基于2020至2022年中国“世界营养日”调查数据的住院患者住院时间延长风险因素分析及预测模型构建  

Analysis of risk factors for prolonged hospital stay and construction of prediction model based on data from nutritionDay worldwide 2020 to 2022 in China

作　　者：周品闻 张玉鹏 章黎 王新颖 中华医学会肠外肠内营养学分会“世界营养日”协作组 Zhou Pinwen;Zhang Yupeng;Zhang Li;Wang Xinyin;nutritionDay Worldwide Collaboration Group of Chinese Society of Parenteral and Enteral Nutrition(Department of General Surgery,Nanjing Jinling Hospital,Affiliated Hospital of Medical School,Nanjing University,Nanjing 210002,China;不详)

机构地区：[1]南京大学医学院附属金陵医院普通外科,南京210002 [2]不详

出　　处：《中华临床营养杂志》2025年第1期16-24,共9页Chinese Journal of Clinical Nutrition

基　　金：江苏省重点研发计划(BE2022822);国家自然科学基金(82370900,82170575)。

摘　　要：目的基于中国“世界营养日”调查数据,分析住院患者住院时间延长的风险因素,构建预测模型并进行验证,为临床决策提供依据。方法本研究为回顾性研究,数据来源为中国“世界营养日”多中心住院患者营养状况调查研究数据库。选取2020至2022年登记的年龄≥18岁、30天预后问卷有效、临床资料完整的2335例病例为研究对象,收集研究对象的人口统计学特征、营养相关指标、疾病信息和结局指标,以住院时间第75百分位数作为分界点,将患者分为住院时间延长组(570例)和住院时间正常组(1765例),结合最小绝对值收敛和选择算子回归算法和多因素Logistic回归分析结果并构建列线图预测模型。采用曲线下面积(area under the curve,AUC)、校正曲线、Hosmer-Lemeshow检验和临床决策曲线,验证模型区分度、拟合度和临床有效性。结果最终纳入体重指数(body mass index,BMI)、住院期间是否手术、住院期间是否入住ICU、NRS 2002评分、是否能独立行走、既往住院次数和过去3个月丢失体重7个风险因素,并构建列线图预测模型。训练队列AUC为0.783(95%CI:0.759~0.807),验证队列AUC为0.797(95%CI:0.746~0.849)。校准曲线及Hosmer-Lemeshow检验(训练队列P=0.735,验证队列P=0.431)显示模型拟合度较好,临床决策曲线显示列线图具有良好的临床应用价值。结论BMI、住院期间是否手术、住院期间是否入住ICU、NRS 2002评分、是否能独立行走、既往住院次数和过去3个月丢失体重是中国住院患者住院时间延长的风险因素。据此构建的列线图预测模型能够预测中国住院患者住院时间延长的风险,为住院时间延长的早期识别和干预提供依据。Objective To analyze the risk factors for prolonged hospital stay in inpatients based on data from nutritionDay worldwide survey 2020 to 2022 conducted in China and to construct and validate a prediction model for clinical decision-making.MethodsThis study was a retrospective study,the data source was the China's multi-centered nutritionalDay worldwide database for nutrition status in inpatients.A total of 2335 cases registered in the database from 2020 to 2022 were selected for the study,comprising individuals aged 18 and above,with valid response for the 30-day prognosis questionnaires,and with complete clinical data.The demographic characteristics,nutrition-related indicators,disease information,and outcome indicators of the participants were collected.Based on the 75th percentile of hospitalization duration,the participants were divided into the prolonged length of stay group(570 cases)and the normal length of stay group(1765 cases).A nomogram prediction model was constructed using the least Absolute Shrinkage and Selection Operator(LASSO)regression and multivariate Logistic regression analysis.Area under the curve(AUC),calibration curve,Hosmer-Lemeshow test,and clinical decision curve were used to verify discriminative ability,goodness-of-fit,and clinical effectiveness of the model.ResultsSeven independent risk factors for prolonged length of stay were identified,namely body mass index(BMI),whether surgery occurred during hospitalization,intensive care unit admission during hospitalization,Nutrition Risk Screening 2002 score,ambulatory independence,previous hospitalization frequency,and weight loss in the past 3 months.A nomogram prediction model was established accordingly.The AUC of training set was 0.783(95%CI:0.759-0.807),and the AUC of validation set was 0.797(95%CI:0.746-0.849).Calibration curves and Hosmer-Lemeshow tests(P=0.735 for training set,P=0.431 for validation set)indicated good model fitting.The clinical decision curve demonstrated the favorable clinical utility of the nomogram.ConclusionsBMI

关 键 词：住院时间 风险因素 列线图 中国住院患者 世界营养日 

分 类 号：R459.3[医药卫生—治疗学]