It's a June morning in the cardiology ward at the brand new Georges Pompidou European Hospital in Paris. From one of the windows, you can see the top third of the Eiffel Tower. Inside the rooms, though, you won't see any bedside charts. Instead, Benoît Diebold, the attending physician, a resident, and two medical students go room to room equipped with a cellphone and a portable computer mounted on a wheeled cart. The computer's network cable trails awkwardly on the floor behind, and a patient standing in a doorway gives it a kick. Diebold announces cheerfully that he ordered a wireless link "yesterday."
Welcome to the (almost) digital hospital, a medical ideal where everyday operations and record-keeping are carried out and maintained almost exclusively with computers. At the heart of this ambitious model is the electronic patient record, or EPR. The idea behind it: to make all of a patient's medical reports, lab results, and images electronically available to clinicians, instantaneously, wherever they are [see ]. And, with an electronic record as a base, software tools that increase the accuracy of treatment can be added to improve patient outcomes and reduce cost.
Opened in August 2000, Georges Pompidou European Hospital, an 825-bed facility in Paris, was meant to be a showcase of 21st-century medicine. But its information technology system remains a work in progress.
EPRs have been a healthcare priority for years in many countries. The Institute of Medicine in Washington, D.C., the leading medical academy in the United States, has been pleading for an end to handwritten clinical data in the interest of safety and efficiency since 1991. But while the quality of the available information technology (IT) has improved greatly, progress in implementing EPRs has lagged behind, and only a minority of hospitals and clinics use them.
Considering what the needs of a modern hospital are, it isn't hard to understand why progress has been slow. Sharing data is fundamental to the concept of the EPR, but hospital medical departments, like surgery and intensive care, act like autonomous guilds, explained Charles Safran, chief executive officer of Clinician Support Technology, in Framingham, Mass., a leading provider of medical software. "Historically [they] haven't cared if their computers integrate with the hospital-wide information systems," he said. On top of that, issues of security, privacy, and confidentiality have not been addressed in any decisive way. Yet, easy access to electronic records holds the potential to make these data a more meaningful resource beyond the settings where they were collected and where, traditionally, they would remain.
As companies and hospitals try to design the software to implement and use EPRs, it is still unclear who will decide what information should be available, to whom, and how. For example, medical staff require a detailed chronological medical history to assess the status and needs of each patient. But researchers want to go beyond this level and integrate data about populations and subgroups that could contribute to medical knowledge. In another area, regulatory bodies and professional accreditation reviewers require documentation of patient care. And insurers of individual patients are interested in their genetic and medical histories.
Undeniably, balancing the rights to privacy and legitimate access will be no mean task. But perhaps the most surprising hurdles to implementing EPRs are the human factors. Finding a comfortable interface between people and machines (that is, making it easy for doctors and nurses to enter and use data) is a continuing challenge. In addition, financial and productivity incentives for adopting EPRs are not always clear, and training staff to use them is time consuming, expensive, and often must be customized for the peculiarities of each hospital.
Several leading teaching hospitals already have EPRs, developed mostly through years of in-house work, while others are trying to build a digital hospital from scratch. One example of the second group is a venture between HealthSouth Corp., the largest U.S. healthcare services provider, and database-supplier Oracle Corp., Redwood Shores, Calif. The pair recently announced plans for "the world's first all-digital, automated hospital," to be completed in 2003 near HealthSouth's Birmingham, Ala., headquarters.
The technical innovations in the HealthSouth/Oracle venture sound remarkably like those the Pompidou European Hospital is trying to implement. But the Pompidou already exists, and by that fact it provides a valuable object lesson on the struggle to digitize healthcare. Following a calamitous first year and a half of operation that included everything from strikes to two fatal outbreaks of Legionnaires' disease among its patients, the digital hospital and the EPR at its core are still a work in progress.
Pompidou, an 825-bed teaching hospital that replaced three outmoded facilities in southwest Paris, was envisaged as a showcase for 21st century medical care and research. Symbolic of the hospital's ultramodernity is the pioneering work that Alain Carpentier is doing in the cardiac surgery unit, using a US $1 million remote-controlled robot to perform minimally invasive heart repair operations. Elsewhere in the facility, robots sort and perform chemical analyses on routine specimens in an automated laboratory.
But the Pompidou was intended to break new ground in a more fundamental manner. As in other countries, medical departments in France have traditionally functioned as baronies, with each unit doing its own separate medical record-keeping and scheduling for patients. The Pompidou changed that policy by developing a physical setup that encourages interactions among staff, and an IT infrastructure that supports the flow of information. The computing system was designed to retrieve all the information related to a patient from wherever it originates, whether in the emergency room, laboratory, or outpatient clinic, and to make all the data available at the point of care.
The EPR, however, was not meant to be just a ho-hum text record of events, but a multimedia report that includes digitized images and video and, eventually, genomic data. The record will also be integrated with decision-support tools for clinicians, such as alarms that signal a patient's allergy to a drug a doctor tries to prescribe, or pop-up pointers with information for treating certain conditions. In time, doctors outside the hospital will be able to access the EPR to provide continuity of care once the patient is released.
Patients themselves will benefit from such a record. An ultimate aim of the EPR is to provide them with the information they need to cooperatively manage their own health, most notably through direct access to their medical record, as mandated by the French government; exactly how that will happen, though, remains to be worked out. Once patient access is available, the EPR promises to be a large factor in establishing a truly integrated health information system.
The ghost of systems past
Since its appearance in the 1970s, hospital computing has been primarily about billing, with little attention paid to the needs of the clinician. But there were some notable exceptions. By the mid-1980s, according to Clinician Support's Safran, the Diogène system at the Cantonal Hospital of Geneva, in Switzerland, and the Center for Clinical Computing system at Beth Israel Hospital, in Boston, had developed into programs that were no longer limited to accounts and billing. Both systems provided access to laboratory results, medication-dispensing information, on-line text of discharge summaries, and more.
What was missing, Antoine Geissbühler, head of medical informatics at the 2200-bed Cantonal Hospital, told IEEE Spectrum, was "putting care providers and physicians in the loop at the time that they make decisions." The records were archival, rather than interactive. According to Geissbühler, healthcare providers are now beginning to use the computer to get advice and to make decisions. "That's a major change in culture," he said.
Geissbühler and his team are retrofitting their enhanced EPR and decision-support tools onto the hospital information system they inherited. But such homegrown systems take time and money to develop. That's why Patrice Degoulet, head of medical informatics at the Pompidou [see photo], chose a commercial best-of-breed solution — a collection of the most effective software for each particular task — in constructing the Pompidou's IT infrastructure from scratch.
To solve what he knew would be a huge integration problem, Degoulet formed a partnership with Alcatel/Thomson (now Thales Information Systems, Malakoff Cedex, France, an information systems integrator). The company would supply and marry all the clinical software components — ordering of treatments, scheduling of X-rays, and so on. For financial, administration, and laboratory analysis, Degoulet kept the software from the old French hospital system, referred to as AP-HP. Separate contracts were negotiated for the radiology information system and its image-display components.
In setting up the IT infrastructure, Degoulet followed the popular three-tier architecture, consisting of data at the bottom, a flexible user-interface at the top, and business logic and process services, or middleware, in between [see illustration]. The data, middleware, and files are distributed over some 40 Hewlett-Packard servers running under either Unix or Windows NT. The software applications themselves are kept on client PCs, except for a few programs, such as the imaging software. The exceptions reside on servers with access by a thin client, software that relies heavily on the server, installed on the PCs.
Thales created the middleware program, known as DOM-H (DOM stands for "dossier médical"; the H is for hospital), to glue the applications together "like Lego blocks," as Thales' Serge Castaigne described it. DOM-H includes components for security, a single sign-on feature (so users of the system have to get authorized only once instead of at each application within the EPR), and a standard called Clinical Context Object Workgroup (CCOW). CCOW allows a user to pass seamlessly from one application to another without losing track of the patient. DOM-H uses a Web-type interface, although once a user has entered the system, each individual application maintains its appearance and mode of interaction.
A 622-Mb/s asynchronous transfer mode (ATM) optical-fiber backbone feeds the Pompidou's local-area network (LAN). Joining the LAN with the PCs and imaging workstations are 10- or 100-Mb/s Ethernet connections. The hospital is linked by a router to the private network of the AP-HP hospitals, which is itself connected to the Internet.
Working with the EPR
Signing on to the Pompidou's hospital information system requires a login and password. A security component keeps track of who is entering the system, at what time, and on which terminals. In recognizing a user, the system authorizes appropriate functions, including what part of the EPR the person may read, change, or delete (although nothing is ever deleted for good). Receptionists at the front desk, for example, have access only to a list of patients and their location in the hospital; physicians are entitled to the full range of functions.
Since the goal of the EPR is to capture everything that happens to a patient in the course of his or her hospital stay, information entered and stored includes admission, discharge, and transfer (ADT in the parlance); laboratory orders and results; routine clinical notes by doctors and nurses; and vital signs. Typing a patient's last and first names and date of birth in the DOM-H main page pulls up a complete history of prior consultations and hospital stays for that patient; the data may be displayed according to the medical problem, type of exam, surgical procedure, or in other ways.
Ordering an X-ray in the main patient management software (Dx-care by Medasys Digital Systems) automatically triggers a central scheduler (One-Call by Per-Sé Technologies), which notifies the radiology information system of the appointment. Reports and images are forwarded to the picture archiving and communication system (PACS) server, and physicians may call them up wherever they are in the hospital.
Clinical decision-making meshes general principles of medical management with the specifics of a particular patient. The EPR makes it possible to relate one to the other.
One area in which the relationship is important is in ordering medication. According to a 1999 U.S. Institute of Medicine report titled To Err Is Human: Building a Safer Health System, medication errors in and out of U.S. hospitals result in more than 7000 deaths annually. One study done at Massachusetts General Hospital and Brigham and Women's Hospital, both in Boston, showed that preventable adverse drug reactions raise costs by $2.8 million per year for a 700-bed teaching hospital.
Decision-support systems can change those numbers by, say, having the computer signal an inadvisable choice of drug, dosage, or type of examination, in accordance with information in the EPR. However, physicians within a hospital, let alone across the discipline as a whole, do not always agree on the rules under which reminders and prompts should occur, and that makes such systems difficult both to build and to make broadly acceptable.
Dealing with Dr. No
Software systems that lower cost and improve patient care are no good unless physicians actually use them, and research has shown they do not always do so. In December 2001, the British Medical Journal reported that 227 doctors in 19 Norwegian hospitals used electronic patient records systems mostly just to read patient data, and that they "used the systems for less than half of the tasks for which the systems were functional." Possible reasons the study's authors cited include a lack of available computers, the flexibility of paper records, and local resistance to disruption of traditional work routines.
So the first step in implementing EPRs and the tools that use them is to get physicians to feel comfortable using computers. One way is to make computers available wherever physicians need them. The Pompidou has roughly 2000 PCs distributed throughout the hospital. Ward stations are all equipped with PCs, four for every 24-bed ward. In addition, every outpatient examination cubicle has a computer.
To facilitate bedside capture of information, Degoulet settled on laptop computers affixed to wheeled carts that can be shared by both doctors and nurses. What began as a pilot program has now expanded so that 85 portables are deployed, about one for every 10 beds. Cantonal Hospital in Geneva has also gone with portables, and is evaluating other devices.
But though bedside capture of clinical information features in most scenarios of the digital hospital, not everyone is pushing it. J. Michael Fitzmaurice, a senior science advisor at the Agency for Healthcare Research and Quality, a U.S. government entity in Rockville, Md., noted that it is not yet even clear, from a cost-benefit point of view, that physicians would be the best people to have at the keyboard. And there are other factors to consider. "Patients really find it distracting for a doctor to use a computer while in the room during a visit," said John Halamka, chief medical information officer of CareGroup Healthcare System in Boston. CareGroup comprises six Harvard University hospitals and has had "entirely electronic records since 1989." Doctors do their data entry and ordering at the nurses' station once they have left the room.
Training staff to use an EPR is a big part of the job of implementing one. For two years before the opening of the Pompidou in 2000, Degoulet assembled a team of people who would be using the EPR — physicians, nurses, secretaries — to decide a policy of access and to customize the software to include hundreds of medical protocols, lists of tests, and drugs peculiar to the Pompidou. A program to train trainers who would then train end-users preceded the ribbon-cutting.
Instead of establishing a complex hierarchy of enabled functions in the EPR, the philosophy of the Pompidou was to give every user some basic functionality and to build on the parts that were working well [see illustration]. It was a compromise that has proved frustrating for more sophisticated users, but has made it possible to bring the 15-20 percent of users who had never worked with computers more quickly into the system. Presently, users can enter and access ADT data, order investigations such as X-ray and ultrasound, access images and obtain results, and enter daily notes on what a doctor observed and found and how he or she made decisions. Electronic prescribing is due to begin soon at the Pompidou, as a pilot program, and will gradually be expanded over a year.
Success is quantified as the percentage of the computer system's functions used by the people responsible for the corresponding hospital task. So, whereas computer retrieval of the results of imaging studies and laboratory tests is undeniably working (used by 100 percent of physicians in 51 out of 51 departments), only a few doctors in a few units enter routine clinical notes on the computer. About 70 percent of clinicians type in their own laboratory orders, either at bedside or at the ward stations — a figure Degoulet is happy with because a year ago, it was 0 percent. Clinicians who do not want to use the computer can fill out a computer-readable form that, for instance, accompanies a bar-coded test tube to the laboratory. The information is then entered into the computer at the lab.
More to come
After years of work, software for EPRs, and for hospital information systems generally, is still not optimal. Like audiophiles, defenders of best-of-breed solutions equate components and quality. Detractors argue that patchwork systems, like the Pompidou's, have too many points of failure. The problem is that elite systems, like the one at Brigham and Women's Hospital and another used by the U.S. Veterans Health Administration hospitals, require sustained effort and resources beyond most hospitals' strategic planning horizon.
Until very recently, no one copied these elite systems, because hospitals are inherently local so their software relies on vocabulary and workflow peculiar to each [see "You Have to Have Standards "]. Consequently, the market for EPR software is still young. According to Paul Brown of Britain's Center for the Advancement of Electronic Health Records, in Chepstow, Monmouthshire, most EPR software companies "don't stay [in business] long enough for anybody to count them." If the Pompidou can pull it off, said Geissbühler, "it will mean the industry is ready."
While the jury is still out on the EPR industry, the Pompidou is forging ahead. The trailing cables on the portables are about to be replaced by an 11-Mb/s IEEE 802.11b wireless LAN. The ATM will give way to a gigabit-per-second Ethernet, and here and there the system is being fine-tuned to solve nagging performance problems that have complicated the staff's acceptance of the system.
Degoulet compared the digitizing of Pompidou to being two-thirds of the way across a stream: you are not yet across, and it is too painful to go back. "Maybe if we build 10 Pompidou hospitals like this one," he volunteered, "the 10th one will work. But first we have to build one, and it will not work so quickly.
Samuel K. Moore, Editor
About the Author
GISELLE WEISS, based in Basel, Switzerland, writes for The Economist and The Times of London.
To Probe Further
One of the most recent studies on the use of electronic patient records is "Doctors' Use of Electronic Medical Records Systems in Hospitals: Cross Sectional Survey," by Hallvard Lærum and others in the British Journal of Medicine, 2001, No. 323, pp. 13441348 (see http://bmj.com/cgi/content/full/323/7325/1344).
For an overview that describes progress in both the United States and Europe, see the revised edition of The Computer-Based Patient Record: An Essential Technology for Health Care, edited by Richard S. Dick, Elaine B. Steen, and Don E. Detmer for the Institute of Medicine (National Academy Press, Washington, D.C., 1997). It's available free on-line at http://www.nap.edu/catalog/5306.html. To Err is Human, mentioned in the article, is also available at this site.
The technology behind implementing electronic patient records and decision support systems is a frequent topic of reports in IEEE Transactions on Information Technology in Biomedicine. It's available on-line via IEEE Xplore at http://ieeexplore.ieee.org/lpdocs/epic03/RecentIssues.htm?punumber=4233.
For details on the state of standards, see "Data Standards in Health Care," by Clement J. McDonald, Gunther Schadow, Jeff Suico, and J. Marc Overhage, writing in the Annals of Emergency Medicine, Septermber 2001, Vol. 38, pp. 303310.