Presentation (CTI)

OBJECTIVES: Regional healthcare data (e.g., hospital electronic medical records [EMR] and/or claims) in China represent an important source for real-world evidence (RWE) generation, but their use is constrained by heterogeneous data structures, local data governance requirements, and restrictions on transferring patient-level information outside of respective secured data environment. To support compliant, scalable use of regional healthcare data for RWE generation, an integrated framework and AI-assisted platform was developed and piloted in the identification of patients with common cancers in China.
METHODS: The framework includes three components. First, a query-oriented regional healthcare common data model (RH-CDM) was designed to harmonize heterogeneous regional data sources into an analysis-ready structure. Second, a data preparation tool was developed to support regional data sources in transforming raw hospital EMR and/or claims into RH-CDM-ready datasets. Third, DataScope, an AI-assisted feasibility assessment platform, was built to facilitate external research groups construct SQL queries against RH-CDM remotely, dispatch approved queries to participating regional databases and receive anonymized aggregate outputs after local execution.
RESULTS: In pilot implementation, the framework was deployed across four regional databases in China, covering approximately 70 million individuals. Fujian Province regional data, the biggest participating source to date covering 37 million individuals, was used to assess oncology cohort availability. The assessment identified approximately 150,000~200,000 breast cancer patients, with average follow-up time of 600-650 days; 300,000~350,000 lung cancer patients, with average follow-up time of 300-350 days; and 150,000~200,000 colorectal cancer patients, with average follow-up time of 400-450 days. The feasibility assessment results were returned within 2-3 working days. As only the anonymized query results including patient count and data value distribution were returned, no patient-level data left the secured local data environments, ensuring full compliance with local data governance requirements.
CONCLUSIONS: The framework provides a practical infrastructure for compliant, rapid, and scalable RWE generation while preserving regional data sovereignty.