Abstract:

Objective We systematically review the hypoglycemia risk prediction models for patients with diabetes mellitus,so as to provide references for nursing practice. Methods A systematic review was conducted. The CNKI,CBM,Wanfang,VIP,PubMed,Embase,Cochrane and CINAHL were searched to collect studies on hypoglycemia risk prediction models from database establishment to 28 December 2021. The reviewer independently screened the literature according to the pre-determined inclusion and exclusion criteria,extracted the data and evaluated the risk of bias of the included studies using the Prediction Model Risk of Bias Assessment Tool. Results A total of 14 pieces of the literature and 16 models were included with the total sample size of 257~ 1 173 820 cases and the outcome events of 72~7 030 cases. All studies reported the area under the receiver operating characteristic curve of the prediction models in the derivation and（or） validation datasets as from 0.63 to 0.96. Among them,12 models reported calibration metrics,and 5 models were externally validated. Independent predictors of multivariable model repeatability were age,history of hypoglycemia,use of insulin and other hypoglycemic drugs,and estimated glomerular filtration rate. Small sample size,lack of internal or external validation,inappropriate variable selection methods,and missing data processing can increase the risk of bias in the studies. Conclusion The overall prediction model of hypoglycemia risk has good performance and low concern for applicability. However,this finding should be interpreted with caution due to the risk of bias in the included studies. Future work should focus on following transparent reporting of multivariable prediction models for individual prognosis or diagnosis to develop and evaluate the hypoglycemia risk score for diabetes and to validate its applicability and feasibility in clinical practice.

Key words: Hypoglycemia, Diabetes Mellitus, Prediction, Model, Systematic Review, Nursing Care

摘要：

目的 系统评价糖尿病患者低血糖风险预测模型,为临床医护人员选择或开发适合的风险评估工具提供参考和借鉴。 方法 计算机检索中国知网、中国生物医学文献数据库、万方数据库、维普期刊库、PubMed、Embase、Cochrane图书馆和CINAHL数据库中与主题相关的文献,检索时间为建库至2021年12月28日,研究者根据纳入、排除标准筛选文献,两名研究员依据预测模型研究数据提取表和偏倚风险评估工具独立进行资料提取和质量评价。 结果 研究纳入14篇文献,16个低血糖风险预测模型,总样本量为257~1 173 820例,结局事件数为72~7 030例,受试者工作特征曲线下面积为0.63~0.96,其中12个模型报告了校准,5个模型进行了外部验证。多变量模型重复报告的独立预测因子有：年龄、既往（严重）低血糖史、使用胰岛素和其他降糖药物、肾功能。部分模型样本量小、缺乏内部验证或外部验证、不恰当的变量选择方法及缺失数据处理增加了研究的偏倚风险。 结论 低血糖风险预测模型整体呈现良好的区分、校准性能及适用性,但存在显著的方法学缺陷和高偏倚风险。未来研究应侧重于遵循多变量预测模型的透明报告来开发和评估糖尿病患者低血糖风险评分,并验证其在临床实践中的适用性和可行性。

关键词: 低血糖, 糖尿病, 预测, 模型, 系统评价, 护理