【摘要】 背景：胃癌患者普遍存在营养风险，但目前尚缺乏精准的营养风险筛查工具。目的：构建胃癌住院患者的营养风险预测模型并进行验证。方法：纳入2023年8月—2024年7月秦皇岛市第一医院收治的295例胃癌住院患者，按照7∶3的比例分为模型组（n=206）和验证组（n=89）。采用营养风险筛查2002(NRS 2002)评分对模型组患者进行营养筛查。以单因素分析和多因素Logistic回归分析筛选胃癌患者发生营养风险的独立危险因素，并构建预测模型。采用Hosmer-Lemeshow拟合优度检验评价预测模型与NRS 2002评分之间的一致性。采用ROC曲线评价该模型的预测效能并进行验证。根据ROC曲线的阈值进行风险分层，评估模型组、验证组中不同亚组患者的营养风险情况。结果：多因素Logistic回归分析结果显示，年龄≥68岁、体质指数（BMI）≤22 kg/m~2、厌食、吞咽困难和贫血是模型组胃癌患者发生营养风险的独立危险因素（P均<0.05）。预测模型与NRS 2002评分之间的一致性良好（P=0.567）。ROC曲线分析显示，阈值为9.5分时，预测模型预测模型组和验证组胃癌患者发生营养风险的曲线下面积（AUC）分别为0.840(95%CI:0.785～0.895,P<0.001)和0.895(95%CI:0.831～0.958,P<0.001)。与低危患者相比，模型组和验证组高危患者发生营养风险的比例均明显升高（P均<0.001）。结论：基于年龄、BMI、厌食、吞咽困难和贫血构建的营养风险预测模型能够有效识别胃癌住院患者的营养风险，可作为临床参考工具。更多还原

【Abstract】 Background: Nutritional risk is highly prevalent among gastric cancer patients, yet accurate screening tools remain lacking. Aims: To construct and validate a nutritional risk prediction model for gastric cancer inpatients. Methods: From August 2023 to July 2024, a total of 295 gastric cancer inpatients admitted to the First Hospital of Qinhuangdao were enrolled, and divided into the model group (n=206) and validation group (n=89) at a ratio of 7∶3. Nutritional screening was performed in the model group using the nutritional risk screening 2002 (NRS 2002) score. Univariate analysis and multivariate Logistic regression analysis were employed to identify the independent risk factors for nutritional risk, and a prediction model was constructed. The Hosmer-Lemeshow goodness-of-fit test was used to evaluate the consistency between the model constructed and NRS 2002 score. The predictive performance was assessed using ROC curve and validated in the independent cohort. Risk stratification was performed based on the cut-off value derived from the ROC curve, and the nutritional risk in different subgroups of the model group and validation group was assessed. Results: Multivariate Logistic regression analysis indicated that age ≥68 years, body mass index (BMI)≤22 kg/m~2, anorexia, dysphagia and anemia were the independent risk factors for nutritional risk in gastric cancer patients in the model group (all P<0.05). The prediction model had good consistency with the NRS 2002 score (P=0.567). ROC curve showed that the area under the curve (AUC) of the prediction model for predicting nutritional risk were 0.840 (95% CI: 0.785-0.895, P<0.001) for the model group and 0.895 (95% CI: 0.831-0.958, P<0.001) for the validation group when the cut-off value was set at 9.5 points. The proportion of patients with nutritional risk in high-risk subgroup was significantly higher than that in low-risk subgroup in both model group and validation group (P<0.001). Conclusions: The prediction model based on age, BMI, anorexia, dysphagia and anemia can effectively identify nutritional risk in gastric cancer inpatients and may serve as a clinical reference tool.更多还原