[Development of a pre-processing workflow for real world data derived from multicenter clinical laboratories].

Abstract

Objective: To develop a pre-processing workflow of real world data (RWD) derived from multicenter clinical laboratories so that the level of data standardization can be improved, and subsequently to produce more robust real world evidence (RWE). Methods: Purpose sampling was used to invite senior experts with experience in clinical research utilizing RWD, covering the fields of clinical laboratory, epidemiology, biostatistics, and clinical medicine. In-depth, semi-structured individual interviews were conducted and thematic analysis was used to analyze the collected data. Results: The in-depth interviews were completed in 16 experts. The experts unanimously agreed that pre-processing RWD derived from multicenter clinical laboratories is necessary prior to its application in research. Based on experts' insights, a comprehensive pre-processing workflow for RWD was constructed, comprising six key steps: ①developing a variable list based on research questions and distributing it to each clinical laboratory; ②conducting an initial quality assessment of RWD based on existing quality control results in clinical laboratories; ③cleaning the data; ④determining whether RWD (including categorical and continuous variables) is heterogeneity among different clinical laboratories; ⑤exploring potential sources of heterogeneity;⑥pre-processing RWD based on identified causes contributing to heterogeneity. Conclusion: The pre-processing workflow of RWD was established, to provide a methodological reference for controlling systematic errors in RWD derived from multicenter clinical laboratories, thereby enhancing the validity of RWE.