Cancer Informatics Scholar focused on Real-World Data and Evidence NCI’s Office of Cancer Centers is offering supplemental funding to the P30 Cancer Center Support Grant (CCSG) to stimulate research in machine learning and artificial intelligence tools that can extract Real-World Data (RWD) from EHRs at NCI-designated cancer centers. It is expected that at the end of the project period, sites selected to receive these supplemental awards will share their tools (i.e. newly developed or reused) with others in the community.
Scholars will develop AI models or a set of tools that rapidly addresses the completeness and quality of EHR data based on two specific cancers. Scholars will develop their algorithms around one common and one rare cancer of the Scholar’s choosing (e.g. prostate cancer and brain tumors, respectively). The EHR data and analyses must be executed within the institutional architecture and no data sharing is expected. Key performance measures for each cancer site chosen include: (1) how quickly the tool(s) can parse all elements (listed below) from the EHR for a given patient; (2) how quickly and accurately the tool(s) can reconcile across both syntactic and semantic data constructs for a given variable (e.g., race, performance status, etc.); (3) the ability to develop a human-readable summary table that includes the variables below; (4) validation approaches for the tool and/or AI approach; and lastly, (5) the portability and interoperability of the tool(s) to prepare a standardized, anonymous analytic dataset for exchange using the Fast Healthcare Interoperability Resources (FHIR) data standard. Scholars are expected to develop a toolset or repurpose an existing/open-source tool(s) that can extract this information from EHRs across the two cancer types they selected. At the end of the supplement period, meritorious Scholars will be invited to demonstrate their tool sets on a de novo cancer data set