Abstract:

Objective To explore the factors influencing enteral nutrition intolerance in critically ill patients and to develop a risk prediction model to provide medical staff with a tool for early identification of patient intolerance. Methods Domestic and international databases such as CNKI，PubMed and Web of Science were searched and supplemented by searching references and grey literature. The search period was from inception to November 2022. Data were independently screened and extracted by 2 systematically trained researchers，and the quality of the literature was evaluated. Meta-analysis was performed using Review Manager 5.4 software. The OR value of the comprehensive effect of the factors was taken as the coefficient of each factor in the formula，and the natural logarithm of the ratio of intolerance incidence and non-incidence was the constant term of the formula. From December 2022 to June 2023，360 patients who met the inclusion and exclusion criteria in the ICU of a tertiary hospital were collected as a model verification group by convenient sampling method，and the collected clinical data were substituted into the formula to evaluate the discrimination and calibration of the model. Results A total of 13 articles were included. 7 influencing factors with more than 3 times of reports and statistically significant results were obtained. For age[OR=0.97，95%CI（0.94，0.99），P=0.010]，Acute Physiology and Chronic Health Evaluation score Ⅱ[OR=1.17，95%CI（1.01，1.36），P=0.040]，comorbidity with diabetes[OR=1.21，95%CI（1.05，1.40），P=0.008]，comor-bidity with neurological diseases[OR=0.85，95%CI（0.74，0.98），P=0.020]，mechanical ventilation[OR=3.21，95%CI（1.82，5.66），P<0.001]，using sedative analgesics[OR=2.27，95%CI（1.66，3.10），P<0.001]，using gastric motility drugs[OR=0.23，95%CI（0.15，0.36），P<0.001]. The incidence of enteral nutrition intolerance was 35.00%. The risk prediction model for enteral nutrition intolerance in critically ill patients was logit（P）=-0.619-0.031×age+0.157×APACHE Ⅱ+0.191×comorbidity with diabetes-0.163×comorbidity with neurosurgery+0.820×using sedatives and analgesics+1.166×mechanical ventilation-1.470×using gastric dynamic drugs. The area under the receiver operating characteristic curve of the model was 0.864. The maximum Youden index was 0.589. The sensitivity was 0.922. The specificity was 0.667. The corresponding clinical diagnostic threshold was 0.536. Hosmer-Leme-show test χ²=13.410，P=0.098. Brier score was 0.195. Conclusion The risk prediction model of enteral nutrition intolerance in critically ill patients based on large sample evidence-based medicine is universal，scientific and practical. It provides a tool for medical staff to identify patients with enteral nutrition feeding intolerance in ICU.

Key words: Intensive Care Units, Enteral Nutrition, Meta-Analysis, Prediction Model, Evidence-Based Nursing

摘要：

目的 探究危重症患者肠内营养喂养不耐受的影响因素，并建立风险预测模型，为医护人员尽早识别患者肠内营养喂养不耐受提供工具。方法 检索中国知网、PubMed、Web of Science等国内外数据库，并补充检索纳入研究的参考文献和灰色文献，检索时间自建库至2022年11月。由2名经过系统培训的研究人员独立筛选、提取资料，并评价文献质量。使用Review Manager 5.4软件进行Meta分析，所得因素综合效应量OR值取自然对数为模型预测公式中各因素的系数，不耐受发生率和未发生率比值的自然对数为公式的常数项。采用便利抽样法，于2022年12月—2023年6月在南京市某三级甲等综合医院ICU收集360例符合纳入标准的患者为模型验证组，将收集的临床资料代入预测模型，以评价模型的区分度和校准度。结果 纳入13篇文献，获得7个报告次数≥3次且结果具有统计学意义的影响因素，为年龄[OR=0.97，95%CI（0.94，0.99），P=0.010]、急性生理与慢性健康状况评分Ⅱ[OR=1.17，95%CI（1.01，1.36），P=0.040]、合并糖尿病[OR=1.21，95%CI（1.05，1.40），P=0.008]、合并神经系统疾病[OR=0.85，95%CI（0.74，0.98），P=0.020]、进行机械通气[OR=3.21，95%CI（1.82，5.66），P<0.001]、使用镇静镇痛剂[OR=2.27，95%CI（1.66，3.10），P<0.001]、 使用促胃动力药[OR=0.23，95%CI（0.15，0.36），P<0.001]。肠内营养喂养不耐受的发生率为35.00%。危重症患者肠内营养喂养不耐受风险预测模型为logit（P）=-0.619-0.031×年龄+0.157×急性生理与慢性健康状况评分Ⅱ+0.191×合并糖尿病-0.163×合并神经系统疾病+0.820×使用镇静镇痛剂+1.166×进行机械通气-1.470×使用促胃动力药。该模型的受试者操作特征曲线下面积为0.864，约登指数最大值为0.589，灵敏度为0.922，特异度为0.667，对应的临床诊断阈值为0.536。Hosmer-Lemeshow检验χ²=13.410，P=0.098，Brier得分为0.195。结论 该研究构建的危重症患者肠内营养喂养不耐受风险预测模型具有普适性、科学性和实用性，为医护人员识别ICU肠内营养喂养不耐受提供了工具。

关键词: 重症监护病房, 肠内营养, Meta分析, 预测模型, 循证护理学