Michael J. Gray

As the technology of picture archiving and communications matures, a key question arises: Is a picture archiving and communications system (PACS) becoming more affordable? If, in saying affordable, we mean is the price of PACS implementation less expensive, the answer to the question is probably no. But if affordable is defined as easier to justify, the answer to the question is a definite yes. But the trick to cost-justifying PACS may be to think bigger.

The three words or concepts most frequently used when discussing the economics of PACS are: affordable, cost justification, and return on investment (ROI).

In most cases, affordable is associated with the ability to find the dollars to pay for PACS. The more interesting use of affordable is its variation: “Can we afford not to invest in PACS?” Sometime during the course of developing a PACS strategy, the steering committee should consider the consequences of not making the PACS investment. A careful look at procedure volumes, both historical and projected, will indicate whether the business of radiology is going up, going flat, or going down. It is especially important to determine the growth rate in the more revenue-rich modalities like CT and MR. It is also useful to separate the growth rates within the main department from those in the outpatient operations. Now pose the key question, “Are any of these growth rates being affected by local competition, or an inability to absorb additional growth, or (worse yet) an inability to provide the expected level of service to the referring physicians?” Any one of these answers would suggest that the hospital or health system cannot afford to delay the investment in PACS.

Another interesting concept in PACS planning is that of cost justification. The simple truth is this: PACS will cost a lot of money and winning support for the project may come down to justifying that cost. Success with cost justification depends on one’s skill with words, not spread sheets. If the business model suggests that the combination of development and operational costs for a PACS outweighs the projected savings and any additional revenue that might be generated by the PACS, the proponents of the project will have to justify that deficit. Perhaps the deficit can be justified by referencing numerous soft savings, or benefits that cannot be accounted for in the business model. If that tactic fails, there is always the fallback position, which is justifying the cost with the cannot-afford-to-delay-the-investment-in-PACS argument.

By far, the most successful way to gain approval for a PACS project is to present an accurate financial view of the PACS plan using the ROI tool. The objective of the ROI is to use spread sheets to argue that PACS can be operated at or above breakeven. Until recently, this no-nonsense look at the economics of PACS has been very discouraging. It has been very difficult to avoid operational deficits without bending the numbers, and the ROI that shows a significant deficit often leads to the end of the PACS project. Yet the ROI is still the most reliable way to gain support for a PACS. Instead of playing with the concept of affordable, or trying to find the words to justify the cost, take the time to figure out a way to build a positive ROI. It is not necessary to resort to bending numbers to create a positive ROI. What is required is to build a business model that uses PACS to create broader savings or new revenue that collectively exceeds the projected development and operational costs of the system.

It is easier to build a positive ROI today then ever before. The first trick is to take advantage of several recent technological advances that make it possible to lower the cost of some of the PACS components. The second trick is to build a bigger system, not a smaller system.

Definition of PACS

Figure. Diagram illustrating the major subsystems of the electronic information management system.

Today the term PACS is not very definitive. Think bigger. The important concept is that of electronic information management, in which information is images, text, and voice. In most cases, this concept covers more than most PACS. When all of the issues and problems to be solved are considered, the system grows to an enterprise scale, and that is bigger than what is traditionally thought of as PACS. The Figure illustrates the major subsystems of the electronic information management system (EIMS). In this view of the system, PACS is one of several component subsystems.

  • The Acquisition subsystem represents such diverse but related components as Digital Imaging and Communications in Medicine (DICOM) modality software, external DICOM interface boxes, computed radiography(CR) systems, digital radiography(DR) rooms or upgrades, and digital mammography systems.
  • The Archive subsystem represents the server or servers most directly responsible for managing the bulk of the image data, both online in the form of RAID and near-line in the form of a jukebox of archival media. While this component formerly was inseparable from the PACS and might still be included in the PACS, it is represented as a separate subsystem in the Figure, because it is now possible for it to be a separate subsystem.
  • The Diagnostic Displays used by the radiologists and the associated servers that support the display and work-flow features constitute the PACS subsystem. This subsystem is what is left of the traditional PACS.
  • The Voice Recognition Reporting system is also a part of the complete EIMS. This subsystem absolutely must be included, since a large percentage of the referring physicians are more interested in the timely delivery of the report than the availability of the images alone.
  • The Web Server subsystem, in conjunction with hundreds of PCs turned Web Viewer, is responsible for electronic image and information distribution across the enterprise. It too could be included in the archive or PACS subsystem, but it is represented as a separate subsystem because it is possible for it to be separate.
  • The HIS/RIS subsystem is also recognized as a separate piece of the EIMS. The hospital information system and/or the radiology information system is interfaced (at the least) or integrated (if possible) to the EIMS. Through DICOM Modality Work List, the HIS/RIS provides the DICOM file header with patient and study data. The Scheduling information triggers the pre-fetching of relevant prior studies. The Order information determines who will have access to the images and final report. The electronic medical record (EMR) or clinical data component of the HIS is destined to become the preferred point of access for all clinicians and caregivers to retrieve all kinds of clinical information on the patient including the radiology images and reports.

The following are examples of the many recent technology advances that have made it easier to build a positive business case for the EIMS.

Acquisition Subsystem

In my opinion four vendors have caught up with each other and created parity in their stand-alone DICOM interface boxes. This has effectively increased competitive pressure and the result can be seen in lower unit prices when the final contract is negotiated.

The emergence of direct or digital radiography and increasing parity among computed radiography vendors have led to some price reduction of high-volume CR units, and the creation of economical single plate readers. The combination of DR and CR required to take the radiography sections of the department digital costs less today than it did a year ago. Furthermore, the resulting reduction in radiographic rooms and technologist FTEs required to meet current and even projected procedure volumes is an obvious example of cost-avoidance that fits nicely in the benefits section of the business model.

But the biggest change in the Acquisition subsection is the arrival of digital mammography. Not only does this technology make the modality digital, it generates significant new revenue. Here is a clear example of thinking bigger. By including the cost of digital mammography in the EIMS model, it is acceptable to include the new revenue generated by this upgrade in service. The new revenue comes from two sources: increased procedure volume and increased technical fees for the procedure. It is realistic to project a growth in mammography procedure volumes simply because the procedure is digital, and a simple effort at marketing that fact should assure an increase in procedure volumes. Hand in hand with the ability to attract additional business is the ability to absorb that new business with existing staff and the new equipment. Digital mammography in a digital department is a very efficient combination. There is also additional revenue in the form of technical fees. A different set of CPT codes apply to digital mammography and most payors will negotiate a higher technical fee for the digital version of the mammography procedures.

Archive Subsystem

The emergence of the generic DICOM archive has had a significant impact on the architecture and cost of the EIMS. This is true whether the Archive subsystem is offered for sale or packaged in a fee-for-service program. This is true whether it is deployed on-site or the server farm is shared and housed at an off-site data center.

The creation of a generic DICOM archive has led to a reduction in the cost of archiving because these solutions are typically based on standard hardware and software platforms from the much larger informatics industry. The cost of disk drives used to build RAID configurations, storage area networks (SAN), and network attached storage (NAS) configurations has dropped by as much as 300% in the past 18 months. The price of high-end, generic storage products from vendors in the radiology market has also dropped significantly. Maybe the prices are a response to the recent business climate, but there is no indication that they will once again go up. Moore’s Law is clearly applicable: every 18 months chip capacity doubles. Consequently, the price of storage drops.

Many of the new archive vendors are actually in the software business and have programs that allow their archive clients to purchase the generic hardware and software directly from the recommended OEMs. In some cases, the archiving applications were developed for multiple platforms. Both of these facts have increased the level of competition between the OEMs that build the disk drives, jukeboxes, and archive subsystems, and this effectively reduces the cost of those components.

Another interesting aspect of this new archive paradigm is its dependence on the use of DICOM for communications between archive server, modality device, display station, or PACS server. Since the archive is generic, any of a number of PACS vendors could be chosen to provide the PACS subsystem (the diagnostic displays and associated servers). The choice of archive vendor, usually a first-year decision, does not automatically lock in the choice of PACS vendor. And the choice of a diagnostic display vendor in an early phase does not automatically lock in the choice of archive vendor. Whenever there is choice, there is competitive pricing. So selection of an EIMS architecture that is based on a generic archive is a proactive strategy that makes it possible to negotiate the absolute best price for pieces of the system that will be bought in years two, three, and beyond.

A final note on the archive subsystem relates to cardiology. It is already possible to include cardiology image sets in some of these generic archives. Although there is no such thing as sharing disk space, there is an economy of scale in the servers and tape libraries that make a combined radiology/cardiology archive less expensive and more efficient to operate than two separate archives. Once again, bigger is better.

Local PACS Subsystem

Not long ago, the PACS server fed images to a variety of display stations: diagnostic, clinical review, take home. In those days, PACS was a big system with a long list of components and a price tag to match. With the advent of sophisticated web viewer application packages, the top-end web viewers have actually surpassed the feature set of all but the diagnostic display. Any differences in performance can usually be overcome with a network upgrade, or clever web viewer applications, like local disk caching and batch downloads, or subscription mode. Local disk caching refers to storing the images on the web viewer’s disk drive. Batch download is an application that allows a user to retrieve multiple image studies in a single operation. Subscription mode lets the user instruct the web server to automatically forward to a specific web viewer certain types of studies as soon as they become available to the web server. With a server that supports one or more of these applications, there may be no need to purchase the more expensive clinical review stations for the emergency department, the critical care areas, and beyond. The local PACS is effectively reduced to a set of diagnostic displays and their matched server(s). And since the archive server subsystem configured with generic storage is where most of the image data is stored, the local PACS server can be scaled down to a smaller configuration. This scaling factor means an even smaller PACS subsystem. The smaller system configuration and competition among multiple vendors for the PACS subsystem piece of the business have effectively reduced the purchase price.

Voice Recognition Subsystem

EIMS projects in the last few years rarely included a voice recognition (VR) reporting component. This was probably due to the belief that such systems were imperfect and incapable of 100% accuracy. While they will probably never be 100% accurate, the current generation of VR is much improved. Nevertheless, many radiologists are still reluctant to consider using VR because they do not want to get trapped into doing their own editing, so including it in the PACS project is frequently a sticking point. Here are several arguments for including it in the system.

The investment in an electronic information delivery system (the web server) that does not include some method for providing the report along with the images is not going to meet the vast majority of the referring physicians’ expectations. This is especially true if one appreciates that 80% or more of those users are looking primarily for the reports, not just the images, if they are looking for the images at all. VR has improved to the point where it is accurate enough to be the difference between a successful PACS project and a failed PACS project. There are two reasons for including it in the EIMS project. The first reason is that of function: making reports available in a few hours, not days. The second reason is financial: putting the 80% savings in transcription costs in the business model to help offset the display equipment. The way to alleviate any concerns among the radiologists is to stress the part of the plan that uses the remaining 20% of the transcriptionist pool to perform any edits that the radiologists chose not to perform.

In some cases, the availability of the reports alone can mean the difference in when a patient is moved from one class of bed to another, or discharged altogether. If this more timely reassignment can be linked to reduced cost of inpatient care, there is yet another savings that could be credited to the EIMS and used in the business model to offset costs.

Web Server

As previously mentioned, the arrival of the web server and its feature-rich web viewer has made a significant impact on the cost of the local PACS subsystem by eliminating all but the diagnostic displays from this formerly expensive subsystem. The web server/viewer subsystem itself is largely a software product. The server hardware is off-the-shelf and does not need to be too beefy since the majority of the local image cache can be assigned to the archive server. The argument that this separation of image storage across multiple platforms from different vendors will hurt performance is easily overcome with the latest technology. This includes appropriate network topology, use of multiple servers to balance input/output, use of SAN or NAS for actual image storage, and the aforementioned viewer applications like batch downloads or subscription mode. Not only is some cost being removed from the local PACS subsystem through the elimination of the clinical review stations, but the web server itself can be built at a reduced cost.

As for the web viewers, many of the computer platforms may already exist on site. Most of the referring physicians may be able to meet their web viewing performance expectations by using the PCs already deployed throughout the hospital, in their offices, in their clinics, and in their homes. Once again, such important features as local image caching and batch downloads, and/or subscription mode, may be required to meet those expectations. If the use of existing computers turns out to be successful for a large percentage of users, the cost can be as minimal as the user license of a few hundred dollars each.

For those locations where user expectations are justifiably higher (emergency department, critical care, and visiting physician stations outside the film library), a high performance web viewer will have to be deployed. Even if this means two flat panel 1,600-line displays wrapped into a $20,000 package, the cost is now under that of most equivalent clinical review configurations.

Another way to achieve reduced system costs is to use web viewers in the surgical suites, but limit their number. Limited use of the web viewer in this case may be the only economical solution for image viewing by the surgeons. Replacing every lightbox panel in each of the surgical suites with some kind of electronic display is neither technically feasible nor cost-effective. Even if a two-channel display in each surgical suite was acceptable, the cost is significant. A better solution may just be to place enough web viewers in the neighborhood of the surgical suites to satisfy availability, then add a high-resolution postscript printer to the local network. The web viewers use simple browser interfaces that allow fast and easy access to the vast array of Microsoft applications, including print. The surgeons themselves or another member of the surgical staff could easily and quickly make hard copy of whatever images they wanted to take into the surgical suite. The low-cost paper or transparency media could be placed on the existing lightboxes. This hybrid solution would be considerably less expensive than deployment of eight-panel display stations in all the surgical suites. And sometimes hard copy is the best display media.

As a final note on the web server, a few of the current web server/viewer products can also provide specific viewing applications for cardiology studies. Even though these viewing packages require a different set of user licenses and therefore additional costs, the additional cost is less than that required to build a duplicate web server just for cardiology. The same argument can also be made for pathology images (slides) and any other imaging modality that can create DICOM Visible Light data objects.


More and more RIS are being equipped with DICOM front ends. These software upgrades or new system features are eliminating the need for the EIMS-RIS interface and the various HL-7 software interface packages that are typically required to complete the bidirectional EIMS-RIS connection. Although a DICOM front end for a specific RIS may still cost some dollars, the cost is usually less than the sum of the parts that it is replacing. In addition to the savings in hardware and software cost, the performance improvements and additional features may further reduce the labor costs associated with managing and servicing the EIMS.

As previously mentioned, many of the existing PCs deployed throughout the enterprise to serve as terminals for the hospital information system may also serve as web viewers for radiology images and reports. This effectively reduces the cost of deploying electronic image viewers required to replace film.

Another cost-reducing strategy is the linking of the radiology web server to a button on the operator interface of the EMR terminal. If the referring physicians and nursing staff can get accustomed to the operator interface of this terminal, then why not use it for access to the radiology information? Rather than making everyone learn multiple interfaces and make multiple log-ons, it would be easier and more efficient to use the EMR interface as the one and only interface and log-on required to access all patient information. This can be accomplished by adding an http link in the EMR graphical user interface to the radiology web server database. While this would improve the efficiency of the users, it would be difficult to assign a value to this efficiency. However, it is possible to draw a correlation between ease of use and cost of training. This single point of access would reduce the cost of user training, which is easily estimated and then included in the business model.

In conclusion, there are several recent technological advances that can be used independently or in combination to reduce the cost of an EIMS. The most significant of these advances are the generic DICOM archive and the web server.

Another major contributor to lower costs is the maturing of several product lines. The maturing of external DICOM interfaces, computed radiography, and even digital radiography has brought either product parity or competitive pressure to bear on pricing strategies. The net effect is lower prices on items that usually are purchased in quantity.

Another contributor to the reduced cost of deploying an EIMS is the concept of splitting the system into smaller subsystems that interoperate using the DICOM standard. This breaking of the traditional PACS into smaller subsystems not only makes it possible to select the best product for the application, but it also encourages competition in every phase of the deployment schedule, and that tends to lower the effective purchase price.

As ironic as it sounds, it is usually better to build more of a system than less. Think bigger. By including such technologies as voice recognition reporting systems and digital mammography, it is not only possible to achieve a higher level of system performance, it is also possible to include additional savings and new revenue in the EIMS business model. These additional dollars drop directly to the bottom line and they just might be the last dollars required to reach or exceed breakeven. If an EIMS business model does not come within at least a few hundred thousand dollars of breakeven each year of operation, it may be necessary to fall back on those cost justification theories or the cannot-afford-not-to argument.

Michael J. Gray is principal of Gray Consulting, Novato, Calif, a consulting company specializing in PACS design for health care providers, (415) 892-0943.