Hospitals and clinics in several states are trying out predictive medicine programs that, when linked to electronic health records, offer a probability of a diagnosis, but the Wall Street Journal reported not all doctors involved in the trials are sold on the technology.

The software being tested at New York hospitals run by Northwest Health pops up when physicians are viewing electronic records. Doctors are asked to answer several questions about a patient’s condition, and then the tool offers a prediction on the probability the patient has the suspected ailment.

The New York trial is limited to possible pulmonary embolism patients, while nearly three-dozen clinics in Wisconsin and Utah are participating in a trial to use the tool to predict strep and pneumonia.

The Wisconsin trials are being overseen by David Feldstein, MD, associate professor at the University of Wisconsin School of Medicine and Public Health (UW Health). Feldstein told WSJ so far, physicians are only using the predictive tool in 20 percent of cases, but he hopes educating doctors will change those habits.

“There is a big backlash against clinical decision support,” Feldstein said.

There is some resistance among physicians about trusting a computer program’s analysis rather than their own medical opinion. John Beasley, MD, a professor of family medicine at UW Health whose clinic is participating in the trial, said when the tool pops up on his screen, he ignores it.

“On one side are the people who want to take a statistical approach to this, versus those of us who say this is a humanistic enterprise and you frequently do stuff that is irrational because it is good for the patient,” Beasley said.

WSJ cited two earlier studies which showed varying results of success using these kind of predictive tools.

In a 2013 study published in JAMA Internal Medicine and authored by Thomas McGinn, MD, MPH, who is running Northwell’s pilot program, physicians who used the predictive models ordered fewer antibiotic prescriptions, rapid streptococcal tests, and pharyngitis throat cultures.

A 2015 study co-authored by Patricia Silveira, MD, of Brigham and Women’s Hospital, however, found problems with a tool to predict deep-vein thrombosis, finding the model overestimated the likelihood of a clot, leading to physicians prescribing more blood thinners.