IBM, Mayo Clinic to mine medical data.

The Mayo Clinic and IBM are cooperating on an ambitious program to eventually use computers and electronic records to design individualized patient treatments.

According to The Wall Street Journal, the multi-year collaboration involves one of the world's largest medical databases, containing complete records on nearly 4.4 million patients. The goal is to allow a doctor to be able to ask an online computer system how the last 100 Mayo clinic patients with the same gender, age, and medical history responded to particular treatments.

The project will use the same pattern-recognition and data-mining tools used by direct-mail companies to determine whom to target with marketing offers and by banks to identify credit-card scam victims within their customer base. Mayo reportedly hopes that applying those same tools to medical records will help it find patterns that currently are discovered by chance.

IBM and Mayo have been working since 2001 to computerize all Mayo's files from its three hospitals in Jacksonville, Florida, Scottsdale, Arizona, and Rochester, Minnesota. The records include lab tests, scans, cardiograms, and X-rays. The IBM project will add more data, such as information on patients' protein and genetic makeup.

Mayo has been systematically keeping patient records and using them in treatment since 1907. For years, it has employed researchers who comb through paper records looking for patterns of diseases and treatments. Now, it envisions doing that electronically so that its information systems and others nationwide will be a repository of all known information on how to care for a patient.

One challenge, however, is that computer companies have not always been able to deliver the results promised by such data-mining, mostly because computers have trouble distinguishing significant from insignificant correlations--a problem that becomes more difficult the larger the database is.

IBM security experts have helped devise ways to make patient information available for research purposes without compromising patient privacy. The database, which planners say will comply with new federal guidelines on patient privacy, will not include information that could identify specific individuals. IBM is also reportedly working on artificial intelligence techniques to spot and scrub identifiable information before humans see it. Doctors say computerized pattern-recognition of genetic data could help researchers make sense of many current medical mysteries.