基于人工智能技术的临床数据资源中心建设实践

Construction of a clinical data resource center based on AI technology

目的:在医院数据中心基础上,进一步融合院外及公开多源异构数据,并通过标准化数据建模和数据治理,构建高质量临床数据资源中心,以支持医院开展高质量临床研究,并将研究成果反哺临床,提升研究成果转化能力,实现从数据到科研到临床的一体化发展闭环。方法:利用大数据及人工智能技术,设计多层次、多维度的数据模型,将原始数据转化为标准化的,可满足科研、临床等不同场景需求的多模态数据集合。结果:临床数据资源中心已入库治理5000余万诊次临床诊疗数据、1000多万条随访数据,整合45个既往与在研课题研究数据,18个气象、文献及基因等公开数据;形成614个数据模型和10479个研究可用字段;支持CDSS等数据服务调用频率高达162万次/天;支撑521位科研人员和团队建立了1532个研究人群和专病库,已有62个研究模型成果转化反哺临床。结论:通过建设高质量临床数据资源中心,可高效满足临床研究团队对相关数据的使用需求,且随用、随查、随取、随分析,降低了数据使用门槛,提升了研究效率。

Objective To further integrate out-of-hospital and open multi-source heterogeneous data on the basis of the hospital data center,and to build a high-quality clinical data resource center through standardized data modeling and data governance,so as to support the hospital to carry out high-quality clinical research,feed the research results back to the clinic,improve the ability to apply the research results,and achieve an integrated development closed-loop from data to scientific research to clinical.Methods A multi-level and multi-dimensional data model was designed using big data and AI technology,and the original data was converted into a standardized multi-modal data set that could meet the requirements of different scenarios such as scientific research and clinical.Results The clinical data resource center has stored more than 50 million clinical treatment data and more than 10 million follow-up data;integrated 45 previous and ongoing research data,and 18 public meteorological,literature,and genetic data,etc.;formed 614 data models and 10,479 research fields available;supported CDSS and other data services with a call frequency of 1.62 million times/day.The center also supported 521 researchers and teams to establish 1,532 research populations and special disease banks,and 62 research models have been translated into clinical results.Conclusion The construction of the highquality clinical data resource center can effectively meet the needs of the clinical research team for the use of relevant data,and it can be used,checked,retrieved,and analyzed at any time,lowering the threshold of data use and improving research efficiency.