾功能不全〔OR=4.809，95%CI（1.456，15.884）〕是PUB患者住院时间延长的独立危险因素（P<0.05）。预测评分模型在训练集中预测PUB患者住院时间延长的AUC为0.77〔95%CI（0.71，0.84）〕，在测试集中预测PUB患者住院时间延长的AUC为0.76〔95%CI（0.66，0.86）〕，最佳界值均为2.45分。结论　年龄≥60岁、WBC和/或NE升高、Forrest Ⅲ级以上溃疡、输血、合并肾功能不全是PUB患者住院时间延长的独立危险因素；本研究依据危险因素制定了一种具有较高辨别力的简单易行的预测评分模型，具有为优化PUB患者住院评估及管理提供参考的潜能。

关键词: 消化性溃疡出血, 住院时间, 危险因素, 预测评分模型