Abstract:

Objective To construct and validate a risk prediction model of hypoglycemia in emergency intensive care unit（EICU） patients. Methods A retrospective study was conducted among 2 093 EICU patients in a department of a tertiary A hospital in Urumqi from January to December 2022，as research subjects. Univariate analysis and logistic regression analysis were used to determine the risk factors for hypoglycemia，and R software was used to establish a nomogram prediction model. The area urder the receiver operator characteristic（ROC） curve was used to test the model differentiation，and the Hosmer-Lemeshow test was used to test the goodness of fit of the model. The risk prediction model was validated by the prospective study with inclusion of 699 EICU patients admitted to the same hospital from January to March 2023. Results The model variables included whether hypoglycemia occurred in the past year，acute physiology and chronic health evaluation Ⅱ score at admission，coefficient of variation of blood glucose，history of renal disease，history of diabetes，insulin treatment，and serum creatinine. The Hosmer-Lemeshow test of the model was P=0.497；the area urder the ROC curve was 0.820（95%CI：0.794~0.847）；the best cutoff value was 0.495；the sensitivity was 0.856；the specificity was 0.751. The model validation results showed that the Hosmer-Lemeshow test P=0.537；the area urder the ROC curve was 0.859（95%CI：0.819~0.898）；the best cutoff value was 0.597；the sensitivity was 0.840；the specificity was 0.757. Conclusion The established nomogram prediction model helps clinical staff to screen patients at high risk of hypoglycemia and provides a reference for optimizing the management of hypoglycemia in EICU patients.

Key words: Emergency Intensive Care Unit, Hypoglycemia, Risk Factors, Prediction Model, Nomograms, Nursing Care

摘要：

目的 构建并验证急诊重症监护室（emergency intensive care unit，EICU）患者低血糖风险预测模型。 方法 回顾性收集2022年1月—12月乌鲁木齐市某三级甲等综合医院EICU收治的2 093例患者作为调查对象；通过单因素分析和Logistic回归分析筛选发生低血糖的危险因素，应用R软件构建列线图预测模型。采用受试者操作特征曲线下面积检测模型的区分度，采用Hosmer-Lemeshow检验判断模型的拟合优度。采用前瞻性研究设计，便利选取2023年1月—3月同一所医院EICU收治的699例患者对模型进行验证。 结果 模型变量包括近1年是否发生低血糖、入院时急性生理与慢性健康状况Ⅱ评分、血糖变异系数、肾脏疾病史、糖尿病史、是否使用胰岛素治疗和血肌酐水平，Hosmer-Lemeshow检验结果显示P=0.497，受试者操作特征曲线下面积为0.820（95%CI：0.794~0.847），最佳临界值为0.495，灵敏度为0.856，特异度为0.751。模型验证结果显示，Hosmer-Lemeshow检验P=0.537，受试者操作特征曲线下面积为0.859（95%CI：0.819~0.898），最佳临界值为0.597，灵敏度为0.840，特异度为0.757。 结论 该研究建立的列线图预测模型有助于临床医护人员筛选发生低血糖的高危患者，为优化EICU患者低血糖的管理提供参考依据。

关键词: 急诊重症监护室, 低血糖, 危险因素, 预测模型, 列线图, 护理