Improving throughput in radiology relies on process, not people
March 1, 2019
CHICAGO – There was plenty of talk at the annual Radiological Society of North America (RSNA) conference about the growing importance of business intelligence and analytics (BIA) as technologies that can solve problems in the here and now.
Unfortunately, on this front, healthcare may be well behind the times. “We as an industry are laggards,” commented Dr. Paul Chang, professor of radiology at the University of Chicago. “We’re still talking about dashboards and scorecards, as if it’s the 1990s.”
Often the terms scorecard and dashboard are used interchangeably in healthcare, which they’re not, he asserted. A dashboard gives real-time information, he said, while a scorecard is retrospective.
He demonstrated this with the example of the dash of a car. “When I’m speeding on the road, the dashboard tells me how fast I’m going, in real-time,” said Dr. Chang. “When I get the speeding ticket, that’s a scorecard. It’s retrospective, and tells me how fast I was going – but in the past.”
Dr. Chang, at an educational session, averred that other industries have gone way beyond dashboards. “What do other industries do? They’re not using scorecards or dashboards, because both are used by humans. Many humans won’t react to them.”
He joked that as a tenured professor at the University of Chicago, he could be given a scorecard or dashboard showing that his performance was poor. “Here’s your scorecard Chang – you suck – try to do better.”
But that doesn’t mean Dr. Chang will do better. After all, he is a full professor with tenure. Why bother?
What other industries are doing, Dr. Chang said, is building improvements right into the workflow, so that people don’t have to make decisions about doing the right thing.
In healthcare there are some successes on this front.
As examples, he cited PACS, which can automatically bring up a patient’s past exams and related medical documents when a radiologist goes through a worklist. These charts and images provide more context, enabling the radiologist to make a more informed analysis.
Another example is the “correct patient” in the IHE scheduled workflow profile, which uses IT to automatically ensure you’re dealing with the right patient.
Of course, closed loop barcoding and medication management, for institutions that are using IT on an enterprise-wide basis, could be cited as another example of a process that ensures “best practices” are carried out.
What Dr. Chang seemed to be saying is, the more you can automate various functions, the less chance there is for human error to occur. “Humans are terrible at remembering to do things,” he said. They are also not so good at doing the right thing, he opined.
And sometimes, they’re just wrong. That’s why peer review in radiology can be a bad idea, said Dr. Chang. “What if your peers are wrong, too? Then you’re reinforcing errors.” He noted that AI, on the other hand, could be wonderful for peer review, as it could consolidate the opinions of hundreds of radiologists to arrive at correct readings of multiple conditions.
Radiologists could then learn from these automated solutions, which would provide the best evidence.
Today, when building improved workflows, Dr. Chang said, you must first decide what your goals are. You must then figure out what the key performance indicators (KPIs) are to achieve these goals. The KPIs should then be measured, ideally in real-time.
Determining KPIs, however, is again something that healthcare is terrible at, said Dr. Chang. But he asserted that consultants are very good at it, and that bringing in consultants for this job is worth the investment.
Trouble arises again when it comes time to extract the data you need, as the data could be housed in six or seven different production systems. There again, investments must be made into integration engines and interfaces. And, of course, the data itself could be wonky. “The data you extract is sometimes not trustworthy,” said Dr. Chang. “Too many times, people grab every data element and use it.”
But despite these challenges, creating business intelligence and analytics systems is still possible. It helps to think about it in a new way, though, said Dr. Chang.
“BIA isn’t something you buy, it’s something you do,” he said.
He noted that at his own institution, the University of Chicago Department of Radiology, doctors and residents are encouraged to spot problems. And when they do it, they’re urged to let the IT department know.
Indeed, IT is seen as the solution, rather than additional staff. “We don’t throw additional FTEs in to solve problems,” he said. As he mentioned, adding staff can compound problems, as people tend to make mistakes, and more people can make more errors.
Instead, the goal is to re-design processes and to automate them, as much as possible.
On a practical note, Dr. Chang advised his audience that when pitching management for funding to improve processes, never position them as “quality” projects. That’s because quality is seen as a “floormat”.
And floormats, when you buy a car, he explained, are expected to be free. “When I pitched a project to the C-suite, and called it a quality project, they said great, go find a grant.” Things changed when he re-packaged it as a productivity effort. “When I called it a throughput project, I got the funding immediately,” said Dr. Chang. “It did double our throughput. And it also raised quality. But management thought the quality part should come free.”