中国老年2型糖尿病患者认知衰弱风险预测模型的系统评价

Systematic Evaluation of a Predictive Model for the Risk of Cognitive Frailty in Elderly Chinese Patients with Type 2 Diabetes Mellitus

摘要: 目的：系统评价国内用于预测老年2型糖尿病患者发生认知衰弱的风险预测模型，为临床选择风险评估工具和后续研究提供参考。方法：计算机检索中国知网、维普网、万方数据知识服务平台、中华医学期刊全文数据库、PubMed、Embase、Cochrane Library发表的相关文献，检索时间为建库至2025年3月。使用PROBAST偏倚风险评估工具对纳入研究进行评估，并对模型区分度指标进行Meta分析。结果：共纳入10项研究，所有研究中的模型均经过验证并使用受试者工作特征曲线下面积(area under the curve, AUC)报告了模型验证时的区分度，其中9项研究中的模型验证时的AUC > 0.8。9项研究报告了模型校准，6项研究报告了临床实用性评估。纳入模型最常见预测因子为年龄、抑郁、营养状况。所有研究的偏倚风险均为高风险，纳入研究的适用性良好。对模型验证时的AUC进行meta合并，结果为0.876 (95% Cl: 0.846~0.907)，提示模型整体具有良好的区分度。结论：老年2型糖尿病患者认知衰弱风险预测模型的建模具有较高的区分度但在应用时存在高偏倚风险，且普遍缺乏外部验证。在研究样本选择、原始数据处理、算法使用和统计分析方面有待进一步提升。未来研究可考虑参照方法学指南构建新的预测模型并进行验证。

Abstract: Objective: To systematically evaluate the risk prediction models used in China to predict the occurrence of cognitive frailty in elderly patients with type 2 diabetes mellitus, and to provide a reference for clinical selection of risk assessment tools and follow-up studies. Methods: Computerized searches were conducted for relevant literature published in CNKI, WEIPU, WanFang Database, China Medical Journal Full Text Database, PubMed, Embase, and Cochrane Library, and the search was conducted from the establishment of the libraries to March 2025. The included studies were evaluated using the PROBAST risk of bias assessment tool, and meta-analysis was performed on the model differentiation index. Results: A total of 10 studies were included, all of which had validated models and reported discrimination at model validation using the area under the curve (AUC) of the subjects’ work characteristics, with nine studies having an AUC > 0.8 at model validation. Nine studies reported model calibration, and six studies reported an assessment of clinical utility. The most common predictors included in the models were age, depression, and nutritional status. The risk of bias was high for all studies, and the applicability of the included studies was good. Meta-merging of the AUC at model validation resulted in 0.876 (95% Cl: 0.846~0.907), suggesting that the models were overall well discriminated. Conclusion: The modeling quality of the cognitive frailty risk prediction model for elderly type 2 diabetic patients was good, with high discriminability but a high risk of bias in its application and a general lack of external validation. Further improvements are needed in the selection of study samples, raw data processing, use of algorithms and statistical analysis. Future studies may consider constructing new predictive models with reference to methodological guidelines and validating them.