Physician wants and needs - as well as budgetary concerns (what else is new?) - often dictate how healthcare organizations structure their data centers and how many of the images are available instantly or even in just a few seconds or minutes. But technology is beginning to make the dilemma less important as image access gains speed.
Three factors are applying great pressure to healthcare organizations as they find ways of storing the data they are producing and gathering on patients: regulations, imaging modalities, namely multislice CT, that are producing larger and larger studies and thus data sets, and increasing physician needs and expectations to get images now. Though technology is a major source of the storage problem, it is obviously also essential to the solution.
For radiology departments in particular, a common model in data storage is to maintain about six months of studies as 'short-term' kept on speedy (but expensive) servers that can call them up in seconds. As data become older than six months, they fall off into the 'long-term' category and are often stored on less expensive technology such as a DVD jukebox, but even these images only take a few minutes to retrieve (although a few minutes can be too long depending on need). And there are ways to fiddle with this scenario so that the waiting period for long-term images is largely invisible.
Such is the case at AnMed Health Medical Center in Anderson, S.C., which is a private, not-for-profit 567-bed inpatient facility. AnMed Health North Campus, located 3 miles from the Medical Center, includes an outpatient center. Having moved to a GE Healthcare Centricity PACS about five years ago, the organization has had an evolving approach to image storage.
In the healthcare world, organizations have to be prepared to store images and supporting data of each patient for a number of years depending on legal requirements. That's a long time if you consider that some of their patients are children. This is driven by a slew of regulations and different authorities dictating how long data must be stored in electronic format.
And the images just keep getting bigger. The organization was the first in the state to convert to all digital imaging, including digital mammography, cardiology imaging, and multislice CT, which all together take up enormous space.
"In 1998, we probably had 20 to 25 servers in our data center. We now have more than 160 servers in our production data center supporting numerous clinical systems," says Marty Stewart, manager of production services. "It was part of that implementation through which we were introduced to the IBM ESS (enterprise storage server) Shark in 1998/99, which is a highly robust disk farm. We currently have two ESS's one located in our production data center and the other located in our disaster recovery center. We are running the IBM PPRC (peer-to-peer remote copy) software that copies transactions at the microcode level from one Shark to another Shark."
As more clinical systems were implemented, they stressed to the vendors that they install all of their applications and databases on the Shark at the main hospital, so the data are instantly replicated to the disaster recovery center located at the AnMed Health North Campus.
Around the same time, they implemented McKesson's Horizon Patient Folder as their electronic medical record (EMR) and this was being stored on several jukeboxes as well, but recently Information Services began archiving this data onto an IBM FastT Storage Server (IBM has since renamed this TotalStorage DS4800). The long term images from PACS were also stored on a jukebox. In 2004, with the jukebox rapidly reaching its capacity, they worked with GE to begin storing long-term images on an IBM FastT. Also, during this time they implemented dual writes for these images (one of a FastT at the main data center, and one to a FastT located in their disaster recovery data center).
All of AnMed's clinical systems connect to the ESS or FastT for their data and applications through their SAN fabric. The SAN "consists of an IBM 2109-M12 Director class switch which can service 128 fiber-attached devices. To connect the AnMed Health North Campus to the AnMed Health Medical Center, we use an ADVA DWDM (Dense Wave Division Mux) on each end of a three-mile dark fiber pair leased from our local phone company," says Tommy Sluder, manager of network & technical support services, part of AnMed's Information Services. "We have two IBM 2109-16 port SAN switches at the North Campus for continuation of the SAN."
In total, they have a capacity of 40 to 45 Terabytes (TB) of data between the FastT and ESS.
Short-term, long-term - finding the line
"What we try to do is keep six months of images on line, and all that means is that in a matter of seconds those images are available," says Sluder. "For the long-term, it takes minutes for data to be recovered, and the longest I've seen is about three minutes for recovery of previous studies, and that's when the system is heavy loaded."
Short-term storage at AnMed uses a smaller disk pool than long-term, and thus there are less data to look through and recall is faster. Facilities also have the option of cost-cutting with long-term storage by using slower-drives, which though reliable, just don't have the same turnaround.
There are certain helpful functionalities such as 'pre-fetching' rules offered by numerous PACS. These rules allow radiology departments to prepare if they know a patient is coming in for, say, a follow-up MRI, says Sluder. They can schedule that exam in radiology and because of the pre-fetch rules, all of the patients' previous studies have been brought back in and are immediately available. There's no delay in looking at the prior images at all. So, that's an example of short-term/long-term invisibility.
But the lines between categories could of course be redrawn at any time. The rules of the storage game are generally created by the workflow needs, and of course, cost.
"As we add more and more devices and the images become larger of course, it's going to take more disk space," says Stewart. "We're kind of at the mercy of the physicians and staff to determine our strategy for short term/long term storage. They may come up and say 'I want to have nine months online, I want to have a year online'. If there's valid reason for that then of course our storage increases. We feel six months is about where we need to be, but that could change."
AnMed is lucky in that no matter where data are designated to sit, the basic architecture they will need is already in place.
One of the biggest issues AnMed faces is to maintain an accurate sizing model to make sure they have appropriate storage available to meet the needs of increasing image sizes. This also leads to an issue in making sure appropriate funding is maintained within the organization, Sluder adds.
Long-term storage: Who needs it?
At the other edge of the spectrum is an organization that has freedom to create a data network from the ground up without any real regard to legacy systems that can bog you down and create the need for compromise. Such a situation is like winning the lottery, and in that case we have a lottery winner in Raymond F. DeCrescente Jr., chief technology officer of the Capital Region Orthopaedic Group (CROG) in Albany, N.Y.
When the privately-owned Capital Region Orthopaedic Group (CROG) was established in 2000, the only electronic data management component was a practice management system for billing, accounting and appointments. The system supported the 14 orthopaedic surgeons practicing at the time who were looking to move to digital and put in place an electronic medical record system. The change was needed, the group was growing fast.
Now they are one of the largest orthopaedic practices in the Northeast and draw 93,000 patients each year, seen by 22 orthopaedic surgeons at their main facility or at any of their four satellite offices, in addition to an MRI office with plans for a spine center underway. They are currently performing as many as 59,000 x-ray studies a year. Given the size and scope of the group, it is getting to be the size of a hospital, DeCrescente says.
Capital Region Orthopaedic installed a Merge FUSION PACS with Merge eFilm Viewer to manage the images coming from three dual-detector digital x-ray rooms, an MRI, a CR system for spine cases, and other systems. All of these modalities are generally sending 25 MB studies into the central PACS, says DeCrescente, adding "One of the things I was so surprised at is the storage needs of an orthopaedic practice with all of the digital x-rays and MRI. It's absolutely amazing how fast that grows."
Designing a single storage system for all data
DeCrescente sat down with all of the physicians in the group to get an understanding of what their needs were for image access.
"When you have 24 different physicians with all of the different specialties you are going to get a wide array of answers as to what they feel their long-term and short-term needs are," says DeCrescente. "We had a subset of doctors who wanted to see everything they've ever done. Because they feel if they are going to go from flat film x-rays, which are in a file room, to a digital format, they shouldn't have to go to any archive for anything."
So, DeCrescente set out to accomplish that. Upon reviewing a multitude of vendors that offered different parts of the overall architecture of the would-be data center, they decided to go with Hewlett Packard as the sole provider. DeCrescente wanted to avoid creating a hodge podge of different systems from different vendors. HP was able to hit each requirement of their hardware spec soup to nuts, and offered local service support in Albany to boot.
"They were able to provide us the blade servers and the backend network, they handled our back-up and they handled our SAN. So they ended up being our entire data center," says DeCrescente.
Some might assume that side-stepping the issue of short-term and long-term storage would be a financial nightmare, as the high-speed servers typically used for short-term storage (six months to a year) are very expensive. Yet, for organizations like CROG that are not encumbered by older systems, they have more freedom to shape their own destiny.
There are, of course, many good reasons why many organizations must deal with hodge podge, and that's because at many facilities systems are phased in piece by piece, says DeCrescente. He looked at doing a ground-up build of their storage as a real opportunity.
From the onset, the facility had 12 TB of "out of the box" capacity. They've now had the data center in use for about two full years and have used about 6 to 7 TB with some more room to grow. All the images and data are instantly available online with absolutely no waiting at all. Sure it is filling up, but they have no plans to implement any additional off-line storage systems. There are some pretty compelling reasons for this.
Firstly, drives and drive technology are coming down significantly in price, and you can get a lot more storage capacity for your buck, says DeCrescente. He added that for them, the cost of building up a long-term solution often based on DVD jukebox technology would be so expensive that it would diminish the ROI of the system, and it would be less convenient for physician users.
The key is to poise your organization for growth from the beginning, making it fairly simple to add more storage cabinets and other components as needed.
Policing what data the SAN are being used for also is crucial. At CROG, they keep a tight control of what data the SAN is being used for, says DeCrescente. Generally, it is used for material that is critical for patient information, which includes images and EMR data, and also business and financial files.
There are many options available for healthcare organizations confronting storage issues. Facilities with legacy systems should plot out ways to use the technologies they have most efficiently, but also have a clear understanding of physician needs and also future storage needs, and build from there. Savvy use of PACS tools, for instance, also can help ease archive headaches and make them more efficient from a workflow standpoint.
For organizations starting from scratch, the storage world is your oyster - provided the money is there and planning for growth is done well in advance. There need be no short-term/long-term division.