In recent years, hospital information technology departments have assumed more responsibility for picture archiving and communications systems (PACS). This has increased demand for best-of-breed systems incorporating components from several vendors, and PACS purchasers have begun to require the ability to integrate third-party elements into their systems. The turnkey system from a single vendor was once the most common PACS, but advances in technology have made it easier for facilities to assemble systems that meet their budgetary and operational needs.

Vendors can be distinguished from one another based on the design of the servers, archives, and interfaces that they offer, but the degree to which they can support sound disaster-recovery plans and meet data-migration needs should also be considered. All of these elements will affect not only the initial PACS acquisition cost, but the long-term cost of system ownership.

SERVERS AND ARCHIVES

Servers are not interchangeable; they differ in hardware design, storage type, and database-management method. In turn, these variations affect both performance and cost. Server prices may increase as much as 20% beyond the amount initially quoted as network cards, redundant power supplies, and similar features are added to ensure high system availability. The overall PACS server price may nearly double through duplication of the server and database, but this fault tolerance is necessary to ensure uninterrupted availability of data. Automated failover, in which internal monitoring detects degraded performance and switches to the duplicate server, is highly recommended (but not available from every vendor). If manual failover is necessary, the value of fault tolerance is significantly diminished, since an extended period of poor performance could precede the decision to switch to the duplicate server manually.

Many vendors are unwilling to guarantee PACS performance unless high availability, fault tolerance, and a hardware service contract are in place. While high availability and fault tolerance are mandatory for the best PACS performance, the hardware service contract adds unnecessary cost because hardware is now so reliable that only minimal service is ever required. For most PACS, the only service contract needed is for software (covering updates, upgrades, and remote diagnostics).

Standards-based database management is vitally important. In addition, the PACS must have the capacity to accommodate the facility’s projected volume (which is typically underestimated 10% to 50% by hospitals). Vendors may either underestimate capacity needs by accepting the hospital’s incorrect estimate or overestimate volume needs (and increase system costs) by presuming that the largest historical volume increases are typical everywhere. If errors in both directions are to be avoided, more precise volume estimates are called for, and they should take into account procedural changes and increases in the number of images generated per examination, not simply projected imaging volume.

PACS purchasers should be aware that they can become committed to one vendor’s system if the vendor uses a proprietary form of data compression, as nearly half of PACS vendors do. Proprietary compression compels one to use proprietary decompression for data export, and this can be true even if a Digital Imaging and Communications in Medicine (DICOM) format is used. Alternatives that allow more flexibility are nonproprietary lossless compression and standards-based lossy compression. If possible, compression should be avoided completely. If compression must be used, the vendor should clarify whether the data-export strategy will lock in the purchaser to future use of that vendor’s products.

PACS may employ direct attached storage (DAS), network attached storage (NAS), or a storage-area network (SAN). Overall costs of ownership for on-site archival seem to be lowest for SANs. NAS is the second choice, but is more costly over the long term. DAS is initially less expensive to acquire, but the savings are eliminated by the higher cost of replicating data for this storage method. Enterprise-wide archiving calls for SAN as its storage method of choice, particularly because SAN costs continue to decrease.

Servers, archives, and all other PACS elements are associated with costs that are often overlooked in calculating cost of ownership. These include personnel costs for system and network maintenance and administration, along with support costs for software, hardware, and the network. Even within the 3 to 5 years typically allowed for return on investment (ROI), a PACS may cost twice as much as its initial acquisition price because of these added expenses. If the cost of ownership is not adequately predicted, additional funding may have to be found later, possibly jeopardizing the integrity of the PACS project or of other vital activities of the institution. Full ROI will be achieved even later if space, electrical, cooling, or other renovations are needed to house PACS components. Security and privacy measures will also add to costs, and data migration and disaster recovery must be part of the expected total cost of PACS ownership.

Vendors usually offer a broad range of archiving options for PACS, but these, like servers, should not be thought of as interchangeable. Roughly four of every five PACS installations use internally managed archives, but storage service providers (SSPs) are providing the rest with remotely managed, off-site archives. These are typically paid for on a per-study or per-megabyte scale, and can be particularly useful for institutions that lack the internal expertise, staff, or initial funding to manage their PACS archives well themselves. This option can also help facilities avoid the obsolescent archiving technologies offered by some PACS vendors.

SAN archive costs have dropped to less than $20,000 for the first terabyte of storage and $8,000 or less for each additional terabyte. Deep archives on digital video discs (DVDs) are also being considered for PACS because they can cost only slightly more than a dollar per gigabyte of storage. DVDs, now used for archiving in roughly 10% of PACS, can be ideal for moderate-to-high storage volumes because the jukeboxes that use them now cost less than $20,000. Nearly all PACS vendors also offer DVDs for online backup. Excluding those PACS that have no true archive per se and instead rely on mirrored spinning disks, DVD archives account for an estimated 5% to 10% of all deployed PACS archives; tape archives of several kinds are used in approximately 15% to 20%; SAN archives account for 65-70%; and another 5% rely on other archice solutions.

The difference between direct market prices and PACS vendor charges for archives can be enormous; for example, a 5-terabyte SAN and tape backup can cost as much as $300,000 and an 80-slot jukebox, $75,000 or more. This is not merely because archiving produces profit for vendors. Pricing takes connectivity, throughput, expansion capability, back-end server architecture, and system design into account, and some vendors also add archive-management software or dedicated servers to their archive costs. If a SAN is used, its software, database-management, and archive-manager costs may be part of its vendor price. Some vendors, however, simply charge a fee covering software and systems integration for a SAN that the facility purchases on its own.

DATA MIGRATION AND DISASTER RECOVERY

Despite the adoption of DICOM standards, many PACS archives still employ proprietary software, hardware, and/or media. Near-line storage and deep archives are available from all vendors, but data-migration and disaster-recovery plans are more rarely offered (although they are of critical importance). It is not uncommon to replace PACS hardware every 4 to 5 years and to upgrade software each year. Data-migration planning is necessary because the institution may need to replace its PACS with a system from a different vendor. Differences between the old and new systems can make data migration very slow and expensive. The facility may have to maintain two archives, including expensive service contracts with two vendors, if it chooses to migrate only the database and leave the older archive in place. Space for old and new archives could be difficult to find, and it may become increasingly hard, without vendor support, to maintain the old archive indefinitely (but the increasing inclusion in PACS of digital mammograms, pediatric examinations, and oncology studies makes this necessary).

An SSP can be used to prevent data-migration problems because it becomes the responsibility of the SSP to manage both the data and the archiving technology; the SSP should guarantee long-term access to the archive and the security of the data. The institution never needs to replace the archive, since it is maintained off-site by the SSP, so data migration is not needed.

A disaster-recovery plan is necessary for every PACS, and on-site backup is not an adequate solution. In case of disaster, both the PACS and its backup are obviously likely to be damaged or destroyed if they are kept in the same place. Ideal disaster recovery calls for two redundant copies of the images and the PACS database to be maintained in separate locations. In the event of catastrophe, these two backups should be made available only to previously authorized personnel; at all other times, access should be restricted.

CONCLUSION

Regardless of the PACS components and implementation plans chosen by a facility, it is not enough to analyze the initial cost of PACS. The cost of acquisition matters less than the overall, long-term cost of PACS ownership. In turn, the cost of ownership must be weighed against the strategic and other business benefits of PACS to the enterprise if the real value of the technology is to be appreciated fully.

Michael J. Cannavo is president, Image Management Consultants, Winter Springs, FL.