Choosing the right health care servers is no easy task

servers and client systems. Their application environments tend to be on the trailing edge of software or hardware technology, which seems paradoxical, given the very high level of science and engineering that drives health care in general. Nonetheless, a premium on risk avoidance by software vendors seems to have resulted in health care IT departments having to nurse along ancient (by IT standards) software far longer than typical businesses do.

A classic example of this is the persistence of MUMPS among two of the major electronic health record (EHR) vendors. Both use InterSystems Corp.’s Caché (the successor to MUMPS) databases in the same data structure they developed more than 20 years ago. They both use Microsoft Visual Basic Version 2 for the fat client software as their front end. Microsoft VB2 was released in 1992 -- not exactly leading edge!

Using decades-old software makes the hardware environment difficult to standardize and costly to establish. The health care CIO has to work backward from the application to determine exactly what hardware to buy. Just for the EHR part, we’re dealing with both the Caché server and a 2 GB VB2 fat client for the front end, which forces this sort of analysis:

• For the server side, the CIO needs to first know which OS is supported by this application, which in this case is Unix. But which Unix vendor and OS version does the application support?
• Which hardware vendors support that version of Unix?
• Which PC hardware for the fat client supports VB2?
• If there is none available, can I purchase the latest and greatest hardware PC and downgrade the OS to the version that supports VB2?

Ultimately, the CIO using traditional server and PC technology will wind up having very limited options. Virtualization has changed that picture considerably, however, for both servers and clients, but virtualization has its own limits and costs.

For servers, it’s useful to start from the premise that there is no such thing as a generic server for health care. The application, the clinical relevance and other factors will determine what kind of hardware and operating system you will want to standardize on. There are at least five main categories of servers you should consider: EHR, clinical database, Web, picture archive and communication system (PACS), and general business servers.

EHR and clinical database servers

Most health care organizations have a plethora of database servers of various sizes: large Hewlett-Packard Co. Itanium servers running Oracle Corp., Sybase Inc. or Caché EHR systems; large IBM Corp. 64-bit Intel Corp. servers running SQL Server Enterprise database farms for radiology imaging storage; small, 32-bit Dell Inc. servers running SQL or third-party database systems storing electrocardiogram images; U.S. Food and Drug Administration (FDA)-approved proprietary servers with databases of fetal monitoring tracks or ultrasound images; and Sun Microsystems Inc. Solaris servers with databases of medical libraries.

With such diversity, it can be daunting to manage these various hardware systems, as well as the maintenance for multiple OS vendors, database vendors and application vendors.

Web servers

Health care organizations’ Web servers are typically not standardized as well. Although many people believe that Microsoft IIS is the market leader in Web servers for most industries, the educational and research branches of health care institutions tend to migrate to Apache, the Web server market leader, with about 50% of the market. Tencent Inc.’s QQ runs a distant third in the Web server market.

For servers, it’s useful to start from the premise that there is no such thing as a generic server for health care.

You might think that a health care organization’s Web server would be the most important server in the data center. That server, however, plays a minor role compared with the intranet server farm.

A health care organization’s intranet server carries the brunt of the day-to-day workload. The intranet server farm will provide the staff of a health care organization its clinical scorecards, document management utilities, clinical forms, training services including video training, clinical and pharmaceutical libraries, emergency readiness plans, and flow charts and procedures, among other things. In addition, the intranet server farm typically has the primary interface to clinical applications. What this all means is that a health care organization’s intranet farm may be the most complex computer environment for a health care CIO to manage.

PACS servers

The most rapidly growing server farms in health care are those for PACS servers. Although these systems at heart are built around a database, you should consider the servers separately from other database servers, primarily because most have proprietary databases. In addition, a PACS typically stores an image as a binary large object, or BLOB so these are not classic small-transaction databases.

Only since 2004 have digital imaging PACS received FDA approval. The growth in the number and size of PACS since then would be considered an explosion. PACS now cover a number of Digital Imaging and Communications in Medicine imaging methods.

Health care organizations now have PACS to store images for almost all those methods. Because of the clinical instruments used to capture images, most PACS are specific to a particular type of image. PACS that capture ultrasound images will not be like those that capture x-ray angiograms, for example. Good EHR systems will have interfaces to the various PACS so all of a patient’s medical imaging records can be viewed through a single source. I cannot imagine a physician having to log into five to 10 PACS to view all the images in a patient’s medical record.

General business servers

A health care organization has the same general server requirements as a business: financial servers for billing and accounts payable, database servers for inventory systems, payroll servers, and human resources servers.

Of all servers, these general business servers are the least costly and easiest to manage and maintain, because they aren’t required to be up and running 24/7. These servers also have pretty standard requirements for OS and hardware support, and are easy to replace and upgrade.

They are key servers for a health care organization, as for any other business, because they keep the bills going out and the payments coming into the company. You can have all the available state-of-the-art Web, PACS and EHR servers, but they will not run for long if your general business servers are not helping the health care organization process revenue.

Of course, each of the server types I have discussed has more detailed requirements for availability, storage and other key needs. I have created a matrix (see “Rating the Requirements for Health Care Servers”) that gives a good general summary of those requirements. From such a matrix I can determine the level of difficulty involved with managing the server and performing hardware and software maintenance for the database, applications and OS updates; and get a good estimate of the human resources required. From each of these cells I can assign a weight, then drill down to determine overall cost over the lifetime of the server.

Rating the Requirements for Health Care Servers

Al Gallant is the director of technical services at Dartmouth Hitchcock Medical Center in Lebanon, N.H. Let us know what you think about the story; email [email protected].