Feature: Clinician 'hands-on' ordering reduces low-yield exams
By encouraging more clinician “hands-on” use of its radiology order entry system, Massachusetts General Hospital (MGH) has been able to reduce the number of low-yield imaging exams, according to a study published in the June issue of Radiology.

According to Vartan M. Vartanians, MD, from the department of radiology, along with his colleagues at MGH in Boston, the hospital was able to reduce a persistent low-yield imaging rate and encourage more clinician hands-on keyboard use of the system by instituting a rule that prevented nonclinician support staff from completing an order that generated a low-yield decision support score.

MGH’s decision support system provides a yield score, ranging from one to nine, based on clinical information provided in support of the exam request. The nine-point scoring system is similar to the nine-point scoring system used with the American College of Radiology appropriateness criteria—scores of one to three indicate low-yield, four to six marginal yield and scores of seven to nine indicated exam.

MGH’s system has been expanded so that scores assigned to particular exams have been modified and adapted according to continuous feedback from practicing clinicians and radiologists.

Prior to the rule change, the fraction of exams that were scheduled through the order entry system with a low-yield decision support score was relatively unchanged—between 3.5 percent and 5 percent on a quarterly basis.

According to the authors, after the policy change was put into place, the proportion of total examination requests by physicians directly logging into the system more than doubled--from 26.31 percent of 42,737 requests to 54.37 percent of 76,238 exams requested. The percentage of low-yield exams requested decreased from 5.4 percent of 38,801 to 1.9 percent of total of 65,765 total requests.

There was relatively no change in the rate (from 90.93 percent to 87.79 percent) of completion of exam requests not affected by the policy change.

The problem with trying to reduce that low-yield rate, co-author Daniel Rosenthal, MD, also from the department of radiology at MGH, said in an interview, is that “decision support can only be effective if it is presented to a person who is capable of changing his mind.”

Rosenthal explained that in healthcare it’s quite typical that doctors work through surrogates, whether it’s a nurse, nurse practitioner, a trainee like a resident or fellow or even administrative staff. And if a doctor directs that surrogate to order an imaging exam and it comes back with a decision support notification that it is low-yield, surrogates “have two choices,” according to Rosenthal. “They can ignore it and do as they were told, or they can stop in their tracks and go back and speak to the doctor who gave the order.

“Given the hierarchical nature of medicine, it’s unlikely they will do the latter,’’ he added. “That’s why we commented in the article that there is a relationship between this barrier to ordering and hands-on-ordering by the responsible physician. We made it a little bit harder for surrogates to place orders.”

According to Rosenthal, it’s quite clear the policy absolutely had a measureable impact. “The utility scores improved and we made it more desirable for the physicians to order the exams themselves and not use a surrogate. Both of those things happened,” he said.

But questions remain, he said. Did the number of low-yield exams ordered decrease because doctors learned something and they decided they needed to order more appropriate tests? Did it go down because doctors realized they had to provide more information in order to get an appropriate score? Or did it decline because doctors “gamed” the system. “These are extremely important and difficult questions,” said Rosenthal. “My opinion is that a little of each happens.”

Rosenthal added that the study suggests there is something to be learned about how small changes can impact the way EMRs work. “There has been this big national rush to EMR,” he pointed out, “and there are lots of little details about the design of an EMR that can significantly change the way it functions. But there’s not enough national experience with them to predict which ones will be successful.

“We’ve been in this game longer than anyone,” he said, “and it always surprises me to see the side effects of the changes we made. In this case the inappropriate use value went down.”