摘要
目的:评价现有脓毒症相关性谵妄(SAD)风险预测模型的研究方法学与报告质量。方法:系统检索中英文数据库,检索时限从建库至2024年8月31日。纳入以构建或验证SAD风险预测模型为研究主题的文献。由两名研究者独立进行文献筛选,采用预测建模研究系统评价的批判性评估和数据提取清单(CHARMS)提取数据,分析模型的预测性能指标和关键预测因子。使用预测模型偏倚风险评估工具(PROBAST)评估模型的偏倚风险和适用性,应用个体预后与诊断预测模型研究报告规范(TRIPOD)声明评价报告质量。结果:共纳入9项研究,纳入模型受试者工作特征曲线下面积(AUC)为0.742~0.910,模型总体预测性能良好,其中出现频率较高的预测因子有年龄、序贯器官衰竭评估(SOFA)、呼吸频率及使用镇静药物咪达唑仑等。方法学评价结果显示:在偏倚风险方面,8项研究为高风险,1项为低风险;在适用性方面,3项研究为高风险,6项为低风险。TRIPOD报告质量评价平均完成度为77.78%(范围:63.64%~90.91%),其中“样本量计算”“风险分层”“补充分析”和“代码共享”等条目报告缺失严重。结论:现有SAD风险预测模型在预测性能和临床适用性方面整体表现良好,但普遍存在较高的方法学偏倚风险,且报告质量有待提高,尤其在关键方法学细节和透明化报告方面存在显著不足。未来研究应严格遵循PROBAST和TRIPOD指南,以提升模型的可靠性、适用性和可重复性。
Objective:To evaluate the methodological and reporting quality of existing risk prediction models for sepsis-associated delirium(SAD).Methods:A systematic search was conducted in both English and Chinese databases,covering all records up to August 31,2024.Studies focusing on the development or validation of SAD risk prediction models were included.Two reviewers independently screened the literature.Data were extracted using the CHARMS checklist to analyze model performance metrics and key predictors.The Prediction Model Risk of Bias Assessment Tool(PROBAST)was used to assess risk of bias and applicability.Reporting quality was evaluated using the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis(TRIPOD)statement.Results:A total of nine studies were included.The area under the receiver operating characteristic curve(AUC)for the included models ranged from 0.742 to 0.910,indicating good predictive performance.Frequently reported predictors included age,Sequential Organ Failure Assessment(SOFA),respiratory rate,and use of the sedative midazolam.Methodological assessment showed that eight studies had a high risk of bias and one had a low risk.Regarding applicability,three studies were rated high risk and six low risk.The average reporting completeness based on TRIPOD was 77.78%(range:63.64%-90.91%).Items such as sample size calculation,risk stratification,supplementary analyses,and code sharing were frequently underreported.Conclusion:Current SAD risk prediction models generally exhibit good predictive performance and clinical applicability.However,they commonly suffer from high methodological bias and suboptimal reporting quality,especially in key methodological details and transparency.Future studies should adhere strictly to PROBAST and TRIPOD guidelines to improve model reliability,applicability,and reproducibility.
作者
章丹凤
刘娟
史广玲
胡大玲
孙晓兰
张颖
沈文瑞
闵艳
ZHANG Danfeng;LIU Juan;SHI Guangling;Hu Daling;SUN Xiaolan;ZHANG Ying;SHEN Wenrui;MIN Yan(Department of Neurosurgery,Sir Run Run Hospital,Nanjing Medical University,Nanjing 211100,China;Intensive Care Unit,Affiliated Brain Hospital of Nanjing Medical University,Nanjing 210029,China;Department of Geriatrics,Sir Run Run Hospital,Nanjing Medical University,Nanjing 211100,China)
出处
《东南大学学报(医学版)》
2025年第6期932-940,共9页
Journal of Southeast University(Medical Science Edition)
关键词
脓毒症
脓毒症相关性谵妄
预测模型
方法学质量
报告质量
预测模型偏倚风险评估工具
个体预后与诊断预测模型研究报告规范
sepsis
sepsis-associated delirium
prediction model
methodological quality
reporting quality
Prediction Model Risk of Bias Assessment Tool
transparent reporting of a multivariable prediction model for individual prognosis or diagnosis
