肿瘤医院肝胆外科住院病人营养现状调查及模型预测

Survey of inpatients'nutrition and model prediction in hepatobiliary surgery of cancer hospital

目的:为更好地了解本院肝胆外科肿瘤病人的营养现状和营养管理情况,以及针对恶性肿瘤住院病人实施的营养评估及干预情况,开展营养日问卷调查,建立预测模型,为临床合理决策提供理论实践基础。方法:根据世界营养日调查要求并取得本院伦理委员会审批后,对本院2016至2020年在营养日当日肝胆外科住院病人进行问卷调查,调查内容除营养项目通用问卷外还包括针对肿瘤病人的专科问卷。通过R语言软件建立机器学习模型对营养风险进行预测。结果:本研究纳入男性病人103人,女性病人61人;问卷回访率97.6%;其平均年龄为(57.3±11.8)岁;平均BMI为(24.3±3.3)kg/m^(2)。约25%病人MST评分大于2分,存在有营养风险。主要病人治疗目标为“积极治疗”,肿瘤分期为“局部浸润”和“局部扩散”,营养支持方案为“无特殊配餐”,有一定比例的肠内营养和肠外营养,5年来肠内营养方式有所加强。在30日随访中,73.2%的结果为“出院回家”。支持向量机模型预测结果远好于Logistic回归和判别分析模型。结论:对本院肿瘤病人的营养干预还未有效实施。医生、护士、药师和营养师营养相关理念和知识需要加强培训,对病人合理膳食和营养教育需要加强指导;对数据预测可选择支持向量机法,有助于提高肿瘤治疗效果,改善病人生活质量。

Objective:In order to better understand nutritional status and management on hepatobiliary surgery patients in our hospital,as well as the nutritional evaluation and intervention for inpatients with malignant tumors,a nutrition day questionnaire survey and a prediction model was carried out,to provide a theoretical and practical basis for clinical rational decision-making.Methods:According to the nutritionDay Worldwide requirements and the approval of the ethics committee of our hospital,a questionnaire survey was conducted on the inpatients of hepatobiliary surgery department in our hospital on the nutrition days between 2016 and 2020.In addition to the general nutrition questionnaire items,the survey also included the special questionnaire for tumor patients.Machine learning models were used to predict nutrition risk using R language software.Results:This study included 103 male patients and 61 female patients with a return visit rate of 97.6%,an average age of(57.3±11.8)years,and an average BMI of(24.3±3.3)kg/m^(2).About 25%patients with MST score greater than 2 are at nutritional risk.The main treatment goals of patients are"Curative",tumor stage is"Localized",nutrition treatment is"No special diet".There was a certain proportion of enteral nutrition and parenteral nutrition.Enteral nutrition has been strengthened in the past five years.The proportion of patients who were"Discharged home"on the 30th day follow-up was 73.2%.Predictive evaluation from support vector machine(SVM)was far better than that from Logistic regression and discriminant analysis models.Conclusions:The nutrition intervention for tumor patient in our hospital has not been effectively implemented.The trainings for doctors and nurses on nutritional concepts and knowledge,as well as the guidance of patients'reasonable diet and nutrition education need to be strengthened.SVM could be selected for prediction,which is helpful to improve tumor treatment and patients’quality of life.