But even hard stops are not as successful as one might think, Dr. Jackups said, thanks to workarounds.
For magnesium ordering, he and his colleagues considered a soft stop at time of order, but “we were afraid it would be too likely to be bypassed.”
An order question—which Dr. Jackups has found to be useful for many targets—requires an indication but can be bypassed easily too: It’s easy to select an indication and move on. “And we know what the indication is—it’s to look for when to supplement.” And there’s too much institutional inertia, he said, for the noninterruptive intervention to work.
He and his colleagues opted for a modified version of the soft stop, meaning instead of the soft stop firing at the time of order, they decided to fire it after they had proved the patient has had normal magnesiums. Thus, the alert, when it comes up, reads: “Daily Mg levels are ordered on this patient. As the last two Mg levels have been within normal range, please consider discontinuing this order if clinically appropriate.” The dates, times, and results of the prior orders are displayed, and the default is to discontinue.
“So if you are not paying attention and click ‘accept’ and move on, you’ve canceled the test.” But no one is being tricked, Dr. Jackups pointed out. “It’s up there. You had a chance to read it. But this requires you to interact.” It forces the orderer from what he describes as (with credit to the late Daniel Kahneman, in his 2011 book Thinking, Fast and Slow) the automatic brain, in which real-world, imperfect decisions are made with limited information and time, to the reflective brain, which makes rational, analytical decisions based on careful considerations. “You have to stop if you want to keep going and get that magnesium, or you can just click ‘accept’ and stay in your automatic brain, and we’ll do the work for you,” he explained.
That default, sometimes called a default heuristic, he said, is a powerful tool in clinical decision support as long as the laboratory is upfront about it and “not too sneaky.”
When to show the alert is one of five “rights” to consider when applying clinical decision support, Dr. Jackups said: the right information, to the right person, in the right intervention format, through the right channel (orders activity), and at the right time in the workflow.
For magnesium, the provider has already ordered the daily test because it’s a one-time order that continues to generate tests daily. Thus, with this alert, the intervention can’t take place at the time of order.
In such situations, it is common to fire the alert when the particular patient’s chart is opened. “We hate chart open alerts,” Dr. Jackups said. It’s the wrong time to fire, he explains, because “who knows why I’m going into the chart. I may just need to get some important information, read some notes, write notes. There’s a lot that could go on, and if I get that alert, I am immediately going to ignore it, so if we did this as a chart open, people will definitely not pay attention.” They’ll click through it all and the magnesiums will be canceled, and “they did not get a fair chance to think about it. To me, that would be a trick.”
Firing the alert during the orders activity is what they chose to do. In opening the orders activity, he said, “they [providers] have declared they want to order something on this patient, so they’re at least thinking about orders. This is as close as we can get to the right time in their workflow to make them think about canceling the magnesiums.” And the laboratory has information to show them at this point: the prior normal magnesium results. “That can be powerful,” Dr. Jackups said.
It also gets to the right person, he noted, and that is a physician or other provider who is authorized to cancel future orders because they opened the orders activity. “If they feel they are not authorized to do it, they can choose to keep the test going.”
Before going live with an alert, he advises, it should be run silently to see how often it fires and whether it fires when it should, and it should be done with sufficient lead time to evaluate it.
When they went live with their alert in March 2025, there was more overriding of the alert than not but still many who opted to discontinue. And each month the numbers of overrides declined slightly. “I see this pattern a lot with soft stop alerts,” he said. “Part of it is simply that people recognize it’s going to happen and may actually be changing their ordering behavior”—in this case doing fewer daily magnesiums.