Authors:
Background: The NCCN Outcomes Database was created in 1997 to understand practice patterns, correlate them with patient outcomes, and inform clinical guidelines. The first version of the database was visionary in intent but difficult and costly to maintain, because it relied on distributed data abstraction at NCCN Member Institutions. NCCN recognized that a more sustainable model was needed, leveraging widespread availability of electronic health records (EHRs), a common data model, and the technology infrastructure to support data-sharing and analysis at scale. Methods: In 2015, NCCN initiated a collaboration with Flatiron Health (New York City, NY), a software company whose mission is to improve treatment and accelerate research in oncology. An interdisciplinary Flatiron team (oncologists, software engineers, informaticians, and database architects) is taking a systems engineering approach to the problem of clinical data curation. In collaboration with NCCN Guideline Panels, a common data model reflecting contemporary care was established across 6 vanguard tumor types. Scalable engineering pipelines were built to ingest and normalize discrete data fields from the EHR (eg, medication, laboratory values). To collect data from unstructured free-text, such as visit notes and pathology reports, Flatiron deployed a “technology-enabled” chart abstraction in which software assists clinically trained human abstractors in identifying data (eg, biomarker results). All data in the Outcomes Database are centrally deidentified and quality-controlled. Results: Flatiron has undertaken EHR integration with 2 large NCCN Member Institutions (>100,000 total patient records) and relaunched the Quality & Outcomes Database in Non–Small Cell Lung Cancer, Melanoma, Breast Cancer, and Colorectal Cancer at these centers. Data are refreshed by Flatiron quarterly, and processed patient-level data tables have been returned to each participating institution. Cohorts in additional tumor types will be launched throughout 2017. A secure Web portal supports database query and cohort visualizations. Quality measurement and guideline concordance solutions are being added to the portal, with benchmarking against peer institutions. Conclusions: NCCN collaboration with a technology group has enabled sustainable, standardized data capture and reduced chart abstraction and resource commitment from member centers. The NCCN-Flatiron Outcomes Database is one example of a data-sharing network that can support collaborative evidence development, quality improvement, and outcomes research.
Sources:
NCCN Annual Conference