HIMSS: Radiology workflow orchestration makes sweet music

2017 03 27 10 26 06 867 Orlando Florida 400

ORLANDO, FL - Workflow orchestration can maximize quality, improve service, and reduce costs for multisite radiology practices, while ensuring equitable workload for radiologists and enabling advanced data analysis, according to a presentation at the Healthcare Information and Management Systems Society (HIMSS) annual meeting.

Since adopting workflow orchestration in January 2018, Renaissance Imaging Medical Associates (RIMA), based in Los Angeles, has realized a number of important benefits, including improved adherence to service-level agreements (SLAs) and a higher percentage of cases read by subspecialist radiologists. It has also made it easier to perform the data analytics tasks needed to manage a large, consolidated radiology practice, according to Dr. Andrew Deutsch, chairman and CEO of the radiology group.

Deutsch, who is also a member of Carestream Health's medical advisory board, discussed his group's experience with workflow orchestration during a Thursday morning session and offered his recommendations for adopting the technology.

Addressing market dynamics

The recent consolidation among healthcare providers is being driven by several factors, including the need for operational efficiencies derived from economies of scale, as well as regulatory requirements that emphasize measuring and tying reimbursement to documented quality and value. These require the use of electronic health records and other IT technologies, and advances in real-time artificial intelligence (AI) and cloud computing can help consolidating sites improve operations and comply with regulations, another driver for consolidation among radiology groups, Deutsch said.

"[These requirements and technologies] are expensive and can really only be implemented in a cost-effective way by large organizations," he noted. "It's our thesis -- and not just ours -- that we're going to increasingly see, probably both in the [U.S.] and elsewhere, the creation of mega provider groups and mega radiology services groups to be able to take advantage of all of the important technology that's available."

Workflow orchestration is beneficial for consolidating providers because they render services over multiple sites. There is also growing demand from payors and clinicians to receive services from experts rather than generalists, Deutsch said. In radiology, that means more subspecialists.

"It is simply not economical to have all of these experts at one place," he said. "You need a way to get these images interpreted by your army of radiologists, and so you need to really start to concentrate on your workflow in a way that was never really done before."

A complex undertaking

Automation to manage the workflow across multiple radiology sites with the goal of maximizing productivity and quality -- and therefore value -- is a pretty complex undertaking, Deutsch said. The goal is to maximize the probability that imaging studies acquired at any of the sites will be read within the specified SLA requirements by a subspecialist located at or affiliated with the site or preferred by specific facilities or referring physicians, while evenly distributing the workload and complexity among the reading group.

At RIMA, radiology worklist orchestration is based on several key business principles:

  • Prioritization: The most urgent cases -- and the cases nearing the turnaround deadline under the SLA -- are prioritized.
  • Expertise: Cases within a radiologist's domain of expertise of subspecialty are selectively assigned to the best available radiologists.
  • Relationship management: Radiologists located at or affiliated with the site or preferred by specific facilities or referring physicians can be given priority for cases from those facilities or referrers.
  • No orphan cases: An acquired study will always appear in a worklist of a logged-in radiologist.
  • Workload balancing and equality: Maximize the probability of studies being fairly distributed among radiologists to minimize unbalanced workload.
  • No "cherry picking": Minimize reading out of order.
  • Promotion of group accountability: Each individual worklist includes valuable, real-time information about the reading performance of the relevant subspecialty and the group as a whole across all sites.

"A problem is that a lot of these [principles] butt up against each other," he said.

A "smart engine" that uses sophisticated algorithms and machine intelligence can help balance many of these competing requirements, he said. Performance tuning and real-time monitoring and analytics also help.

Adaptive workflow orchestration

RIMA, which has approximately 65 sites, 120 radiologists, and an annual volume of 1.5 million studies per year, has pursued an adaptive model of workflow orchestration. Rather than allowing each radiologist to pull studies from multiple worklists, this approach dynamically finds the most appropriate radiologist to read the study at any given time based on subspecialty, location, affiliation and special relationships, and the overall availability of resources. The entire group uses a single, short worklist, in which the content presented to the radiologists is dynamically tailored and optimized to their reading profiles, Deutsch said.

This simplifies setup and control of the worklist and provides a structured way to increase productivity, he said. The worklist is also divided into subworklists arranged by clinical urgency. Within a subworklist, studies are organized by the time remaining before the SLA deadline. A case may also be escalated and presented to an increasing number of available radiologists if it remains unread over time and is about to breach its SLA, according to Deutsch.

"This promotes group accountability," he said.

To prevent cherry picking, radiologists are presented with a limited number of the total cases available for reading. While other cases are accessible, radiologists must report their reason for not following the rules, he said. Dashboards also promote individual and group accountability, according to Deutsch. Each individual worklist includes real-time information about the reading performance of the relevant subspecialty and the group as a whole.

In addition, RIMA has adopted AI-based triage of certain types of cases; if an AI algorithm detects brain bleeding on a case, it's prioritized on the emergency department subworklist, he said.

Radiologists at RIMA log on to specific shifts, which are configured to allow different orchestration rules. Deutsch noted that in addition to workflow features, the orchestrator includes quality assurance functions, such as peer review, critical results notification, and discrepancy management, as well as communication tools, such as chat, collaboration and screen-sharing, and interfaces to secure email systems.

Improved results

Since RIMA implemented its workflow orchestrator in January 2018, the group has improved its adherence to SLA agreements by more than 20%, Deutsch said. On average, 84% of reports are also now signed by a subspecialized radiologist, he noted. In addition, the group has experienced improved workload balancing and significantly reduced cherry picking of studies. The orchestrator has also ensured that all sites receive a similar level of service.

Deutsch offers the following recommendations for implementing radiology workflow orchestration:

  1. Centralized workflow management seems to be the best approach.
  2. Appropriate overlay technology -- which will centrally index cases from multiple sites with normalized metadata and a master patient index -- can integrate with a wide variety of legacy PACS, RIS, and electronic health record systems likely to be installed at disparate sites. This will enable all radiologists to read and report as if they were reading from a single site using the same workstation and reporting system.
  3. Orchestration must balance multiple, often competing requirements.
  4. Orchestration should be user-friendly, easy to administer, and perceived as equitable by radiologists.
  5. The software should support ancillary services, such as critical results notification, peer review, collaboration, communication, and AI-based triaging.
  6. The software should provide robust real-time and long-term analytics.
  7. Implementation should be planned, and this process should include all major stakeholders. System configuration and user training are vital to a successful rollout.
  8. Academic centers, which must also provide measurable value and increase efficiency, will likewise benefit from orchestration of the resident-attending workflow.
Page 1 of 775
Next Page