Big Data to Big Story: Data-mining 5 million health records to reveal the long-term effects of drugs
Using an open-source platform for mining electronic health records, we can illuminate the most important patient experiences that would not otherwise come to light, including long-term drug effects, and share them publicly.
This project would be a collaboration between Harvard Medical School’s Center for Biomedical Informatics [Harvard professors Dr. Isaac Kohane, director of Harvard’s Countway Library of Medicine, and Dr. Kenneth Mandl] and WBUR’s health reporting team [Carey Goldberg and colleagues].Harvard’s informaticians have helped develop a tool for mining electronic health records called i2b2, which now extends to a total of some 50 million patients worldwide, and they have direct access to some 5 million patient records.Big Data to Big Story would allow the public access to what has heretofore been mainly a tool for academic researchers. It would create a board of doctors, health economists and patient advocates to help determine which queries should be put to the i2b2 network. It would then help the public learn from the results by presenting them as stories and graphics.An example of i2b2 already at work: In 2009, there was concern that the diabetes drug Avandia was causing heart attacks, but the data were unclear. So using i2b2, Dr. Kohane’s team inspected the data from just a couple of large Boston hospitals. Within weeks, they found that Avandia was, in fact, causing a much higher rate of heart attacks than other drugs in the same class. The FDA cited their study when it issued a “black box” warning for Avandia — and that research may have saved thousands of lives
Big Data to Big Story leverages academic medicine’s foremost health-record data-mining tool to allow the public to gain critical, previously unavailable answers about medical treatments and their outcomes.
Who is the audience for this project? How does it meet their needs?
Patients depend on doctors, who depend on an “evidence base” largely built from short clinical trials; any patient making a medical decision could benefit from the longer-term, commercially unbiased, real-life data that i2B2 can produce. Big Data to Big Story would allow a panel of doctors and patients to choose the most critical queries to be put to the data-mining tool, and all patients with relevant conditions would benefit from the results. It would also benefit the consumers of journalism, because stories would be conceived based not on anecdote but on significant trend data.
What does success look like?
Success looks like a series of valuable insights into the real-life results of treatments that would not have come to light if not for a set of carefully chosen i2B2 investigations. It would include proof of widespread dissemination of that information and, in the longer run, metrics showing that the information had affected treatment choices by patients and doctors.