Data-Mining in Doctor's Office Helps Solve Medical Mysteries
(This story was corrected to remove remarks from a Microsoft executive that the company said were made in error.)
When hospitals look to Microsoft Corp., it's no longer just for the latest in office software. Some are getting help from the technology giant in diagnosing their patients.
Medical providers are experimenting with new ways to sift through their data and find patterns. The results can be surprising: Microsoft recently examined health records from a research partner, MedStar Washington Hospital Center in Washington, D.C., and found that the location of patients' rooms is one factor in whether they get sick. Such insights can help identify hot spots for bacteria and illness and stop their spread.
As hospitals digitize patient records and amass huge amounts of data, many are relying on companies such as Microsoft, SAS Institute Inc., International Business Machines Corp. and Oracle Corp., whose data-mining technologies can help them detect patterns and improve medical care.
The analytics industry generated more than $30 billion in revenue last year, according to research firm IDC. That is forecast to increase to $33.6 billion this year as the health-care industry becomes a bigger customer.
The use of data-mining technology has already led to some measurable improvements in patient care.
New York-Presbyterian Hospital has reduced the rate of potentially fatal blood clotting in patients by using analytics software, Dr. Nicholas Morrissey, a surgeon involved in the effort, said in an interview.
As patients were being admitted, the hospital started using Microsoft software in 2010 to scan records for risk factors, such as cancer, smoking and the amount of time patients were bed-bound. By letting the software rather than the hospital staff make the assessment, doctors saved time and made better evaluations, Morrissey said.
The clotting now happens at a rate of 0.23 incidents per 1,000 patient days, versus 0.33 incidents per 1,000 patient days before implementation of the software, Morrissey said. He stressed that further study of the software's impact is needed.
"I wouldn't be out there saying we've solved the problem, but we're definitely making progress -- that was a significant drop," he said.
Seton Healthcare Family, a hospital system in central Texas, started using data-mining software from IBM last year to sift through transcribed doctors' notes and other "unstructured data," such as numbers printed on medical images, for ways to lower readmission rates for congestive heart failure patients.
It found some surprising themes. Two of the strongest predictors of whether a patient would be rehospitalized were lack of an emotional support network and the presence of a bulging jugular vein -- two things hospital staff can easily look for when admitting patients.
"We've gotten some really tremendous results," said Ryan Leslie, Seton's vice president of analytics and health economics. The hospitals narrowed a list of 113 predictors that medical staff thought would be important down to 18, and even identified some that weren't on the original list, such as the bulging vein.
Besides improving patient care, hospitals' use of analytics software can make them eligible for millions of dollars in government funding if they can demonstrate "meaningful use" of their computer systems. The payments are part of the Obama administration's $14.6 billion program to prod health-care providers to adopt electronic medical records.
Federal law prohibits medical providers from disclosing certain health information without patient consent, but there is an exemption for activities that fall under "quality improvement," said Susan McAndrew, deputy director for health information privacy at the U.S. Department of Health and Human Services' Office for Civil Rights.
That means health-care providers generally don't have to tell patients that their records are being analyzed for the purpose of improving care, which has raised some privacy concerns.
"People do not like to have researchers of any stripe using their electronic health records," said Deborah Peel, founder of Austin, Texas-based Patient Privacy Rights. "As a matter of respect and autonomy and patient-centeredness, patients want to be asked. When they are asked, by and large they support this. It's the not being asked stuff that's really bad."
She pointed to a 2007 survey that found just 1 percent of respondents felt researchers didn't need to seek permission before using their medical data.
Some privacy experts disagree, warning of the dangers of over-notification.
Deven McGraw, director of the health privacy project with the Center for Democracy & Technology in Washington, said notifying patients too often could be unnecessarily confusing.
"You ask for it when it's used in ways that people might not expect," she said.
New York-Presbyterian Hospital and Seton Healthcare Family didn't notify their patients. MedStar Washington Hospital Center did not respond to a request for comment.
Morrissey of New York-Presbyterian, which published some of its results, emphasized that no personal information is ever publicly disclosed. He said the benefits outweigh the concerns.
"I think we're going to impact things that are most life-threatening," he said, emphasizing that while the results need further research, "it seems real, and it's significant."