机器学习临床预测模型构建策略与建议

Strategies and Recommendations for the Development of Clinical Machine Learning Predictive Models

目的提出临床预测模型的构建策略,帮助研究人员规范化开展临床预测模型研究。方法通过文献检索,总结构建临床预测模型的操作步骤和内容;通过专家咨询,总结提炼临床预测模型构建方法学框架。结果通过总结456篇临床预测模型研究经验和专家咨询,提出临床预测模型构建的11个步骤,并对细节问题进行了分析和阐述。结论该研究针对临床预测模型构建方法学提出了一些策略和建议,预期可为研究人员提供参考。

Objective To propose strategies for developing clinical predictive models,aiming to assist researchers in conducting standardized clinical prediction model studies.Methods Literature review was conducted to summarize the operational steps and content for developing clinical predictive models.Then,a methodological framework was summarized and refined through expert consultation.Results The 11-step methodological framework for developing clinical predictive models was obtained by synthesizing the experience of 456 clinical predictive modeling studies and expert consultation,and the details were analyzed and elaborated.Conclusions This study presents methodological strategies and recommendations for the development of clinical predictive models,intended to serve as a guide for researchers.