Stanford Cancer Institute News
Supporting the Growing Data Needs of Cancer Researchers
SCI Research Database Integrates Clinical, Molecular, and Imaging Data
If the wealth of information on patients’ medical histories, tumor biopsies, imaging results, and even genetics could be integrated into one database, it could greatly facilitate the performance of large studies spanning multiple cancer types or using many different types of information.
That’s the purpose of the Stanford Cancer Institute Research Database (SCIRDB).
“Many cancer centers create databases from their electronic medical systems, but what makes SCIRDB unique is that it’s pulling data from multiple sources,” said Daniel Rubin, MD, MS, an Associate Professor of Biomedical Data Science, Radiology, and Medicine; and the Director of Biomedical Informatics for the Stanford Cancer Institute.
EPIC, the electronic medical record system used by Stanford Health Care, includes primarily clinical data, Rubin said, and isn’t designed for research purposes. Making changes to the EPIC database, to add research data fields or by trying to expand the scope of data that it includes, involves a long and costly process. In the past, this has often led individual investigators who need a research database to design their own. But Rubin and others at the SCI thought it would be most cost and design efficient to have a single all-inclusive database that could be used for SCI research purposes. So they hired a team of programmers to build the SCIRDB, and then they gathered information from researchers on what data and features they wanted.
Adding imaging results to the database, for instance, was important to many researchers. Images associated with tumors were already stored in patient records, but studying them involved analyzing each scan one by one. Accordingly, Rubin and his colleagues developed methods to extract quantitative information from radiology images, and those data are now stored in the SCIRDB.
“It’s key to not only have a text report associated with an image, but also the actual structured measurements,” said Rubin. “When we have that, we can use information on how tumors are growing or shrinking to conduct studies at a population level.” If a researcher wants to study progression-free survival—how long tumors remain stagnant—after a particular treatment, they can now do that more easily with the measurements included in the database.
The SCIRDB was helpful when Rubin collaborated with Michael Gensheimer, MD, a Clinical Assistant Professor of Radiation Oncology and SCI member, to analyze more than 13,000 patients seen at Stanford from 2008 through 2017 who all had metastatic cancer. They mined the database to build a computational model that could predict—using raw medical records data—which patients would survive more than or less than three months, which could help guide clinical care decisions.
As the database grows and expands, Rubin expects it will allow more types of studies and queries to be carried out. For instance, if a patient with a rare cancer is deciding between two treatments, and there is no published clinical trial result that could guide decision making, a physician could search the SCIRDB for the outcomes of similar patients, and he or she could use the information to inform patient care.
“Clinical care is evolving, with the potential of becoming much more data driven, and we’re here to help clinicians and researchers leverage that,” said Rubin.