Health data exchange aims to improve cancer imaging

April 2, 2006

The ever increasing resolution and specificity of diagnostic imaging are certainly among the great success stories in the fight against cancer. However, whether used for screening, work-up, staging, or follow-up, imaging is at its best when radiologists have the benefit of comparing new images to historical priors, a practice well known to increase the accuracy of interpretation1,2 and one that represents the standard of care.3

The ever increasing resolution and specificity of diagnostic imaging are certainly among the great success stories in the fight against cancer. However, whether used for screening, work-up, staging, or follow-up, imaging is at its best when radiologists have the benefit of comparing new images to historical priors, a practice well known to increase the accuracy of interpretation1,2 and one that represents the standard of care.3

A fundamental challenge to the quality, impact, and cost of cancer imaging remains the portability of patients among multiple providers and facilities. In seeking multiple opinions or simply the "best" providers of surgical, pharmacologic, and/or radiation therapies, cancer patients aim to maximize the quality and effectiveness of their care.

Movement by these patients across geographic and institutional borders often works at cross-purposes, however, by creating gaps in their physicians' knowledge of recent results. This can lead to delays and increased costs. In addition, it can obstruct collaboration among their physicians instead of promoting it.

Recent data suggest that, in up to 20% of all imaging cases, relevant prior images desired by radiologists for comparison with current images are located beyond the borders of their own institution.4 Given the higher rates of patient mobility for cancer care relative to other clinical conditions, one should expect even higher percentages of relevant but out-of-reach priors for cancer patients. Beyond the accuracy of interpretation, the mobility of cancer patients introduces a high degree of duplicative studies that increase the cost of care dramatically while exposing the patient to additional radiation and other procedure-related risks.


As patient portability is certain to remain a fixture of healthcare in the U.S., maximizing the efficacy of cancer imaging requires cross-enterprise integration. Such integration should provide physicians with on-demand, online access to a patient's complete longitudinal record and imaging history, irrespective of where images and reports were generated and stored.

The idea of institutions, even competing ones, securely exchanging digital clinical data on shared patients is certainly not limited to diagnostic imaging. In just the last 18 months, more than 100 initiatives, many of them to form so-called regional health information organizations, have sprouted around the U.S. with the goal of establishing health information exchanges (HIEs) that create and support regional "womb to tomb" longitudinal electronic health records.5

Diagnostic imaging represents an ideal starting point for exchange efforts, given the widespread adoption of PACS/RIS in most healthcare markets along with the near universality of DICOM as an established interoperability standard. Only a few additional ingredients are required to rapidly achieve large-scale adoption and critical mass across a region.


This next step has been taken in the Philadelphia region, the country's fifth largest healthcare market and among its most competitive, with more than 30 integrated delivery networks and health systems. The Philadelphia Health Information Exchange is the nation's first such digital network connecting a region's competing imaging providers to securely share images and reports in support of patient care.

Funded by $2.3 million in grants from the National Cancer Institute and the National Institute for Biomedical Imaging and Bioengineering, the exchange links PACS and RIS from different vendors, of varying ages, and spanning multiple, even unaffiliated, medical facilities into a distributed network.

As of this writing, the network includes the University of Pennsylvania Health System, Thomas Jefferson University Hospitals, Children's Hospital of Philadelphia, and Albert Einstein Healthcare Network. The available imaging data cover some 300,000 patients and eight years of history.

The exchange serves as both a clinical and research platform. On the clinical side, it has been designed to accommodate the most demanding of healthcare environments: multiple competing facilities, multiple vendors, both legacy and new PACS/RIS, continuous uptime, and volumes of imaging as large as 400,000 exams per year for some participating sites.

The network was designed to scale to an unlimited number of participants and to accommodate future expansion into other forms of clinical data. Accessible via a Web portal, the network will soon also be accessible directly from PACS workstations, electronic medical records, and other applications using the open standards specified by the cross-enterprise imaging document sharing profile (XDS-I) established by the Integrating the Healthcare Enterprise initiative (see accompanying article). Most important, the exchange was built to be easily replicated in other regions, based on a minimal set of commercially available components.


On the research side, the network is enabling investigators to ask and answer a wide range of questions at the forefront of health services and medical informatics research. The most recent initiatives are focused on cancer care. Researchers at competing institutions are collaborating to quantify:

- the extent to which relevant priors are available across a region;

- the extent to which duplicative imaging unnecessarily exposes patients to risk and the resulting impact on healthcare costs;

- the ways in which imaging-specific HIEs can improve patient outcomes across different clinical conditions and specialties; and

- the best way to integrate imaging exchange with workflow in the reading room and with the practices of referring physicians.

The coming year will also see expanded clinical use of the network with an enhanced user interface geared specifically at real-time clinical collaborations across enterprise borders as typified by tumor boards and case conferences.


As a model for HIEs sure to follow in other markets, it is important to note that the architecture of the Philadelphia network is strongly related to technical, political, and legal measures of success.

Specifically, the Philadelphia HIE follows a distributed (or "federated" or "peer-to-peer") model. As such, it leverages existing investments in PACS and RIS by each participating imaging provider, keeping the patient data at that site, behind its firewall, with release subject to the site's disclosure policies and procedures.

Authorized and authenticated physicians use the network by means of a Web portal that provides access to patient-centric virtual folders that span multiple imaging sites.

The imaging sites contribute data to the network by means of a drop-in network appliance that:

- enables the facility to field electronic requests from outside physicians for its patient's data;

- provides the means for such outsiders to rapidly meet the site's disclosure procedures; and

- enables the site to securely transmit the digital images and/or reports directly to the requester's Web browser or clinical application.

This approach has the critical advantage of maintaining the political and operational status quo, thereby lowering the barriers to adoption by providers and achieving the critical mass required to ultimately see a benefit for patients. The disadvantage of the distributed approach is the additional overhead that comes with managing multiple nodes and maintaining the availability of data 24/7/365.

The distributed approach taken in Philadelphia contrasts with an alternate vision that has yet to prove successful in addressing the obstacles to achieving health information exchange in the real world. This "medical banking" approach involves establishing clinical data accounts owned by the patient and centrally hosted and managed by one or more third-party organizations (perhaps even actual financial institutions). Under this model, patients collect their own digital data or have their care providers submit the data they generate to the appropriate account. Subject to the patient's permission and direction, care providers could access the account to visualize the information contributed by others.

While some might find the patient's role as gatekeeper and the mimicry of personal financial accounts attractive, the disadvantages are significant. First, real-world implementation would require familiarity and access to technology that may be well beyond many patients, particularly those who most need the benefits of health information exchange. Second, the model introduces practical barriers to physician adoption by requiring providers to relinquish control of the imaging data that many consider a core business asset. Third, a multiplicity of such "banks" would introduce the need for multiple integrations at each physician practice (or perhaps require yet another party to act as a clearinghouse to broker communications with each bank). Fourth, this centralized archiving model requires significantly more storage capacity and bandwidth consumption than the distributed approach.


Given the maturity of PACS and RIS, imaging data exchange is perhaps best seen as the next step in the evolution of digital diagnostic imaging. A move in this direction could mitigate the impact of patient portability, increase accuracy and timeliness of care, and contain the cost of imaging.

Healthcare providers should not only expect their future PACS and RIS to natively support the ability to reach across regions to access priors from networks at other facilities, they should demand such functionality from vendors. Many vendors are already taking important steps in this direction, by including compliance with IHE cross-enterprise integration profiles in their product road maps.

Requiring such compliance in future requests for proposals will ensure that PACS/RIS investments can make the most of health information exchange initiatives. Such encouragement by PACS/RIS customers will only hasten the time when the full capabilities and impact of cross-enterprise image sharing are achieved.


1. White K, Berbaum K, Smith WL. The role of previous radiographs and reports in the interpretation of current radiographs. Invest Radiol 1994;29(3):263-265.

2. Aideyan UO, Berbaum K, Smith WL. Influence of prior radiologic information on interpretation of radiographic examinations. Acad Radiol 1995;2(3)205-208.

3. ACR standard for communication: diagnostic radiology. Standards: American College of Radiology, 1995.

4. Lakhani P, Menschik ED, Goldszal AF, et al. Development and validation of queries using structured query language (SQL) to determine the utilization of comparison imaging in radiology reports stored on PACS. J Digit Imaging 2005 (online).

5. Marchibroda J, Covich Bordenick J. Emerging trends and issues in health information exchange: selected findings from eHealth Initiative Foundation's second annual survey of state, regional, and community-based health information exchange initiatives and organizations. Washington, DC: eHealth Initiative, 2005. (

Dr. Menschik is founder and chief executive officer of Hx Technologies, a healthcare services and IT company in Philadelphia that builds and operates health information exchanges with particular emphasis on diagnostic imaging.