The importance of Java and CORBA in medicine

David W. Forslund, PhD, , James L. Cook, MD*

Los Alamos National Laboratory
Los Alamos, New Mexico
*National Jewish Center for Immunology and Respiratory Medicine
Denver, Colorado
One of the most powerful tools available for tele- manner without regard to the location of the patient
medicine is a multimedia medical record accessible or even the physician. For example, the healthcare
over a wide area and simultaneously editable by system needs to support synchronous and asynchro-
multiple physicians. The ability to do this through an nous remote consultations. That is, the physicians
intuitive interface linking multiple distributed data may need to work together in real time for diagnosis
repositories while maintaining full data integrity is a or treatment or be able to provide consultation serv-
fundamental enabling technology in healthcare. We ices off-line at his or her convenience. This should
discuss the role of distributed object technology us- be done without having to use special equipment or
ing Java and CORBA in providing this capability go to special locations.
including an example of such a system (TeleMed)
which can be accessed through the World Wide Web. Having data made available simultaneously over a
Issues of security, scalability, data integrity, and us- wide area implies a protection of an individual's pri-
ability are emphasized.. vacy in a way that is not normally done in medical
institutions today. Clinical records need to be made
INTRODUCTION available to people only if they are authorized and the
protections need to be commensurate with the risk
Healthcare is undergoing a rapid change due to the associated with a broader network accessibility. This
enormous improvements in computing and commu- implies secure management of a public/private key
nications technology. It is no longer adequate in infrastructure to authenticate access and to manage
healthcare to think of the electronic medical record as access control
an analog to a paper record which is contained in a
hospital's medical records department.' Because One of the weaknesses of most telemedicine systems
healthcare is rapidly taking on a distributed nature, it being put in place is the lack of a provision for data
should be regarded as a collaborative tool that can integrity. The information generated in the telemedi-
incorporate all types of data including sound, images, cine encounter are not necessarily captured in the
and video in an integrated manner even though the clinical record, or worse, the information gets entered
sources of data and processing may be widely dis- multiple times and is out of synchronization between
tributed2. This requires a robust distributed comput- the multiple locations, because each site manages the
ing infrastructure accessible on a wide range of plat- information differently and copies are sent around
forms. The premier open architecture which can from location to location. It is important, then, that
meet this requirement is based on the Object Man- telemedicine systems fully address the management
agement Group's (OMG) Common Object Request of the data they generate and minimize errors associ-
Broker Architecture (CORBA)3. We describe here ated with copying and separate management of these
the motivation for the use of Java and CORBA in copies. Basically, the principle is single entry and a
telemedicine, its benefits, limitations and an example provision for a unified view of the clinical encoun-
implementation and use in the TeleMed system de- ters.
signed and implemented at Los Alamos. With its
basis on object technology it provides support over a The management of medical information over a wide
wide area network of proper containment and man- area cannot be done without some means of unifying
agement of data and processes involved in healthcare the view of disparate data. For example, it should be
and in other industries. possible to locate patient data anywhere on the net-
work by means of a distributed master patient index
BACKGROUND as well as a means of resolving the distinction be-
tween disparate terms which are used to describe
Although we will use the term telemedicine through- drugs, encounters, diagnoses, treatments, etc.
out this work, we believe the term is inappropriate
because it tends to distinguish itself from the normal Once the data is managed over a wide area in a con-
practice of medicine. The goal of telemedicine sistent manner, it should be possible with a telemedi-
should be to bring advanced communications tech- cine system to provide tools for much more auto-
nology to bear in the normal practice of healthcare. matic outcomes analysis. That is, one should be able
For the purposes of this paper we define telemedicine to utilize the system itself to manage the aggregation
to be the practice of moving healthcare to patient of data to facilitate the evaluation of the telemedicine
with communications technology. This amounts to system itself as well as the determination of factors
being able to practice healthcare in a distributed needed to improve healthcare treatments. For the

1091-8280/97/S5.00 0 1997 AMIA, Inc. 364

telemedicine system to be truly cost effective it must day to manage this problem in areas besides health-
demonstrate that it can reduce the time required to care. Appropriate authentication as well as authori-
reach a diagnosis and begin treatment of the patient. zation to limit access to certain types of data are re-
This should then result in improved outcomes for the quired. The data also needs to be encrypted at least
same or less cost. between the source and the destination with a man-
One of the foundational requirements for the system ageable risk associated with its interception. Finally,
to be effective even if it can aggregate consistent data audit trails of access are required so that data access
over a wide are is for it to be easy to use by the phy- and modification is non-repudiable.
sician and easy and fast to access. Because a medical A method of locating patient information is also
record can be very complex and data intensive, it needed to support the virtual patient record so that
must provide a mechanism for intuitive navigation locations can be "automatically" queried to see if
through the information by a health care provider. It they have relevant patient data. Although a central
must provide the knowledge the caregiver needs repository may seem to be the most logical solution,
without overwhelming her with unnecessary infor- it is, in fact, not desirable because of issues associ-
mation not relevant to the current decision process. ated with manageability and ownership of the infor-
At the same time the relevant information must be mation. An Internet-based "DNS-like" solution in
immediately accessible to reduce the time to reach a which there are peers which exchange relevant data
diagnosis. Basically, the system must give the illu- pointers is more resilient to system down time and
sion of providing the entire medical record on de- allows for the local management of patient privacy in
mand to the physician while actually only delivering the institution in which the knowledge is created.
small parts of it.
VIRTUAL PATIENT RECORD To achieve a truly usable virtual patient record will
require industry cooperation on an unprecedented
REQUIREMENTS scale and the use of infrastructure support relevant to
a broad-base of industries. Signs of this new level of
Considering these above mentioned requirements for cooperation is already apparent in the various con-
distributed healthcare, we realize that we need to be sortia that are seeking to improve healthcare.
able to manage a "virtual" patient record that is as-
sembled from multiple locations dynamically. Al- THE ROLE OF THE INTERNET
though this is highly desired, there are many obsta-
cles to being able to create a "virtual" patient record The rapid acceptance of the Internet can provide a
in healthcare. One important requirement is the abil- substantial impetus to meeting these above require-
ity to map the disparate data and process representa- ments. It provides standard communication mecha-
tions that are prevalent in healthcare into a common nisms (e.g. TCP/IP, HTTP, IIOP), platform inde-
structure. This would enable very different systems pendence, low cost distribution methods (e.g. Java)
with their own internal data structures to communi- and a high level of connectivity. It supports various
cate effectively. This approach means more than distributed services including standard naming
simply a messaging infrastructure, but a common mechanisms (e.g. URN's, i.e., Universal Resource
way to describe the processes as well as the informa- Naming4), security (e.g., DCE, certificate servers,
tion in healthcare. It should rely on infrastructure etc.), tools for distributed data abstraction and encap-
usable in a variety of disciplines and capture the sulation (CORBA). At the same time, there is a rapid
processes of the "business" of healthcare. Sometimes evolution and acceptance of multimedia technologies
this concept is referred to as "common business ob- (internet-base audio and video), as well as scalable
jects". The actual internal layout of data is not par- storage systems (e.g., object relational databases)
ticularly relevant to this approach. which support a variety of data-mining capabilities.
A second requirement is high accessibility of the The requirements we have outlined to support a vir-
information while simultaneously providing high tual patient record are formidable. Is there any infra-
data integrity. To provide this, low-latency networks structure being put in place which could simultane-
are needed (more important than high bandwidth) so ously meet all of these requirements? Probably not.
that locations can be queried quickly and data refer- However, there is an approach within industry which
ences transferred. When quality networks are not provides many of the mechanisms needed to support
available, technology for database replication and a virtual patient record. It is based on the work
consistency management must be provided. A com- coming from the OMG consortium. At the first
bination of these technologies is required for proper level, now traditional object oriented analysis and
support of the virtual patient record. A corollary of design (OOAD) methodology is used to provide the
this requirement is the support for distributed trans- right level of abstractions and definitions of data and
action processing so that data can be updated consis- processes involved in healthcare. This work is being
tently between multiple data sources. spearheaded by the healthcare task force in the OMG
As data is merged to produce the medical record "on known as CORBAmed. It is insufficient at this point
the fly", the maintenance of privacy for the end user of technology to simply describe the data that needs
to be exchanged between healthcare facilities. The
becomes even more important. Although this can be processes must be encapsulated and common defini-
complex, the infrastructure is being put in place to- tions must be agreed upon, not simply in the aca-

365
demic sense but in the truly practical sense of com- repudiation service so that events placed in the medi-
mercial applications. cal record cannot be denied at a later date and their
source can be verified with confidence. The Lifecy-
As powerful as the OOAD methodology is, the wide cle and versioning services provide for the long term
area management and access control of objects is maintenance of the distributed objects.
much more important. It is this work of the OMG
that can provide the greatest benefit to the healthcare The popularity and ubiquity of the Internet is leading
industry. As part of the infrastructure support of many vendors to provide WWW or HTML interfaces
CORBA, a variety of services are being defined for to their data repositories both in healthcare6,7 and in
the management of the objects being accessed over other fields. Although extremely useful and easy to
the network. The strength of the OMG comes from support, there are several difficulties with this trend.
the fact that it is an industry-led consortium repre- The Web is primarily a publishing paradigm by de-
senting a very broad commercial market and thus can sign because of the stateless nature of the http proto-
leverage the infrastructure investment across many col. Although state can be maintained on the server,
industries. The current membership consists of over the client really has no knowledge of it which can
700 companies and institutions. lead to confusion in the data if not a sacrifice in data
security and integrity. The IIOP protocol provides
The CORBA architecture5 separates the object inter- support for transactions and full interactivity for In-
faces from implementations. Besides providing for ternet operations. Also because of the stateless na-
language and platform independence, this approach ture of the protocol, there is only poor support for
enables an open specification supporting proprietary authentication and authorization. In any case, it is
implementations which facilitates competition and essentially impossible to interact with server objects
object reuse. Also part of the Object Management in near real time which is required for full collabora-
Architecture of CORBA are the various COR- tive capabilities. In addition, the limitations of the
BAservices, such as Naming, Event, Query, Transac- URL for locating information are profound since it is
tion and Persistence, Security, Versioning, and Life- typically merely a disk address and a machine name.
cycle. Each of these are important to the manage- Improvements in the identification of Internet data
ment of the distributed object infrastructure of value are in the works as result of efforts of the IETF and
to healthcare. In addition, there can be a number of the URN working group. The addition of Java to the
facilities, some of which would be special to health- Internet may result in a substantial alleviation of
care such as Medical Objects and a Master Patient some of these problems, although Java, by itself,
Index, although there may be common services un- does not deal with the interaction between the client
derlying similar facilities in other domains. Many of and server and is only as powerful as the protocol
the components of CORBA are commercially avail- underneath it. But it does substantially enhance the
able from a variety of vendors. The Object Request ability to interact with the data in a less constrained
Broker is the core of this technology which negoti- manner and is much more maintainable than CGI
ates connections between locations and provides the scripts.
support for the various services and facilities. The
interoperability of the ORBs with the Internet Inter- The linking of Java client applications through
ORB protocol (IIOP) has been demonstrated between CORBA8 to multiple servers provides a whole new
almost all of the commercially available ORBs. Ef- dimension to the management of healthcare informa-
forts are now underway to develop the components tion and support for a multimedia electronic medical
for the various vertical markets that will be managed record. Besides supporting a multi-tiered architec-
by these ORBs. ture instead of a mere client/server architecture it
provides major advantages in making it easy to dis-
A fundamental service of CORBA is the Naming tribute client software. The client infrastructure is
Service which manages the pointers to distributed provided completely by the Java Virtual Machine
objects. This is of significance to the Master Patient (JVM) or the Internet web browser, reducing the de-
Index as well as clinical lexicons and for providing veloper's task. The JVM also enables other lan-
the management of the variety of information sources guages to be supported by various compilers which
required to support the electronic medical record. extends Java's functionality considerably. In any
The Event service provides the foundation for case the flexible interface supported by Java allows
telecollaboration through the electronic medical rec- for advanced human computer interfaces to be deliv-
ord so that when objects or data are added to the rec- ered to the physician for ease of use.
ord, they can be seen simultaneously at multiple lo-
cations. The ability to fully interact with the medical TELEMED OVERVIEW
record from multiple locations is sustained by the
Transaction and Persistence Services which ensure TeleMed9 is a system developed jointly with National
data integrity both on the short term and the long Jewish Center for Immunology and Respiratory
term. If a multimedia data object is inserted into the Medicine in Denver, CO with almost all of the above
medical record, these services ensure that it is pre- principles in mind. It was designed with a clear sepa-
served for viewing by others when allowed. The ration between the underlying hardware, the services
CORBAsecurity service provides for full support of layer and the application layer, with the idea of sup-
secure authentication, authorization, and network porting a range of application domains. Both the
encryption of transmitted data while providing a non- services layer and the application layer are designed

366
with open distributed object technology. rity commercially provided as it becomes available.
CORBA/ORBs are used for all communications, and Fifth, we pay close attention making the system easy
the multimedia graphical interface has support for to use by the non-computer professional. The proto-
audio objects. Multiple servers can be used and the cols and procedures and captured in the client inter-
medical records from multiple Object Oriented Data- face so that the physician is dealing with information
bases can be combined through the "merged patient in a way that is entirely familiar. This principle with
record". The data at multiple remote locations can be major contributions from National Jewish Hospital
accessed logically as a single data set or serially with has resulted in the creation of the graphical patient
explicit control over the location. The CORBA in- record (Figure 2) which shows at a glance the longi-
frastructure has allowed us to create multiple servers tudinal medical history of a patient including various
which carry out various decision support functions reports, lab tests as well as a variety of treatments.
such as comparing a subset of images to see which Through this interface we also make available data-
ones are similar. These servers can transparently mining services such as comparing a CT image with
take advantage of parallel architectures for scalable other CT images which are deemed to be similar
concept extraction techniques. In addition, the indi- based on algorithms which extract and compare im-
vidual objects can have their own security and age features.
authentication so that different portions of the medi-
cal record can be made visible for different uses.
Finally, all the objects can come from multiple loca-
tions and thus support the fully distributed paradigm.
Figure 1 shows a diagram of the TeleMed architec-
ture illustrating some of these features.

Figure 2: Web-accessible, java-based, graphical pa-

tient record showing longitudinal patient history with
graphical references to remote data.
Figure 1: TeleMed 3 tiered architecture showing TELEMED DIRECTIONS
multiple persistent storage servers delivering a
merged patient record to a client along with auxiliary In order for the principles of TeleMed to be proven
servers for decision support. and shown practical for a wider audience, we have
undertaking several efforts to more broadly deploy
There are a number of principles that were used to TeleMed. First, we have joined with Northern New
guide the development of the TeleMed software. The Mexico Community College in Espanola, NM to
first one is to move data only as necessary. We pass provide TeleMed as a tool to link together 18 clinics
persistent references to the data which are used to in Northern New Mexico. It will be used to share
obtain the data or objects only when they are re- data such as immunizations, prescriptions and health
quested. This can radically reduce the bandwidth encounters between the clinics to more effectively
requirements for use. The second principle is to treat these rural patients. Using Java and CORBA,
manage complex high-volume data in an understand- we will be providing a simple electronic medical
able manner. That is, we want to give the illusion record over POTS. In addition, we have undertaken
that the whole data set is available, and can be navi- to evaluate and adapt TeleMed for us in the military
gated through in a straightforward manner. Third, we both in combat situations as well as in a few Tricare
require that the same architecture be used for real- regions. In order to provide the necessary linkage for
time consultations as for asynchronous consultations. wider use of the TeleMed technology, we have un-
That is, users can interact with the data simultane- dertaken to coordinate efforts with other organiza-
ously or provide their comments off-line. Fourth, we tions for the adoption of a wide area master patient
seek to leverage Internet standards and use them index which would enable healthcare providers to
wherever possible rather than develop our own. For more readily locate patient information. We also
example, we have developed our own CORBA secu- anticipate the use of TeleMed as a tool to manage
rity model but are replacing it with the CORBA secu- infectious disease information at the state level both

367
for the individual and aggregately. In addition, we
are working with the CORBAmed healthcare task- nual Fall Symposium (Washington, D.C., Oct.
force of the OMG to collaborate in bringing this ob- 26-Oct. 30). ed. J. J. Cimino, American Medi-
ject technology to the healthcare community. cal Informatics Association, Hanley & Belfus,
Inc., 1996, pp 608-612
For more information on TeleMed please visit our 7. McDonald, C. J., et al., The Regenstrief Medical
Web site: Record System (RMRS): Physician use for input
http: / /www. acl . lanl . gov/TeleMed. and output and Web browser based computing. in
Proceedings of the 1996 American Medical In-
formatics Association Annual Fall Symposium
CONCLUSION (Washington, D.C., Oct. 26-Oct. 30). ed. J. J.
Cimino, American Medical Informatics Associa-
Our experience with CORBA indicates that with the tion, Hanley & Belfus, Inc., 1996 p. 989
most recent CORBAservices, it is sufficiently mature 8. Orfali, R. and D. Harkey Client/Server Program-
to support the requirements of the healthcare indus- ming with Java and CORBA, John Wiley and
try. Indeed, multimedia patient records can be made Sons, Inc., New York, 1997.
a reality in a scalable system usable over a wide area 9. Forslund, D. W.. et al., Experiences with a Dis-
network, in which location information can be made tributed Virtual Patient Record System. in Pro-
transparent to the end user. Security and integration ceedings of the 1996 American Medical Infor-
of disparate systems to make such a distributed vision matics Association Annual Fall Symposium
a reality remain significant issues which will need (Washington, D.C., Oct. 26-Oct. 30). ed. J. J.
continual work. The former, as much from user ac- Cimino, American Medical Informatics Associa-
ceptance and ease of use as with actual protection of tion, Hanley & Belfus, Inc., 1996, pp 483-487
information. The latter involves pulling together
legacy data sources in a sensible way. Finally, our
experience indicates that data-mining techniques can
be provided in such a way as to be readily usable by
non-technical users.
Acknowledgments
This work was performed under the auspices of the
United States Department of Energy and partially
supported by the United States Army, Medical Ad-
vanced Technology Management Office of the U.S.
Army Medical Research and Materiel Command
References
1. Dick, R. S., and Steen E. B., Eds. The Computer-
Based Patient Record, An Essential Technology
for Health Care, Institute of Medicine, National
Academy Press, Washington, D.C., 1991
2. Phillips, R L., A Network-based Distributed, Me-
dia-rich Computing and Information Environ-
ment, Paper presented at the Digital Media and
Electronic Publishing conference sponsored by
the British Computer Society in Leeds, UK, De-
cember 1994. Available at:
http:/lwww.acl.lanl.gov/TeleMed/Papers/keynote.
pdf
3. Object Management Group, Ed.. Common Object
Request Broker Architecture, Wiley-QED, 1991.
4. Daniel, R, and M. Mealling, Resolution of Uni-
form Resource Identifiers using the Domain
Name System, an Internet draft specification:
draft-ietf-urn-naptr-04.txt, 1997.
5. Orfali, R., Harkey, D., Edwards, J., The Essential
Distributed Objects Survival Guide, John Wiley
& Sons, Inc., New York, 1996
6. Kohane, I. A., et al., Sharing Electronic Medical
Records Across Multiple Heterogeneous and
Competing Institutions. in Proceedings f the 1996
American Medical Informatics Association An-

368