Moving data may take months, but it can be a smooth process with proper planning
Technology ages fast. According to Moore’s Law, the number of transistors on an integrated circuit doubles every 18 months. It is no surprise then that early PACS systems, which were first commercialized a little over a decade ago,1 are already being replaced with newer releases. In upgrading, a facility often switches its PACS vendor altogether.
Yet even with the DICOM backbone, such a venture is not easy. Migration of data between different systems can pose compatibility challenges resulting in large commitments of time, management, and resources.
“Some people have gone into a PACS purchase without making plans for data migration and are surprised by the requirements. The key message is that it is a part of your PACS system replacement, and it is not a trivial part,” said Fred M. Behlen, PhD, president of Migratek Data Migration Services, Homewood, Ill.
Raym Geis, MD, a radiologist at Poudre Valley Hospital in Fort Collins, Colo, concurred. “Everyone’s issues are different, but the bottom line is that it is trickier than it looks on the surface,” he said. Facilities about to undergo a migration to a new system should plan for the unexpected.
Think Long
“There is nothing typical about data migration projects. There are twists and turns to each and every one,” said Wayne DeJarnette, president and CEO of DeJarnette Research Systems, Towson, Md. The diversity in projects means a wide variety in project time, with completion times ranging from 1 month to a number of years.
Community Hospitals and Wellness Centers (CHWC) in Bryan, Ohio, represents the quick end of the spectrum. Its data migration was completed in 8 weeks. The speed, however, was enhanced by the fact that both the old and new PACS systems were from the same vendor. “I think that not switching vendors was a part of our success,” said Greg Slattery, CHWC’s vice president of information services.
Physicians Regional Medical Center (PRMC)/Pine Ridge in Naples, Fla, decided to migrate only 2 years’ worth of data when they switched PACS vendors. The team originally estimated 4 months, but the project took about 11 months.
Poudre Valley Hospital, Fort Collins, Colo, has experienced a longer time investment. The facility first signed its new PACS contract in January 2006; data migration began in August 2006. Kevin Kirsch, RTR, (R), (CT), Poudre’s PACS coordinator, expects to complete the electronic data migration in late November. Then, the team will attack those files that could not be moved; Kirsch estimated that could take anywhere from an additional 1 week to 6 months. “We’re at a point where we are migrating about 2,500 exams per day, but it’s a scary proposition to have an unknown amount of time to complete the project,” said Kirsch.
Behlen said this feeling is not unique, noting that most people are stunned at the length of time it takes. He noted that recent experiences suggest a straight DICOM data migration will take roughly one quarter to one third of the length of time it took to originally acquire the data. So if a facility wants to migrate 5 years’ worth of data, it should estimate a migration length of approximately 15 to 20 months. “If you want to migrate the data in a tenth of the time that it took to acquire it, you will need to find 10 times the capacity in your system to handle the data flow,” said Behlen.
Think Hard
It makes sense for facilities to begin their migration planning early?even as they begin to look for the new PACS system. “Migration always seems to be the last thing the customer thinks about, but it should in fact be one of the first. Customers should start talking about it with their vendors from the beginning,” advised DeJarnette. Earlier planning will result in a better experience and greater satisfaction.
Few PACS companies archive images using simple DICOM. Many alter the process in some proprietary way. “They may compress data with a proprietary algorithm or put certain types of flags on specific studies,” said Geis. The old and new PACS vendors will have to work together to successfully migrate the data in an efficient manner. And just because they have done so in the past, Geis also advised others not to assume it will work as easily for them.
“PACS archives are often institution-specific with their own idiosyncrasies, so the new vendor may still not know what is going on in the legacy system,” said Geis. Poudre Valley Hospital had visited another site that had completed a migration between the same two vendors, but later learned that the reference site did not store as many nuclear medicine studies, which proved to be a challenge during Poudre’s migration. “We weren’t diligent enough in our research,” said Geis.
That research should cover all of the options for a migration, including who will handle the work. The most obvious, and one of the most popular, is the destination PACS vendor. Many facilities will bundle migration into the purchase of the new system. It is possible, though less likely, that a facility will contract with the legacy PACS vendor to handle the migration. A third option is a third-party vendor, whether a migration provider or an IT services company.
A fourth option is to handle the data migration in-house. “For a small migration job, that sometimes makes sense,” said Behlen. However, he also warned that handling the project in-house can strain resources during a time when they are likely to be strained already. “In principle, it could be as simple as writing some scripts to do 100,000 DICOM move commands for 100,000 studies. In practice, though, those moves will not all work, and someone has to handle the exceptions. There is enough chaos with a new installation and migration that it is usually easier to have someone else handle it,” said Behlen.
Stay Involved
This does not mean, however, that responsibility for the project should be handed over as well. DeJarnette suggested an in-house liaison who stays involved in the process?”if not on a daily basis, certainly weekly,” he said.
Slattery assumed this role for CHWC. “My role was at a high level, making sure the vendor understood the goals of the migration and provided the reporting we sought,” said Slattery. CHWC requested weekly reports, summarizing where the migration was in the total picture as well as noting any studies that had not been able to be moved. This allowed them to recover several studies from caches that would have been lost had they waited until the end of the project to address them.
“The customer has to be engaged in the process, keeping the parties talking to one another and monitoring progress,” said Kirsch. And vendors agreed, citing greater success when the facility assigns someone to stay involved.
“Migrations where the customer is more involved tend to go more smoothly because the entity performing the migration is very much dependent on the PACS and network resources provided by the customer,” said DeJarnette.
Know Your Data
The customer also is very much involved in dictating which data must actually be moved. In some circumstances, it may be possible to leave the data on your old archive and query it when you need it, but this requires maintaining the legacy PACS. Often, one of the goals in upgrading a system is to dispose of the old one.
However, even if migrating most of the data, there still may be some files that can be purged from the system. “Vendors don’t necessarily have a good handle or algorithm to determine what might not even need to migrate, for instance, if the law requires data be kept for 7 years, is it valuable or necessary to move studies older than 8 years?” said Kirsch.
These requirements vary by state and institution, but the in-house liaison will be aware of the specific idiosyncrasies. Jackie Story, RT, radiology department director with PRMC/Pine Ridge, explained that they decided to move only 2 years of data as a practical measure. “Because of our history, we believe that most of the time, if a patient returns for a comparison, they do so within a 2-year period,” said Story.
Save the Data
Once a facility knows how much of its data needs to be moved, the next question is how much is good enough? It is very unlikely that a facility will be able to migrate 100% of its data, and there will be some studies that cannot be recovered. “You need to draw a line between data migration and data recovery,” said Behlen.
When PACS systems were first installed, the standard of practice was film, and a loss rate of 5% or more over 5 to 7 years was considered acceptable, recalled Behlen. A 5% digital loss rate, however, may be perceived as not very good. “There will be a certain amount of data that will not be worth moving over, but you won’t know how much until you begin, so there is a certain unknown amount that needs to be incorporated into the service agreement,” said Behlen.
Correct the Data
Facilities also should consider whether they wish to incorporate data cleanup. Sophisticated algorithms can perform corrections and cleanup of patient demographic data automatically, flagging instances where manual data correction is needed. “It can be an opportunity to correct data errors, such as small spelling mistakes and links to PACS images and RIS reports,” said Behlen.
In some cases, however, it is more than an opportunity; it is necessary. Different PACS vendors have a wide variety of data cleansing requirements. Some require very little data cleansing in order for prior studies from other legacy PACS systems to be useful to the end user, while others require more in-depth cleansing.
“Common DICOM fields that we perform data cleansing on are patient name, medical record number, accession number, date of birth, and even study description. Some corrections are just format changes, while others are based on cleansing the prior studies so they accurately reflect order information and/or patient information generally generated by a HIS or RIS,” said Brian Baker, service/migration manager of DeJarnette Research Systems.
Accept Some Loss
Even if a facility opts to proceed without data correction, it should at least perform data maintenance. “Staying with the maintenance of your old PACS is important all the way through migration. Often, there is a tendency to let things go, but it can have a huge impact on the data migration project,” said Behlen. He advised that retrieval errors be investigated and files rebuilt when necessary.
But even with that effort, some files will be lost. “You have to expect that you will not get 100% of files migrated,” said PRMC/Pine Ridge’s PACS administrator Luke Wademan, who estimates about a 95% rate for his project. During the first 6 months of the migration, 60 files had to be handled manually, with 10 unrecoverable, he said.
Slattery achieved better results with a 99.9% migration rate. “Of 225,875 exams, about 6 million axial images, there were only 25 studies that would not migrate, and they were all on one MOD [magnetic optic disk]. We sent it off to see if the data is recoverable,” said Slattery.
Geis reported more difficulties with his project. “Our nuclear medicine files were tagged by the legacy vendor, and we’ve been unable to strip those tags yet to make the file a true DICOM study,” said Geis. Even if these studies were to represent only 10% of the hospital’s volume, with a million studies to migrate, those 100,000 studies are not a trivial number. Geis expects to work with the vendors to determine how to strip the tag and make the nuclear medicine studies migratable.
Facilities are advised to work with their contractors to determine how to handle the exception data. “You may be able to get it on a disk,” said Behlen.
Find the Right Balance
Facilities also will want to work closely with the vendor to manage workflow. Achieving the right balance between migration and workflow needs may require some adjustments. Attributing too few resources to the migration will extend the amount of time it takes. Attributing too many can impact user workflow.
“Our first goal was to not lose data. Our second goal was to not interrupt the day-to-day operations during the migration?previous studies would still need to be accessed. Our third goal was to complete the migration as quickly and efficiently as possible,” said Slattery.
When CHWC first began migrating the data, the resources given to the project were too high. But the vendor decreased the allocated bandwidth during normal business hours and increased it at night. “During the first few weeks, we had to find the right balance,” said Slattery.
Kirsch reports similar challenges. “We did bulk migration in reverse chronological order at night from about 9 pm to 6 am, and during the day, we did relevant prior presets. It was a challenge to make sure it was adequately balanced,” said Kirsch.
Have a Plan B
Key to that balance is collaboration, not only to achieve balance but success. Kirsch collaborated with the new and old vendors to clarify everyone’s role. He obtained quotes based on total data volume and the work each vendor would handle. “Start early. Make sure there is a plan A and then test it. Then have a plan B and test that, too, since it’s very likely plan A will fail. We ended up with plan C,” said Kirsch. Apparently, Murphy’s Law is as good to keep in mind as Moore’s.
Renee DiIulio is a contributing writer for Medical Imaging. For more information, contact .
Reference
- Wiley G. The prophet motive: how PACS was developed and sold. Axis Imaging News. May 2005. Available at: www.imagingeconomics.com/issues/articles/2005-05_01.asp. Accessed March 30, 2007.
Migration Path
There are a number of different data migration techniques from which a facility can choose. ?These choices will determine the speed with which the migration will occur as well as how useful the data is once it has been migrated,? said Wayne DeJarnette, president and CEO, DeJarnette Research Systems, Towson, Md.
One technique is to do a simple DICOM migration from the old system to the new system. ?The speed with which that will progress depends to a very large extent on the old PACS,? said DeJarnette, noting that simultaneous clinical usage will slow down the process. Facilities may choose to complete this migration before the new system is actually installed, although this option may not always be possible.
A second technique is to perform a DICOM migration based on prefetch. ?You connect the migration engine to the active RIS, and as orders are entered, prefetch from the old PACS storage those studies that will act as priors. In the background, you continue to migrate off of the old PACS,? said DeJarnette. This, however, requires completely and immediately turning over the old PACS to the migration, allowing no clinical interference.
The third option is the medium migration. Known storage formats, such as tape, are brought in-house to the migration vendor. ?This is especially useful in instances where your long-term archive for the old PACS is tape or optical jukebox,? said DeJarnette. The method can speed things up tremendously, but the records will not be in the facility?s control or available for prefetch.
The fourth, and most expensive, methodology is to perform a non-DICOM migration, moving the raw database and image sets of the old archive. ?This is a much more difficult thing to do. It is generally labor-intensive and prohibitively expensive,? said DeJarnette.
However, it can be much more efficient. Fred M. Behlen, PhD, president of Migratek Data Migration Services, Homewood, Ill, noted that rapid migrations are possible with certain legacy PACS systems. ?In a rapid migration, you can suck data from the old system five to 20 times faster than a conventional migration, but it requires engineering for each type of legacy system,? said Behlen.
Facilities also could choose to perform some combination of these methods. One hybrid method keeps the old PACS system in use while feeding data to the new PACS. ?You start on day x and collect current data from day x forward and begin migrating from day x back. At some point, you reach a critical mass where you switch over the new system,? said DeJarnette.
?R. DiIulio