Overheard at SIIM14

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - SIIM14

In May, informaticists, radiologists and other imaging IT stakeholders flocked to Long Beach, Calif., for the 2014 annual meeting of the Society for Imaging Informatics in Medicine (SIIM). The conference offers a chance for professionals to learn from one another and collaborate on the biggest issues in informatics.

Big data was the phrase on everybody’s lips. J. Raymond Geis, MD, chair of SIIM, set the tone with his introduction to the 2014 Dwyer Lecture that kicked off the meeting. “There’s a big new wave of images being generated in every clinical specialty,” he said. “From the first responders and EMTs up through primary care and the medical and surgical subspecialties.”

Geis also quipped that there are actually thousands of image viewers in each facility, if you count all the smartphones. This may feel like a bit of a nightmare for IT security, but it’s the reality.

Other presenters focused on the overwhelming tidal wave of data imaging continues to generate, and you can read more about what SIIM 2014 had to say about big data in this month’s cover story starting on page 6.

Big data also shared the SIIM headlines with discussions of data integration and bridging gaps across the enterprise—both in terms of breaking down imaging silos and bringing staff in different departments together—and clinical decision support and reporting. Read on to learn about a few other highlights Health Imaging saw trending at the meeting.

Decision support dilemmas

Integrating clinical decision support (CDS) in imaging is certainly more than a simple plug-and-play process. Thought has to be put into both when and how recommendations from a CDS system are presented, according to a presentation from Curtis P. Langlotz, MD, PhD, of the University of Pennsylvania Health System in Philadelphia, and Luciano M.S. Prevedello, MD, MPH, of The Ohio State University Wexner Medical Center.

The idea is to blend imaging appropriateness recommendations with other steps in the imaging process, such as safety checks, protocoling and scheduling. “In our view, this is all connected,” said Prevedello.

He offered the example of a patient referred for CT, but decision support recommends MRI would be better for the given indication. If processes are not in place to properly screen patients, an implant that’s not MRI compatible, for example, may not be discovered until the patient arrives for image acquisition. This then triggers another request, and can snag scheduling. A patient contraindicated for contrast is another example of when the optimal scan for an individual patient may not be what’s recommended by the CDS system.

CDS must be well integrated with existing systems and fine-tuned as decision support messages are displayed to users, Langlotz stressed. At his institution, initially the CDS was launched by the EMR upon order validation. The user was then prompted to go through the CDS process and provide information before a recommendation was given. Once the process was complete, the user was returned to the initial screen that prompted the start of the CDS process, which confused some users into unnecessarily activating CDS again, trapping them in an infinite loop at order entry.

CDS integration also should be mindful of workflow by not interrupting a user if the selection is deemed appropriate, advised Langlotz. Only if CDS recommends a different procedure should an alert be displayed.

Lastly, he suggests providers consider the tradeoffs between comprehensiveness of CDS and inclusion of only high-quality evidence. There are hundreds of procedures in a given order catalog, and not all have strong evidence-based guidelines. While some procedures will have appropriate use criteria from a given organization, it is up to a provider to determine whether the criteria is worthy of inclusion in the CDS process or merely the opinion of a small panel that may be beneficial to consider but is not based on extensive evidence.

Bringing radiology & IT together

Radiologists rely on quality IT professionals to keep systems running, but breakdowns in communication and misunderstandings between the two groups can lead to barriers and a less than cordial work environment.

Speaking at SIIM, Adam H. Kaye, MD, MBA, of the Hospital of the University of Pennsylvania, revealed the results of a survey of radiologists and IT staff at his institution that showed both groups had much to learn about each other.

The survey, completed by 95 radiologists and 14 IT staff members, indicated that most in both camps had few issues with professionalism during their encounters, but they weren’t in sync when it came to understanding each other’s workloads. For example, radiologists’ mean estimate of the number of workstations in the department was 87, while IT staff pegged the number at an average of 224. Radiologists guessed the number of calls to the overnight IT staff was 18, while IT staff had a mean estimate of six.

The largest gap came in the estimate of the number of studies read per day. Radiologists submitted a mean estimate of 1,729, while IT staff estimated the number to be more than 85,000. This wildly high average was inflated by three unreasonably high guesses—one IT staff member responded that 650,000 studies were read per day—and once these outliers were removed, the mean estimate for the IT staff fell below that of the radiologists to 1,118.

Communication is key, and Kaye explained that after the initial survey, four IT staff and four trainee radiologists volunteered to shadow a member of the other group for about an hour to learn the workflow. A post-survey completed by those who participated in the shadowing experiment indicated a very positive response. Participants commented that they had a better understanding of the issues their colleagues faced, and a number of them even said it helped them come up with future project ideas. One lingering IT issue, a problem with pre-fetching studies for cardiovascular imaging, was solved directly as a result of shadowing by the IT staff.

“We found that it opened up more possibilities for collaboration and innovation,” Kaye said. “It created a better environment for problem solving and it reinforced this idea that just sitting down with the other group that we’re all on the same team and understanding each other’s motives and needs is paramount to better teamwork,” said Kaye.

Understanding VNA needs

When considering investment in a vendor neutral archive (VNA), providers should have a clear understanding of what exactly they are trying to achieve, advised a panel of industry experts at SIIM.

Led by moderator Steven C. Horii, MD, of the University of Pennsylvania Medical Center, the panel tackled questions dealing with ensuring different systems can work together and involving an entire enterprise.

But before working out the specifics of integration, it’s important to have a clear goal. “You really need to be very certain what the problems are you’re trying to solve when you pick a VNA,” said Kevin Collins, vice president of research and development for Sectra North America. He added that it is not a good idea to get a VNA simply to avoid data migrations, as acquiring a VNA will likely force a migration from existing PACS and providers may end up looking for a new solution in the future anyway.

The problems VNAs solve well, however, are consolidating different department storage and archiving strategies, making sure they are all secure, said Collins. Because of this, it’s important to ask vendors if they will support object formats produced by cardiology, pathology and all the other departments in an enterprise that will need to be included.

The enterprise is where the complexity lies, noted Lenny J. Reznik, MBA, director of enterprise imaging and information solutions for Agfa HealthCare.Whether it’s wound care photos, portable ultrasound, DICOM or non-DICOM, it’s important to understand how the data are taken in and retrieved from the system. “Question is, are you really buying an IT solution or are you buying a clinical solution? Ask yourself that question before you begin this journey,” said Reznik.

When considering the whole enterprise, the archive must be seen as simply a piece of the puzzle with a lot of moving parts. “Simply having an archive without any intelligent logic, without any kind of information lifecycle management and audit capabilities and security won’t do much for you,” said Collins.

Reznik agreed that it’s important to think about how systems work together. For example, if a VNA from one vendor and a viewer from another aren’t in sync, it can provide quite a headache.

“VNA is only as neutral as the PACS lets it be,” added Fred M. Behlen, PhD, president of Laitek. “If the PACS is holding back any information or storing information in proprietary or private tags, you may not see that interoperability in another viewer, especially in areas like annotation.”

The discussion was billed as a comparison of options between a VNA and an archive neutral vendor, which Horii defined as different perspectives on the same issue. A VNA should be able to connect to different PACS without having to do a lot of migration, and offer the ability to switch vendors relatively painlessly again in the future. Archive neutral vendors, on the other hand, promise to be able to communicate with any storage, archive or enterprise image management system.

Whichever strategy a provider chooses, the panelists agreed on one important piece of advice. “Never give up ownership of your data,” said Behlen. If a provider is unable to take the data and easily move to a new system, any future changes to an image management strategy will be very difficult.

MIA clinical history

While missing information about a patient’s clinical history can hinder a radiologist’s interpretation, pertinent chronic conditions are often omitted from the clinical history provided by referring physicians. Moreover, if a condition is not mentioned in the clinical history from a referrer, it’s more likely to be omitted from the radiology report, said Piotr R. Obara, MD, of the University of Chicago, who presented at SIIM.

Results were based on a retrospective analysis of cases featuring a patient with one of seven conditions considered pertinent at the point of radiological interpretation: Crohn’s disease, HIV, lupus, cirrhosis, non-Hodgkin’s lymphoma, astrocytoma, and head and neck cancer. Twenty patients per condition were selected who had a baseline study confirming the condition, and then a follow up study of the same procedure a least 11 months after the baseline.

Looking at the follow up cases, 40 percent of the clinical histories provided by referrers did not mention the pertinent chronic conditon. In 62 percent of the cases in which the condition was not mentioned in the history, the radiologist also did not mention the condition in the radiology report, explained Obara.

In speculating as to why the referring clinician would neglect to include a relevant chronic condition in the patient history, Obara suggested it may be as simple as the ordering physician not knowing about the condition. A technical hurdle also could be to blame, with the computerized physician order entry system not allowing adequate flexibility to include information about chronic conditions that aren’t on a standard list of options.

Stat cases were more likely to omit a condition, indicating that a time crunch can affect whether a condition is mentioned or not. “Because of their busy schedules, clinicians might delegate some of the order entry duties to non-physicians such as nurses who might not know the patient as well or just might not be aware of how important specific information is to the radiologist,” said Obara.

Natural language processing technology could be used to summarize relevant data and compensate for an incomplete clinical history, according to Obara.