FDA launches data platform

The Food and Drug Administration is opening up information on adverse drug events for developers who want to build applications using the data. It's a big step in a larger effort, dubbed openFDA, to create an open-source platform for information on the medications and devices the agency regulates.

OpenFDA is an application programming interface (API) that connects agency data to developers who want to build applications that harness FDA information.

In the case of adverse drug events, the data comes from a variety of sources, including clinician reports, consumer complaints and the pharmaceutical industry, which is required to send information on adverse reactions to prescription and over-the-counter drugs to the FDA. Before openFDA, a Freedom of Information Act request was generally needed to assemble all the information on adverse events associated with a drug.

The data, which covers 2004 to 2013, is stripped of any personally identifiable information and can be searched in a variety of ways. There are structured unique identifiers for drugs and ingredients, but officials also felt it was important to have unstructured access to the information, said Dr. Taha Kass-Hout, the FDA's chief health informatics officer. Developers can build apps that give access to drug information by trade name, generic name or other information, and openFDA allows for misspellings.

"We want to be able to give you something back regardless of how you ask the question," Kass-Hout said.

The API lives on FDA's cloud. Developers get a key for bulk access, which supports a maximum of 60,000 queries per day. A developer whose app exceeds that maximum must make support arrangements with FDA -- a prospect that seems a long way off on Day One of a pilot project. The code is available for review, reuse and collaboration on GitHub.

Although the data is ready for developers and researchers, it is not intended to drive treatment. "This is a beta release," Kass-Hout said. "We have plenty of disclaimers not to use this for clinical decisions."

The concept of openFDA was developed about a year ago, he added, and the work to identify relevant datasets and build openFDA began in September 2013. Internal developers identified more than 80 datasets in the public domain, and through consultation internally and externally with other developers, they decided on adverse events, product recalls and labeling information as a good starting point. The FDA plans to release the datasets on product recalls and labeling information by the end of the summer.