How to prevent cherry picking on radiology worklists

One of the problems in radiology today is the selective cherry picking of easier and more desirable cases from the DICOM worklists and leaving the more complicated studies for other radiologists to read. This is partly due to human nature, and partly because of the way radiologists are paid for their work based on relative value units (RVUs), which reward those who read more studies, regardless of complexity. 

This is a subject discussed over and over in sessions at radiology conferences, but new technology is now enabling artificial intelligence (AI) to sort through the radiology reading lists and distribute studies more fairly among radiologists. Workflow orchestration software is now offered by several PACS vendors, and some radiology groups have programmed their own software. Orchestration is becoming a key component of how practices are leveraging new technology to address the cherry picking problem. 

"Cherry picking is a big problem, and there is a productivity problem associated with it," explained Elizabeth Bergey, MD, a diagnostic radiologist at Quantum, chairman of Quantum’s Board of Directors, president of the physician group and CEO of the professional corporation, who discussed the problem in sessions at the Radiology Business Management Association (RBMA) 2022 meeting in April.

"Obviously, if you have someone cherry picking, that will artificially inflate their relative production and deflate contributions of other workforce members. That is also a big dissatisfier," Bergey said.

Bergey said cherry picking becomes a problem when you start to evaluate work loads to figure out RVUs used to pay radiologists. She is not a fan of the RVU system and prefers using time spent reading a case and looking at complexity factors of each study such as patient age, type of exam in calculating RVUs, rather than how many cases someone reads. She said these parameters are built into their workflow orchestration system to try and make things more fair and to discourage cherry picking.

Active and passive radiology cherry picking, garbage cans and stopper cases

One thing she has found is there are two types of radiologist cherry picking, both active and passive. Cherry picking actively is when you see someone going all over the list to pick the studies they want. Passive cherry pickers will just not open a case they don't want to read or get bogged down in.

"This can be upsetting because there might be a case at the top of the list that was not read. So obviously there is a problem with it, either it is really old patient disaster or they hover over it and see the referring physician is an oncologist and say, 'no, no, I don't want to touch that one,'" Bergey explained. "When you are monitoring the list for cherry picking, you will see no one will take that case." 

She called those wanted studies that no one wants to read "stopper cases," because it stops the normal flow of productive, orderly work on the reading list, and it acts as a bottleneck where radiologists slow down or look busy and wait until the stopper case disappears from the list so they don't get stuck with it. 

"You can watch the list and you can see the stopper case because it just sits there and finally one of the garbage cans pick it," she said. "Now what happens, the next five cases go flying off the list by different radiologists. They were sitting there waiting for that case to be uncovered. We know that this happens, but it is difficult to manage because it is hard to prove someone is sitting there waiting to uncover that case."

Cases that tend to sit around include non-contrast chest CTs, which are often nodule followups, and thyroid ultrasounds.

"All of us who have been around for a while know which cases are going to be stopper cases," Bergey said.

Use of workflow orchestration to automatically manage radiologist reading lists 

Over the past few years, several vendors have introduced workflow orchestration engines on their PACS and enterprise imaging systems to mitigate cherry picking. These systems also look at studies to determine which hospital they came from and the service level agreements (SLA) that defines how long radiologists have to turn around an exam. The AI in these systems also looks at factors such as the speciality of the radiologist, STAT reads and other factors to load balance the list of studies system presents to radiologists. These systems can help give a mix of complex and easy reads, and will present them in order so that SLAs can be met and not drop to the bottom of the reading list.

Bergey refers to the AI as the "brain" built into her practice's orchestration system. It looks at each radiologist's profile, noting their roles, specialties and preferential hospitals they do reads for. The AI also knows the speed of radiologist's ability to read, which is used to estimate read times and if they can take on a STAT read when other readers are busy. This real-time list manager also can show if there might be a bottleneck due to the number of exams and their complexity, which might require additional radiologists to come in early to help out.

"We get an order, our 'brain' then pulls in the study that goes with that order, then our unified PACS queues the archive through an algorithm to look for and pre-fetch relevant prior studies," Bergey said. The process has significantly improved workflow and saved time for radiologists.

What is the ROI for using radiology workflow orchestration?

The cost of an AI-enabled workflow orchestration system is not inexpensive, but Bergey said there is a big return on investment when you look at the time it saves radiologists and the ability to streamline reading lists, meet SLAs and determine when additional staff are needed or not.

"In this time of staffing shortages, we were able to drop our man-power down to its leanest while keeping our turn-around times optimized," Bergey explained. "It's a management and financial advantage to consolidate like this. It's not inexpensive to do this, but the cost of one radiologist is roughly half a million dollars. So if we save one, two, three, four, five radiologists, then you are really far advanced. Not to mention, you cannot even get radiologists anymore." 

She said radiologist job satisfaction has gone up with orchestration. This in turn is helping avoid burnout and the urge among radiologists to join the great resignation at a time when it is getting much harder to fill open positions. 

Orchestration systems also integrate analytics so the practice can perform apples-to-apples comparisons between radiologists, average read times, SLA performance, estimate on the complexity of cases being read, aid in quality assurance reviews and turn-around times based on STAT reads, and help identify the need for additional resources.

Counting time to read radiology studies rather than RVUs as a new measure

Some workflow orchestration systems use an AI algorithm to look at studies and determine the complexity. Bergey said the AI her radiology group uses can identify if a 69-year-old patients with numerous comorbidities will be a more complex case than a serial follow-up exam for a 30-year-old patient.

"You get credit in minutes rather than credit in work RVUs, because with RVUs there is no distinguishing easy or complex cases, and that is why you have cherry picking in the first place," she said. This helps out the radiologists who become the "garbage cans" who end up with all the unwanted studies that take longer to read. 

Orchestrated radiology workflow moves away from conventional reading lists 

"We are still thinking in terms of lists, when we should be thinking in terms of an entire environment, and we are thinking in terms silos because that is the way we had to work for so long," Bergey said. "So when talking to a vendor, make sure their system is used to pinging multiple archives between VNAs, PACS archives or wherever the images are stored."

She said a vendor's system also needs to be able to open exams regardless of their format or compression. Bergey said some vendors save things like mammograms in proprietary formats.

Another suggestion is to eliminate creating a worklist that the radiologist works off of during their shift and use a more dynamic system. She said the workflow orchestration system her practice uses sends studies that need to be read to specific radiologists throughout the day based on SLA turn-around-times, STAT reads, the radiologists availability and their expertise. 

"Our system constantly is looking at the work, and as things bubble up to the top the system sends it to the best available person," Bergey said. 

Watch the following related interviews with Bergey on her RBMA session discussion:   
   • VIDEO: Approaches to intelligent radiologist worklist orchestration
   • VIDEO: How to Prevent Radiologist Burnout


Related Radiology Workflow Orchestration and Cherry Picking Content:

Rand study suggests hospitals are exorbitantly overpaid by private insurers; AHA reacts

Healthcare reform: fodder for findings

Carrot Approach to Radiologist Compensation

This Is Enterprise Imaging

Intelerad Introduces InteleOne Maestro, its New Enterprise Workflow Orchestration Solution

Major healthcare providers in Israel choose Sectra’s digital pathology solution for full-scale primary diagnostics

Nuance AI Marketplace Accelerates AI Adoption for Radiologists to Improve Patient Outcomes and Reduce Burnout

Exclusive Webinar: 5 Strategies for Implementing a Successful Enterprise Imaging Strategy

GE Healthcare launches new program to simplify AI development, implementation

Trimed Popup
Trimed Popup