David Artz, MD, medical director, represents Memorial Sloan-Kettering Cancer Center, New York City.

 How are you approaching the challenge of easing enterprise access to image data, particularly as it relates to radiology? The participants in this electronic roundtable, teams that included a CIO and representation from radiology, agreed to address questions investigating image data storage issues.

John. P. Glaser, vice president and CIO, Partners Healthcare

 Memorial Sloan-Kettering Cancer Center, New York City, is represented by Patricia Skarulis, vice president, IS, and CIO; and David Artz, MD, medical director, information systems.

Partners HealthCare, Boston, is represented by Keith J. Dreyer, DO, PhD, vice chair of radiology, computing and information sciences, Massachusetts General Hospital; and John P. Glaser, PhD, vice president and CIO, Partners HealthCare System.

Keith J. Dreyer, DO, PhD, vice chair of radiology, computing and information sciences, Massachusetts General Hospital represent Partners Healthcare, Boston

 Sutter Health, Sacramento, is represented by Debra Sleigh, CIO, IT enterprise strategic development and integration; Fred Gardner, radiology and PACS Sutter information systems product manager; Mario Garcia, senior technical project manager; and John Hummel, senior vice president, IS, and CIO.

Q: Does your institution have, or plan to have, a common repository for all image data throughout the enterprise? How important is it to devise a storage strategy now?

John Hummel, senior vice president, IS and CIO , represents Sutter Health

 Memorial Sloan-Kettering: We have a common repository for all images generated in radiology. These are all Digital Imaging and Communications in Medicine (DICOM) images. We have separate repositories for other images, generated by other modalities, that are specialty specific. For example, we have freestanding repositories for echocardiography, dermatology, and endoscopy. Of our images, 99% are in the picture archiving and communications system (PACS). These images include CT, MRI, ultrasound, positron-emission tomography, nuclear medicine, fluoroscopy, bone-mineral densitometry, and radiography. The exceptions are in department-specific systems and, in general, are not DICOM images. For example, we have Joint Photographic Experts Group (JPEG) images for endoscopy done by our gastroenterology department. It would not be economical, nor would it be practical, to try to go after that last 1% of images. This is our strategy.

Partners: There is a common image repository used by most, but not all, of the hospitals within the enterprise. Each hospital has its own clinical storage in the form of a redundant array of independent/inexpensive disks (RAID), while the legal archive is common at the enterprise level and is managed by a single information-technology team using a commercial hierarchical storage management system.

It is very important to plan your storage strategy before you purchase PACS. It is not necessary, or financially advisable, to purchase the entire solution years in advance, but the plan should be set.

 Sutter: From the very first planning meetings on PACS, the need to have a centralized storage solution was part of the system strategy. Sutter Health is an integrated delivery network made up of more than 32 hospital campuses and 18 clinics across a large geographical area in Northern California and Hawaii. Our patients are very mobile in where they work and live, so we have multiple access points to the Sutter Health System.

 There are inherent cost savings in having one main database for PACS data (all modalities), as compared with over 50 separate databasesall supported by hardware, software, and database administrators. If Sutter were to have one database at each site, it would also mean having 50 different fail-over/high-availability systems, and that would be a huge expense to justify for these clinical systems. It would also require interfaces between systems to allow for sharing of files in a real-time environment. This would include remote diagnostic readings, community physician referrals, and plate-storage strategies. Sutter Health chose a commercial central PACS data-storage partner and was able to complete the central storage system in 2003.

For Sutter Health, the need to strategize concerning a system-wide storage solution was part of the earliest of return-on-investment calculations. By facing the fact that a proliferation of independent PACS storage databases all across the enterprise would create a logistically, operationally, and financially prohibitive system, we knew that we had to think outside of the traditional PACS box and find a solution (and a technology partner) that could resolve these issues for us.

Fred Gardner, radiology and PACS Sutter information systems product manager

One of the challenges of having a long-term storage system was deciding how and when to implement an enterprise master patient index (EMPI), along with determining how best to coordinate the implementation of an EMPI with the PACS long-term storage system. We were able to choose a very effective EMPI system. This artificial intelligence program allows us to incorporate our current medical records numbers by creating a cross-reference table with an EMPI. In this way, we could more rapidly roll out our EMPI along with our PACS long-term storage system, with a lot less disruption of operations at our various sites.

One of the more worrisome aspects of a long-term storage system was its effect on our wide-area network (WAN). Since our WAN carries 100% of our data traffic, degradation in the WAN's bandwidth and latency would be critical for our electronic medical records (EMR) and other clinical systems during most hours of the affiliates' operations. For this reason, we started a review of our current frame-relay WAN circuits, and we have started the process of upgrading all our circuits. We are also ensuring that we have cost-effective bandwidth.

We are working with a vendor to gain maximum value from data-compression techniques to ensure that our long-term storage system does not have too negative an impact on the WAN. One other way of working through our concerns about the impact of long-term storage on the WAN was to work with a vendor to build a short-term storage system built on a storage-area network (SAN) technology to cache images at each site as they are transported to long-term storage. In this way, we can ensure that our PACS images are in the proper DICOM 3 format for long-term storage, are compressed, and are associated with the EMPI.

Debra Sleigh, CIO, IT enterprise strategic development and integration, represents Sutter Health

We also worked with our radiology information system (RIS) vendor to build a RIS-PACS interface to allow for full association of all PACS modalities and images with our EMR and e-health (web) strategies and with PACS work flow through the RIS.

Q: What are the drivers promoting the common repository? What are the drawbacks of common-repository use?

Sutter: There are several important business drivers that created the strategic need for Sutter Health to adopt, early in the PACS strategy, a long-term centralized storage system. First of these was the need for availability of all radiology modalities for review by any authorized clinician at any time. This meant that a way to share files and to identify patients uniquely would be needed from the very beginning of our PACS project. The second driver was the cost savings produced by aggregating multiple archive servers. Maintenance, database administration, and hardware costs could all be mitigated by having one long-term storage system that could be plugged into any of our PACS components. Third, we needed a system that would allow us to have nearly instantaneous access to all modalities' images locally, but that could be used in a WAN-friendly format across the enterprise.

Mario Garcia, senior technical project manager, represents Sutter Health

Our SAN and long-term storage solution allow us to have the PACS locally do the work that it is intended to do, but still allow us to store the images for across-the-enterprise access. Fourth, by having the long-term storage centralized, we could build into the planning and architecture the high availability and fail-over redundancy that would not be available in multiple smaller systems.

One of the biggest drawbacks of a long-term storage system would be the need to build the interfaces among the RIS, multivendor PACS, and each inpatient and outpatient facility. By using a long-term storage architecture and a SAN local storage system, we could, in fact, have the interfacing done at the local level, thereby having only converted images transmitted and kept in long-term storage. This allows a greater degree of flexibility in interfacing and a huge reduction in costs.

Memorial Sloan-Kettering: The common repository is driven by economics, efficiency, patient safety, and disaster recovery. It is cheaper and more efficient, from a storage standpoint, to use a single system. It is easier to back it up, and it is easier and faster for clinicians to obtain images from a single system. We can distribute images and train users on the front-end viewer of just one system.

Drawbacks are that the proprietary PACS will address the greatest volume of images (DICOM images from radiology sources), but it will not address all of our imaging needs, particularly for tagged image file format and JPEG images in endoscopy and dermatology. Vendors in nonradiology areas do not use DICOM, so it would require way too much work on our part to implement.

Partners: There are a limited number of vendors that provide hierarchical storage management (HSM)-based legal archives. Several PACS vendors still try to provide these solutions themselves, but they are all fraught with limitations. If you wish to stay on top of the removable, medium-price performance curve, the only hope you have is to use a storage vendor for your legal archive. These facts are true for single or multiple hospital systems. In addition, if you are under a multiple hospital system, it is far more advantageous to use the same storage vendor for HSM but not for the media behind it. With a single-vendor solution for storage media, such as RAID, competition and great pricing will be lost. Do not select a storage vendor whose solution requires their own storage media.

One drawback to common data storage is the need to gain buy-in from all of your institutions. Unless your enterprise has total control of each institution's clinical information technology actions, it is best to provide a common storage solution as an option, not a requirement.

Q: What steps (if any) are you taking now to meet projected future storage needs?

Partners: Storage is purchased centrally at 6-month intervals. Vendor selection and price negotiations occur at each of these intervals. This approach allows for unanticipated growth, which is not uncommon in radiology these days, and offers a pricing strategy for storage that is rapidly decreasing even while your annual data requirements are increasing.

Sutter: Sutter Health is implementing a system that is scalable to meet future growth. We built our hardware and databases in a way that allows for our expected growth of three to five terabytes per year and still has our fail-over and high availability built into the design. We have analyzed image acquisition across the enterprise and have taken into account the new generation of scanners that will be producing much larger studies than in the past, as well as the desire to store all modalities of patient imaging (including all raw slices, films and ultrasounds).

Memorial Sloan-Kettering: We are implementing a new storage architecture that will provide 100% of the PACS archive on RAID, completely replicated on two sites, with a third copy on optical disk. Each of these archives will be at physically discrete sites. This is being done for disaster recovery and high availability.

Q: What investments are expected in hardware, networking, software, and personnel?

Memorial Sloan-Kettering: We expect our spending patterns and staff support to remain at current levels.

Partners: Naturally, this varies depending upon size of your enterprise, existing information technology infrastructure, and HSM vendor support agreements. A rule of thumb for enterprise storage is that you will need approximately the same number of staff and new infrastructure as you would for your largest hospital's PACS. The addition of new hospitals will simply require the cost of their interface with the HSM and a new storage calculation for purchases of media at 6-month intervals.

Sutter: There will be significant initial investment in scalable servers, as well as incremental additions of required storage space. Historical network infrastructure will not be acceptable and will be upgraded. Investments will be required in training and retention of the best and brightest RIS personnel, as well as competent information-technology personnel to support the system from the clinical modality to the archive.

Investment in the proper help-desk model and escalation models will be critical to the operation of the system, since an error detected at an affiliate may be related to a local issue or could be related to the central archive.

Q: What effects have the image solutions devised for radiology had on your strategy for an enterprise approach to data storage and accessibility?

Sutter: With the recent development of multihead scanners, with perhaps thousands of images per patient, we are making sure that our long-term storage system and WAN have the capacity to hold these new images. We also work closely with our various radiology groups to keep up with the latest trends in the business work-flow process. We actively participate in radiology conferences and vendor meetings to stay abreast of industry changes and trends.

Memorial Sloan-Kettering: We operate in a paperless environment; therefore, the online PACS is a key component of our environment. I could not even imagine what it would be like if we were transporting films around.

Partners: The implementation of medical image storage has followed a path that is parallel to, but not identical with, the path of text-based clinical data. The two systems are growing at far different rates and are supported by different teams, with a variety of common infrastructure strategies to obtain economies of scale. The same will probably be true for our centralized genetics and genomics initiative.

Q: Is the application service provider (ASP) model viable for enterprise-wide storage?

Partners: With the cost of storage media decreasing rapidly and continuously, the critical nature of the data in question, and the recent history of ASP storage companies leaving the medical market through lack of interest and bankruptcy, ASP storage seems to be an unsafe and unnecessary proposition.

Sutter: In effect, Sutter Health is the ASP for its affiliates. The organization sees no advantage in storing images outside the organization, since this could result in worrisome data issues related to the Health Insurance Portability and Accountability Act. We did review the cost of ASP models, and for one of our smaller, remote affiliates, we have installed a modified ASP system. In this particular case, we have the hardware inside our firewall to ensure that we can have responsible monitoring and auditing of the data (and access to it).

Memorial Sloan-Kettering: ASP is a viable model, depending on the size and technical sophistication of your information-technology organization. It is a good idea for smaller hospitals; we have a large and technically sophisticated staff, so we maintain our own systems.

Q: How does your strategy relate to the EMR?

Memorial Sloan-Kettering: Right now, we feel that the best-of-breed EMR vendors and best-of-breed PACS vendors are separate companies, so we need to integrate functionality from each. We have developed our own single sign-on and context manager, so one can quickly access a patient's images or that patient's chart from either place.

Partners: While strategies for the image archive and EMR are common in their centralized approaches, the two do not interface. Instead, compressed images are delivered to the EMR users via web services directly attached to our PACS.

Sutter: Over successive years, the Sutter Health strategy has increasingly integrated all applications, from infrastructure and networking through financial and clinical applications, to the EMR. With the additions of an EMPI and an enterprise long-term storage solution, Sutter Health is positioned to integrate most aspects of the patient's experience into the EMR. Our long-term storage then becomes a pointer location within the EMR and thus prevents our having to store information twice within the actual EMR database.

Q: Has your thinking on this subject changed in recent years? What is the EMR likely to be, and how does it differ from legacy ideas for it?

Sutter: Historically, the EMR has been affiliate-specific and patient data have been isolated to that specific site. With the advent of EMPI and PACS long-term archival, patient images across the enterprise will be available with the written reports from multiple sites in a secure central environment.

The success of long-term storage also shows that there is an opportunity to create a virtual EMR system, where we have large databases like PACS that are associated with the EMR through a EMPI and RIS. These pointers then allow for access to the needed information from long-term storage while minimizing full integration into the EMR. Ideally, as other information is created that is mostly data centered, a similar model could be created within the EMR. Having these pointer locations embedded within the EMR allows us to keep the EMR database and retrieval documents to a minimum level.

Memorial Sloan-Kettering: The EMR has been a greater area of focus for the past 4 or so years due to its proven benefit in patient safety (with physician order-entry of medications). Safety is a greater focus in the industry in general. We consider an EMR to be an integrated system for review of patient diagnostic data and entry of patient-care orders with associated decision support. I see the EMR moving much more to structured data rather than binary large objects, or BLOBs, of text.

Partners: The EMR is an evolving concept and, as such, will continue to mature. While we have been using an EMR for well over 10 years, we continue to add new functionality and accessibility requested by clinicians through technical advances and new system integration.

Kris Kyes is technical editor of Decisions in Imaging Economics.