Years ago, hospitals upgrading to a new PACS and archiving system had some fairly straightforward decisions to make. Purchasers – for the most part, just the radiology department – considered things like workflow, support and service and, of course, price.

These days, with the rapid increase in hospital mergers, a focus on the EHR, a shift toward enterprise imaging and a look ahead to the promise of artificial intelligence (AI), providers have much more to consider than they did just five years ago.

Dr. Rasu Shrestha, chief innovation officer at University of Pittsburgh Medical Center, noted that 2020 was always thought of as “the future” and “the future is here.”

Providers need to think about the strategy around imaging from an enterprise imaging perspective, looking at imaging “as an asset and enabler of strategies throughout the health system,” Shrestha said.

As a result, radiology departments today rarely make this decision alone.

“Decision-making has gone to the C-suite, but it's important for radiology to have a seat at the table,” Shrestha said.

Learning to share
“Ten to 15 years ago, an organization would be looking for a PACS that met their own needs,” said Don Dennison, a consultant who specializes in imaging informatics. “The most common pattern now is a shared system where multiple organizations use a single instance of PACS.”

Four years ago, after Mount Sinai Medical Center merged with Continuum Health Partners, the system went through the process of switching PACS vendors. Continuum had a system from McKesson Radiology while Mount Sinai used one from GE Healthcare, which was ultimately awarded the new contract for the merged systems.

While the Continuum radiologists needed to get used to the new system, with its new user interface, it was important to move to one, said Dr. David Mendelson, associate chief medical information officer for the Mount Sinai Doctors Faculty Practice and the vice chair of radiology IT of the Mount Sinai Health System.

“In this age of mergers, harmonization is an important principle because it’s not feasible to support numerous different systems,” Mendelson said.

Each organization may have a range of services, from trauma to general radiology, that have different requirements in terms of workflow.

The PACS and vendor-neutral archive also need to be capable of handling multiple patient IDs, a change from when more hospitals were independent.

That leads to the question of whether to upgrade to an enterprise PACS – a system shared among multiple facilities and specialties, including cardiology, neurology, ophthalmology and even dermatology – a vendor-neutral archive, or both.

It is important for healthcare organizations to consider the big picture of image sharing and viewing, according to Michael Gray, a consultant specializing in the digital management and distribution of medical image data, and decide whether or not they are ready to migrate to an enterprise viewing and archiving system.

For Gray, the challenge is analogous to being able to see the “forest from the trees” in the sense that if you don’t see the big picture it’s hard to invest in the right components.

If deploying an enterprise imaging system, facilities also must assess whether or not the components that will be supplied by the EHR are acceptable, according to Dennison.

“What if the reading worklist is incapable of providing advanced features?” Dennison said. “It’s not so much a technical incompatibility. It’s more about the parts that radiology users will be using – does it work and does it work well? If not, how are we going to mitigate it? Are we going to use third-party software?”

For example, a community hospital may have more general radiologists, while a larger health organization will often employ more subspecialty radiologists that require a specialized workflow.

“Some organizations take a lot of time to define their worklists very specifically,” Dennison said. “Often, two or more rads will share a worklist. If the reading worklist logic isn’t very sophisticated, it requires a person, like a radiology coordinator, to manually assign cases. Radiology needs to be engaged in the decision about the reading worklist application that will be used, or risk getting whatever IT thinks is best.”

Dr. Eliot Siegel, chief of radiology and nuclear medicine for the Veterans Affairs Maryland Healthcare System, said that while it’s critical that IT is part of the team that makes the decision on the PACS and archive system, it’s equally important that radiologists, who have important training, be involved in decisions regarding medical imaging.

Radiologists need to consider computer monitors and brightness, and also how peer review – in which radiologists judge each other’s interpretations – is done.

“It’s imperative that a new PACS system be able to do that,” Siegel said.

This means that while radiology departments may take a back seat to IT, they still need to help navigate the decision.

“For many years, radiology would buy their own systems,” Dennison said. “Now, when components are supplied by the EHR, they’re not going to the vendor. I’ve seen radiologists who are a little too trusting that the right thing will happen, and it doesn’t. You’re relying on other groups for your success.”

However, when implementing an enterprise system, the special considerations don’t only extend to radiology.

“Radiology people don’t have a full perspective of the imaging needs of non-radiologists,” Gray said. “IT needs to think, ‘how are the new generation of EHR users going to visualize images?’”

Healthcare organizations also need to think about what kind of images may need to be securely viewed and archived, such as iPhone photos from dermatology or a burn unit, Gray said. They also need to think about their budget now and what they might lose by waiting to move to an enterprise system.

“Ask all the questions,” Gray said. “What do we want to put in this, aside from radiology pictures? … You need to understand the big picture and understand the timetable to achieving it,” Gray said. “Once you have decided that you understand that, then how long are you going to give yourself to get there?”

It’s also important for many other non-clinical teams to now be involved in a PACS purchase.

“There should be input from risk management, teams that look at communications, infrastructure,” Siegel said.

Facilities also need to consider future imaging needs when planning a PACS and VNA switch.

“If you have a 16-slice CT scanner and you know you’re going to 128, that means thousands of images,” said Cris Bennett, a PACS specialist with MD Buyline. “You really have to look ahead, budget wise.”

Planning the transition
With so many stakeholders involved, it is important to plan for a transition to a tightly integrated and/or single new system, which is likely to be more complicated than one hospital switching to a new PACS vendor.

After the merger four years ago and switch to the new PACS, the two New York City health systems are “still migrating the data,” said Mendelson of Mount Sinai.

Large health systems converting to a new PACS and VNA need to look at the “quality and consistency” of the data they’re migrating, Dennison said. Procedure descriptions can differ from hospital to hospital.

“Everyone’s calling the same exam by a different name,” Dennison said. “You have to convert all of the historical values, and you can have thousands of unique descriptions per site. Or you create a mapping, which is more complex. It has a big impact on productivity. A lot of organizations, in the rush to convert, have to spend a lot of time cleaning up after the migration to get their software features to work as they want them to.”

Dennison said there are a number of data migration service providers who can help with the data conversions.

“They have an understanding of it and they can give you guidance on the best techniques,” Dennison said.

Bennett said it is important to note that “there’s always loss of data” during a transition.

“There is just so much more data and more images” these days, Bennett said.

Attention to workflow
It is increasingly important for radiology departments to more broadly consider how PACS features fit into a radiology workflow.

“Think not just about bells, whistles and tools, but how to integrate more tightly into the overall workflow, adding value beyond a stand-alone application, such as a liver lesion management tool,” Shrestha said. “At the same time, they’re not really connected to the rest of the workflow.”

It’s also important to bring in additional context, called image-related clinical context, about patients from the EHR, Shrestha said.

“At UPMC, we’re working with GE to specifically bring in context,” Shrestha said.

Sometimes, integration into the EHR just isn’t in the budget. David Alexa, RIS/PACS administrator for the Dickinson County Healthcare System in Iron Mountain, Michigan, said that when upgrading to Carestream Health’s Clinical Collaboration Platform in October 2017, they weren’t able to add EHR integration.

Instead of viewing images directly in the EHR, in-network physicians log on to a different system to view the images after reading the radiologist’s report in the EHR. If patients are transferred to an out-of-network facility, images are shared using a virtual private network, a secure tunnel between the two sites that allows DICOM traffic.

“It’s more of a workflow issue,” Alexa said. “But we haven’t gotten any complaints because they didn’t have that functionality before.”

However, physicians in the Dickinson County Healthcare System do have the added functionality of being able to view images on their mobile devices with a zero-footprint viewer.

On the ground or in the cloud
Another big consideration when upgrading is whether to have on-site storage or use the cloud.

Michael Cannavo, an imaging IT consultant, said that while a lot of facilities are embracing software as a service so money can be taken from the operating budget instead of more limited capital budgets, he recommends that facilities don’t rely 100 percent on cloud storage.

“If the network connection to the cloud goes out for any reason, it’s like the engine being taken out of your car,” Cannavo said. “You need a hybrid system, with a small-scale server on-site.”

Hospitals, especially in rural areas, also need to factor in the availability of high-speed networks, in order to have a cloud-based system that runs on the same speeds as one on-site.

“Anything less and you may impact workflow,” Cannavo said.

Radiology departments may also want to invest in their own backup networks to prevent downtime, Siegel said.

“Downtime significantly impacts work in a radiology department,” Siegel said. “It’s important to have contingency plans to allow us to operate independently when we need to. … Before purchasing systems, look at that level of vulnerability. The hospital is so dependent on imaging. Having radiology at least be able to continue operations within the department itself is really critical.”

Alexa of the Dickinson County Healthcare System said that when upgrading, they went with a hosted system to ensure that exams loaded quickly.

“We feared if it was out in the cloud, our doctors might not be happy with the speed in which they were loading,” Alexa said.

Facilities also should realize that burning a CD may become extinct, as most new computers don’t come with CD drives.

“Image sharing is a big deal,” Cannavo said.

Making room for AI
While AI still seems like a tool of the future, forward-thinking imaging departments need to look ahead and consider how algorithms can be integrated into the radiology workflow.

“We’re already using AI-based capabilities around clinical decision support for lung nodules and liver lesion management,” Shrestha said. “It’s there, but there will be a lot more yet that will come down the pike. And AI, done right, will be much more tightly integrated into the very fabric of the clinical workflow.”

Siegel said there are a large number of innovative start-ups creating AI applications. While many of these AI vendors allow radiologists to interface with the applications via a web portal, it’s not the ideal solution for a radiologist who wants to use many applications.

“The PACS vendors themselves need to allow their users to interface with these AI systems,” Siegel said. “Although AI vendors need to have ability to interface with workstations, they shouldn’t have to come up with a complete set of interfaces for all types of PACS. I think the solution, moving forward, will be the ability for radiologists to pick best-of-breed AI applications that are part of their workflow, with the ability to be able to designate certain types of studies for certain types of AI applications.”

This also means that radiologists and healthcare providers will have even more to think about when it comes time to purchase a new system in the years to come.

“There’s no doubt that PACS is going to look different in five to 10 years than it does now,” Siegel said.