The computer that decides if we live or die

A serious injury leaves a loved one in a coma. Relatives may face the hardest decision of their lives: to wait it out or turn off the life-support machine.

The program may help families decide what action to take if a loved one is critically ill

But now, that critical decision may be turned over to a sophisticated computer program. New software should soon be able to predict more accurately than loved ones how comatose patients would choose to be treated, if they were able to make the decision themselves.

Bioethicist David Wendler at the U.S. National Institutes of Health in Washington D.C., and his colleagues, used very elementary past research to build up patterns in patients' choices. "There was very little data available and the approach we used was incredibly simplistic," Wendler concedes. "But even with a little amount of data, we did very well."

The study compared how accurately their computer-based tool predicted a patient's preferred treatments compared with what loved ones said. Results showed both methods got it right around two-thirds to three-quarters of the time.

Wendler hopes to build up a broader data bank of personal profiles, which will include age, gender, religious and ethnic background, to advance the software. He is confident that will enable more accurate patient predictions. "We have very good reason to believe we can get significantly better results," Wendler says. "Maybe ten or fifteen percent more accurate than (next of kin)."

Patients have gained more control over their medical care in recent years but many still fail to sign a directive looking to the future. Few discuss treatment preferences they would elect if they lost the ability to make decisions. Without a self-directed advance medical plan for a patient, relatives are often asked to step in and act on a loved one's behalf.

"We've always gone with the idea that people who know the patient best are also best positioned to make the decision about treatment," Wendler says. "My concerns were that this process puts a burden on families. I wanted to develop an alternate approach."

Wendler is acutely aware of the problems a software program like this might pose for the community at large. "Some people say, 'of course this is good' and others think 'this is crazy'," Wendler says.