Jul 07 2009

Data Dose

As agencies amass evermore health data, the quest to optimize file access and management for users intensifies.

Like healthcare organizations everywhere dealing with increasing volumes of data, the Defense Health Information Management System has begun revamping its storage infrastructure to handle what can only be called explosive growth.

The military agency, which oversees patient records for more than 9 million beneficiaries (including active-duty military personnel, eligible family members and veterans), logs 120,000 patient encounters per day, all of which are stored in its clinical data repository. That DHIMS database currently holds 56 terabytes of healthcare data and is growing at approximately 1.5TB per month, says Capt. (select) Michael Weiner, DHIMS’ chief medical officer and acting deputy program manager. And that doesn’t even include imaging data — dozens of terabytes more that currently aren’t integrated into the system, he adds.

Healthcare organizations are at the forefront of the information storage charge. They need a way to house everything from patient records and lab results to medical images. Not only must the technology be scalable — it also must provide secure and easy access to users without being a budget buster. And federal agencies that handle healthcare data have a special challenge, simply by virtue of the staggering amount of data they must store, notes Brian Babineau, a senior analyst at Enterprise Strategy Group of Milford, Mass.

The government’s organizations are vast and serve millions of patients, so moving to more efficient technology is paramount, Babineau says. On the front end, that means migrating from the typical client/server paradigm to environments that employ virtualization and thin-client technology. A desktop virtualization technology such as that offered by VMware or Citrix can support a data repository that can back itself up continuously and be accessed centrally by widely dispersed users.

On the back end, it means moving from siloed storage deployments to a centralized backup system with virtualized servers and shared pools of resources. For example, a healthcare organization might use a blade center chassis, such as IBM Blade Center, to run its application environment. The blade system can let a data center add more processing power as needed without taking up a lot of space in a data center, making an agency’s IT operation more dynamic and adaptable, Babineau says. The data center probably also would employ Fibre Channel switches, such as Cisco MDS 9000 Series Multilayer Switches, to connect servers to a centralized storage device.

For long-term data retention, a system such as EMC Centera — a magnetic hard-disk-based data storage device — would be ideal, he says. It works well with the picture archiving and communications systems technology that medical facilities use to electronically store and display digital images because it can identify and store PACS metadata as well as the actual files, Babineau explains.

In many ways, virtualization is the key to managing the vast storage needs on both the front and back end of a healthcare infrastructure, says Steve Picot, manager of the federal data team for Cisco Systems. VMware Infrastructure 3, for instance, allows for end-to-end, self-adjusting optimization across a virtualized environment.

“The old days of having a software version on every desk and a traditional client/server scenario is being replaced by a virtual desktop where you centrally manage a pool of virtual desktop images,” Picot says. “If healthcare organizations are going in that direction, they are probably also going to explore the idea of virtualized servers on the back end. In fact, you’d be hard-pressed to find an IT organization that isn’t moving toward virtual computing in some way.”

A deeper look at how DHIMS and the Social Security Administration are following these trends illustrates the types of storage initiatives that agencies are taking to improve the ability of users to access ever more healthcare data electronically and quickly.

Taking a Virtual Step Forward

DHIMS has a clear plan for moving forward, and modernization tops its agenda, Weiner says. Today, doctors and other medical professionals access DHIMS patient records through a traditional client/server environment. Each electronic health record (EHR) is stowed in a secure data repository in Alabama that houses a 56TB Oracle database. DOD’s EHR system is known as AHLTA (which once stood for Armed Forces Health Longitudinal Technology Application but today is referred to only by the acronym).  A storage area network consisting of a Hitachi Data Systems storage array and redundant Brocade fiber-attached SAN switches serves up the AHLTA records, with processing managed by a 128-processor HP 9000 Superdome server. For redundancy, DHIMS uses Oracle Data Guard.

“Hardware spending will center on the massive data storage requirements that are overwhelming hospitals. The stated goal that ‘every medical record is digital within five years’ will continue to be a focus.”

SOURCE: Gartner, April 2009

DOD medical images are a different story. The first phase of the Defense Department’s Healthcare Artifact Imaging Management Solution will be a comprehensive image repository that DOD expects to take live this fall. HAIMS will go a long way toward automating the storage of images, Weiner says. It will consist of a federated architecture, meaning that images will be stored throughout the world but that access will be centralized and will appear to clinicians as if the files are located in a single repository.

Although the images will be stowed separately from the AHLTA files, the combined information will form each service member’s electronic health record, Weiner says. Clinicians pulling up a patient’s health information will see a full record, with data and images linked. Eventually, HAIMS will also include patients’ clinical encounters with doctors. The system will generate images using data tags that will associate them with the appropriate clinical encounters and any ensuing medical reports, forming complete EHRs, Weiner explained.

Portability Goes Thin

The Social Security Administration, another user of medical records, faces different storage challenges but is attacking them with similar gusto. SSA is in the midst of a long transition from paper and image-captured records to fully digitized health files. SSA provides no point-of-care health services, but health records remain crucial to one of its missions: evaluating the eligibility of taxpayers who file disability benefits claims.

The agency has made significant progress in automating this mountain of information, says Tom Grzymski, SSA associate commissioner for enterprise support, architecture and engineering. State agencies used to compile paper files for each disability claim, which would be shipped from one place to another until a claim had been approved or denied. But starting in 2002, SSA made a big push to move to electronic files, and it began imaging paper documents in 2004. Currently, SSA has stored 400 million documents that take up 100TB of disk storage space.

Most of the data resides in a DB2 database on three IBM z10 2097-E26 mainframes at an SSA data center. SSA uses IBM Content Manager for imaging, digital asset management, web content management and content integration. Based on these files, SSA creates validation materials and medical evidence files that it stows in an EMC disk storage subsystem, which claims adjusters access via thin clients — the newest model is the HP dc7900 — using web browsers.

But even after digitizing all paper records, SSA will still face user support challenges.

“Today, our systems are geared toward having a disability claims examiner or doctor look at images of the medical evidence. We want to be able to take eXtensible Markup Language documents that we get from the agencies and store them in that format because it’s electronically readable. But our business process today is such that doctors and claims examiners mostly expect an imaged format,” Grzymski says. “So for quite some time, we’ll be in a mixed mode where there are both images and structured data that are stored and accessed.”