Mayo Clinic Taps IBM Tools to Mine Patient Data

The Minnesota medical clinic recently joined with IBM to build a data repository to harness massive amounts of medical records containing clinical information from Mayo's 4.4 million patients.

In an effort to improve overall patient care, the Mayo Clinic is tunneling deep into its data archives to link patients genomic profile, personal history and background characteristics.

The world-class medical institution in Rochester, Minn., recently joined with IBM to build a data repository to harness massive amounts of medical records containing clinical information from Mayos 4.4 million patients. As part of its collaboration with IBM, the Mayo Clinic will tap IBM technologies and the companys Blue Gene supercomputer, providing access to specialized algorithms for areas such as molecular modeling for disease research.

Once all Mayos federated data sources are integrated—a process thats expected to take up to two years—the new technology will perform data mining, analysis and tailored diagnostic queries that will be used by physicians in daily practice for immediate prescribing.

"As we look into the future, the care of patients and how its delivered will change fairly dramatically over the next decade. The types of tools were developing with IBM will help us get there," said Dr. Nina Schwenk, chair of the foundation for IT at Mayo.

One major benefit of the collaboration will allow physicians to turn away from broad-based medical research by giving them access to a database that features an individuals genome characteristics and patterns, said Eric Brown, an analyst at Forrester Research Inc. "Theres something really different about having your physician—instead of turning to general information about people—turning to a database that has special information about people like you, your diseases, your background, what works with that small group," said Brown in Cambridge, Mass. "Its basically a research study tailored to you in your particular situation. ... Call it personalized medicine."

However, discovering and tying together a wide range of data from electronic medical records to doctors unstructured notes, for example, required careful adherence to privacy rules, said Drew Flaada, director of the Mayo-IBM Collaboration, at IBM, in Armonk, N.Y. "We had to retrieve federated data from about five different data sources and bring the information together in a single data warehouse—it has forced a new level of security and privacy far beyond HIPAA [the Health Insurance Portability and Accountability Act] and any federated regulations when they look at all this," Flaada said.

The federal government has its own plans to catapult medical records into the electronic age. Last month, Department of Health & Human Services Secretary Tommy Thompson outlined a 10-year plan to transform the delivery of health care by building a new health information infrastructure, including digitized health records and a new network to link health records across the nation.

IBM and Mayo may be getting a head start. According to Flaada, once Mayo put IBMs technology to work—a project that took five Mayo staffers 13 months to complete—the time it took to find a list of potential candidates to join a study group was cut to 16 seconds.

At the heart of the Mayo project is DiscoveryLink, IBMs data federation product. Other offerings that were deployed include IBMs DB2 Universal Database and the companys pSeries servers running AIX. In addition, to incorporate a data virtualization approach, IBM built DDQB (Data Discovery and Query Builder). DDQB provides an abstracted view of queries that allows physicians to execute searches against medical codes and languages with which theyre familiar. The abstraction engine allows different data sources, such as public-record domains, to be plugged in via a set of APIs to write to. It also offers decision support techniques.

The second phase of the project entails mining vast amounts of data in available public databases on genomic research, such as the National Cancer Institutes database. Beyond that, the project, which extends the data search across Mayos Jacksonville, Fla., and Scottsdale, Ariz., campuses early next year, will also add proteomics and molecular modeling under its umbrella.

"Hopefully, with genomics information, we can say, This is the drug for you and get patients [treatments] individually controlled for a shorter period of time rather than do clinical trials," said Mayos Schwenk.


Check out eWEEK.coms Health Care Center at for the latest news, views and analysis of technologys impact on health care.