Advertisement

Electronic Health Records: Promises and Realities

Part II: Some Early Voyages in Partially Charted Waters

      Concessions to Posterity

      The next few years—with subsidies for accelerated electronic health record (EHR) adoption coming at the same time as the national venture into health insurance reform—may be a watershed of sorts. American medical informatics has the chance to pursue the capabilities foreseen for hypothetical Hospital A, its best-case scenario of added value. With the EHR market in a state of high sensitivity to initial conditions, to borrow a term from chaos theory, it also faces the risk of “lock-in” to a scenario closer to that of Hospital B, where EHR is a cumbersome albatross. (See part I for a full description of Hospitals A and B). These systems are extraordinarily complicated to design; the harmony or dissonance between their structure and purpose may make them either a potent force multiplier in efforts to upgrade the health care system's performance or a millstone dragging those efforts down.
      The myriad investments of clinicians' time that add up to a useful EHR, comments Mark D. Olszyk, MD, MBA, deputy chief of staff at the VA Maryland Health Care System and chairman of the National VA Advisory Committee for Emergency Medicine, express a value choice at odds with the immediate incentives present in highly fragmented institutional settings. “Any medical record, paper or electronic or otherwise, is a concession to posterity,” he says. “I would submit to you that the medical record rarely benefits the provider at the time they're writing it. And [in] a lot of the debate and a lot of the groaning … practitioners say, ‘Hey, you're building this record, you're forcing us to do this, and it's ever-increasing in scope and magnitude and complexity, but you're building it on our backs.'”
      The record's eventual benefits in strengthening continuity of care, facilitating research, or providing “lawyer garlic” generally accrues to others, whereas “everyone who has some input has no real skin in the game.” Consequently, in a well-planned system the record-generating clinicians and staff should either share professionally in those benefits or be able to generate data with a minimal burden on their immediate operations.
      Incentive structures apply on the vendor side as well. George Hripcsak, MD, MS, department chair and professor of biomedical informatics at Columbia University and cochair of the HHS Health Information Technology Policy Committee's workgroup on meaningful use, contends that the current stimulus is just what the field needs to provide suitably scaled incentives for system developers to flesh out that key concept through purposeful design. “I think that one of the things that has stifled innovation in HIT [Health Information Technology],” he says, “is the small size of the market, with such little adoption [that] there's not enough capital in the system to do big R&D.” Until there is a strong network effect that adds increasing value for each new participant and generates visible benefits for new adopters and developers, he notes, the burdens of building an EHR system outweigh the potential advantages, “like having a telephone in a town where you're the only one with a telephone: there's no one to talk to.”
      The Health Information Technology for Economic and Clinical Health Act states objectives instead of specifying system characteristics, Dr. Hripcsak says; by mandating a quick interval for adoption, it can motivate vendors. The Certification Commission for Health Information Technolog, in this regard can perform a coordinating function like that of Underwriters Laboratories in product safety, the American National Standards Institute in engineering standards, or the US Green Building Council in sustainable construction. “Done right, standardization can promote innovation,” Dr. Hripcsak comments, “because then every little company doesn't have to reinvent the wheel; they can just invent their innovative part and plug it into the larger infrastructure.”

      Regional Information Exchanges: Herding Competitive Cats

      Some advances tailored to the needs of emergency departments (EDs) involve the work of one of Dr. Hripcsak's protégés, Jason S. Shapiro, MD, MA, assistant professor of emergency medicine at New York's Mt. Sinai School of Medicine. Dr. Shapiro has been active in the development of a regional information network, the New York Clinical Information Exchange (NYCLIX), taking on one of the more demanding tasks in the field: getting a group of 12 unaffiliated provider organizations, in a catchment area served by numerous institutions rather than a single major group or a small number of them, to share information in ways that are beneficial to patients and to the community but not to each of the competing corporate entities. The H1N1 influenza outbreak in spring 2009, which hit the New York area hard, provided an informative use case for NYCLIX's emergency response capabilities, as reported earlier.
      • Shapiro J.S.
      • Genes N.
      • Kuperman G.
      • et al.
      Health information exchange, biosurveillance efforts, and emergency department crowding during the spring 2009 H1N1 outbreak in New York City.
      Health information exchanges such as NYCLIX are a useful intermediate step, Dr. Shapiro points out, between the current prevalent situation, with information siloed inside individual proprietary systems, and a truly interoperative national-scale or even global-scale network. As an independent not-for-profit organization, an exchange can broker the sharing of clinical data among institutions that share the goal of improving quality and efficiency but do not share a financial bottom line. “It's been 5 years now that we've been doing this,” Dr. Shapiro observes, “and none of the organizations has said ‘This is a bad idea; we don't want to do this, because we take a proprietary stance on our data.' The organizations are the custodians of the data …. [T]he patients own the data.”
      Opinions differ on whether a phased rollout or an all-at-once turnkey launch is preferable, Dr. Shapiro adds, but in hospitals that have designated certain departments as the first canaries to enter the EHR coalmine, the ED is often a good place to start. The time-sensitive and information-intensive aspects of emergency practice make EDs ideal settings for the workflow efficiencies that well-designed EHRs allow. (The fact that emergency patients do not generally choose where ambulances take them, he says, has also “helped calm a lot of these competitive concerns” within the NYCLIX group.)
      As NYCLIX's clinical advisory committee chair, Dr. Shapiro convened emergency physicians in 2005 to discuss data needs and optimal presentation formats. Emergency physicians, Dr. Shapiro notes, have been most interested in ECGs, medication lists, and discharge summaries, along with text reports of advanced imaging studies (radiologists and cardiologists most likely prefer the actual images from magnetic resonance images, computed axial tomography scans, cardiac catheterizations, and ultrasonography). A role-based user interface based on these priorities and New York State Department of Health recommendations (including a “break-the-glass” functionality for emergency data access when a patient lacks the capacity to give consent, limited by state law to the ED setting) is being piloted among approximately 100 physicians at 3 sites. Studies measuring the system's effect have begun, and early anecdotal reports describe the exchange as an effective way to reduce duplicated tests and unnecessary admissions.
      Curiously, Dr. Shapiro reports, EDs, in contrast to ambulatory providers, are not listed among eligible providers for the stimulus funds being made available from the Centers for Medicare & Medicaid Services through reimbursement changes. “The way that emergency physicians will most likely be affected by this large bolus of money,” Dr. Shapiro says, “is in their hospitals deciding to adopt an EHR for the first time. I think there've certainly been plenty of incidences of hospitals choosing a system based on aggregate need, and I would say the ED's voice needs to be heard in that part of the process to make sure that the specific needs of emergency medicine are being met.”
      The systems offered by the estimated 400 vendors currently crowding the field vary widely, and none can be installed without some degree of customization. “If you've seen one implementation,” remarks Dr. Shapiro, “you've seen one implementation.” They can either smooth or impede workflow. Considering the heterogeneity of both the systems and the institution-specific clinical practices they are meant to serve, he adds, the setup of an EHR system should be an opportunity for clinicians to engage in process redesign, analyzing the elements of a department's practice, “figuring out what makes sense and what maybe has just been built up in a patina over years of [saying] ‘this is how we do it.'”

      Veterans Health Information Systems and Technology Architecture: Not Built in a Day, Perhaps to Be Rebuilt

      The EHR system in the United States with the deepest roots and best-documented benefits is difficult to hold up as a general model in all respects because it is inseparable from the scale and unique culture (not to mention the single-payer financial advantages) of the Veterans Administration (VA).
      • Millard W.B.
      Emergency medicine in the VA: the battleship is turning.
      It is also, by the digital world's evolutionary standards, a dinosaur. Still, the experience of physicians, developers, and patients with Veterans Health Information Systems and Technology Architecture (VistA) is an indispensable part of today's EHR debates. Both by serving as a general proof of concept and by bringing some of the inherent problems of EHRs to light, VistA helps point toward solutions.
      Tom Love, PhD (a cognitive scientist; former research director for several major corporations, who worked on the development of the first fully distributed digital telephone switching system for ITT; and cofounder and chief executive officer of the informatics firm ShouldersCorp, currently consulting on efforts to update VistA), views its development as an artifact of its history and in some ways a cautionary tale. “I've often accused the VistA system of looking like a Ferrari on top and a '57 Edsel underneath. It provides functionality—much-needed, wanted, sometimes referred to as much-loved functionality—but from an engineering perspective, underneath it's an absolute screaming disaster.” Originating in the era of UNIX-style interfaces with scrolling command lines on green screens (“It doesn't take you quite back to console model radios and vacuum tubes,” he comments, “but it was back in a period when people still remembered those things and might have had one or two in the living room”), VistA now has a Delphi graphic interface layer placed atop the older code, which still reflects technical choices considered reliable in 1978. Dr. Love calls it “a system that actually works quite well, but it has this defect that it's a few generations of technology back,” too complex and inflexible for the expansions that would allow state-of-the-art interoperability.
      “Probably first witnessed inside the VA,” he continues, “was the fact that their systems had grown up originating as an inpatient system, and then the nature of medical care changed.” A builder of enormous bed towers in the post–World War II expansion era, the VA has converted much of its space to clinics and admits far fewer patients than before. Adapting inpatient-oriented record systems for a distributed practice model involving more work on an outpatient basis was one difficulty, Dr. Love says. Moreover, neither inpatient nor outpatient systems ideally serve the distinct information needs of EDs, where unscheduled patient loads, urgent time constraints, and unpredictable triage-driven queuing patterns often make speedy access to essentials preferable to comprehensive system reviews. “Conventional technology had the property of seeming intolerably slow to an ED surgeon, for example. Two-second response time is painfully slow in that world; in other worlds it's perfectly acceptable.”
      Dr. Love praises VistA's depth of detail, recording “literally every blood pressure reading” and other clinical variable, for enabling discoveries through analysis of aggregate evidence. “The optimal way to store the information to give doctors real-time access,” however, “might be exactly the opposite way that you need to store the information in order to be able to do the analysis.” Data arranged in a hierarchic binary-tree format (allowing rapid retrieval of a datum about a single patient) and relational data allowing broader epidemiologic analyses (eg, “for all the blond-haired blue-eyed guys that served in this region of Vietnam in 1968 and that were subjected to Agent Orange, what medical problems have they presented with?”) are qualitatively different; VistA's polymorphic FileMan database system allows both, but at the cost of duplication and inefficiency. It is said to be “easy to get data in, hard to get data out.”
      Dr. Love's colleague Robert A. Stults, MS, director of ShouldersCorp, trained as an architect and then spent a decade performing media research at Xerox's fabled Palo Alto Research Center before consulting with a range of organizations on telecommunications, security, and business processes, culminating in work on VistA during the past 5 years. He emphasizes the interdependence of information structures and physical phenomena, including day-to-day hospital operations. “The VA is not billing driven,” Dr. Stults notes. “There are billing components in it, but it is very strongly care driven and physician driven,” with a “strong heritage of being longitudinal.” By emphasizing long-range care over immediate aims, VistA is an embodiment of Dr. Olszyk's “concession to posterity” principle. “When you're asking a provider … to spend time and do a patient record,” Dr. Stults says, “you are not asking them from the point of view of a private-sector provider—‘you need to do this so that you can establish codes, so you can get billing.' You're doing this so that the next person down the line is going to be able to know what happened.” This evokes relatively little resistance in an integrated organization such as the VA, but any setting in which encounters are isolated and patients may never reappear (eg, many US EDs) places a premium on ease of recording.
      Dr. Stults acknowledges another conceptual continuum expressed through system design. At one end is the position that operations and EHR are separate functions, connected only through deliberate manual effort, as in most current systems. At the other end, record creation is integrated with hospital operations, minimizing the problem of an immediate “tax” for the record creator and an indirect or deferred benefit. In an optimal work environment, as in the simulators that he studied and developed at Xerox PARC, virtual spaces and inhabited physical spaces are overlaid. Technologies such as direct data input from clinical instruments, the infrared radio frequency identification tracking devices suggested above, and conjectural telemedicine applications connecting bedside operations with distant decisionmakers, Dr. Stults says, “create the record as a consequence of providing the care.”
      Given its orientation toward long-term care, VistA in its present form would be hard to adapt for non-VA institutions. However, it may soon be reworked in ways that could make it more broadly exportable. A working group convened by the American Council for Technology's Industry Advisory Council
      American Council for Technology Industry Advisory Council
      VistA Modernization Report: Legacy to Leadership.
      has recently unanimously recommended an open-source, open-standards re-engineering of VistA, Dr. Love reports, replacing essentially every line of code while retaining the business rules, workflows, and screen definitions. “The underlying vacuum tubes really need to be replaced with semiconductors,” Dr. Love says, transforming this legacy system into a standards-based “ecosystem” capable of wider deployment, easier maintenance, and smoother transactions with commercial systems. VistA 2.0, if rebuilt along these recommendations, may fulfill the council group's aim of being “offered up as the international standard information system for medical centers,” not only saving the VA money but also supporting “advances in evidence-based medicine, medical research, and data standardization and portability.”
      Such a revision is subject to the usual budgetary uncertainties and is by no means guaranteed. It was only 3 years ago that a contract for a Cerner proprietary laboratory information system had some industry observers speculating that the VA might be turning away from VistA altogether.
      • Blankenhorn F.
      Is the VA going proprietary?.
      “In the VA,” notes Dr. Olszyk, “information technology does not work under the same umbrella as the health administration. IT now reports directly to the Secretary. That was done so they would be their own department, have a more streamlined budget … but now they're a little bit further away from the frontline people.” Clinicians' input is vital for decisions with such far-reaching implications, inevitably made in shifting political contexts.

      In Indiana, the Data Switzerland

      That medicine lags so far behind other fields in embracing information technology remains puzzling. “If Wal-Mart knows where every pallet of widgets is in their supply chain,” asks Dr. Shapiro, “and FedEx or UPS can come to my door to drop off a package and have me sign a wireless keypad and immediately take my item out of their database, why don't I know when Mrs. Jones went down the hall for a CAT scan?”
      He is far from the first to observe this incongruity. When Samuel Regenstrief, an operations engineer and Indiana's “dishwasher king” in the 1960s and 1970s, grew old enough to have frequent encounters with the health care system, recounts John T. Finnell, MD, MSc, FACEP, an associate professor of emergency medicine at Indiana University and research scientist at the Regenstrief, he was baffled at how little his cardiologists and internists knew of one another's roles in his case. Assuming the role of a philanthropist, he looked to technology to resolve information bottlenecks in medicine as it was already doing in industry. With his support, a small endocrine clinic expanded into a regional information coordinator and research organization affiliated with Indiana University School of Medicine and focusing on health services research, geriatrics, and informatics.
      The Regenstrief Institute does not provide health care itself; it provides information and related services to care providers. This role has been beneficial in an environment in which “everybody wants to do the right thing, but they also want to get credit for it,” Finnell says. “When you've got 2 competing institutions, how do both get credit and share the limelight? I think that's one of the wins that Regenstrief has benefited from: that we've been known as the Data Switzerland.”
      Greater Indianapolis hospitals and patients now benefit from the Regenstrief Medical Records System (RMRS), a clinical data repository built throughout 3 decades with a special-purpose programming language originally known as Arden syntax, developed by Columbia artificial intelligence researchers in the early 1990s. This decision support framework and a later iteration known as Gopher Care Rules use “medical logic modules” to link queries to appropriate clinical recommendations. With the aid of the RMRS, clinicians can take a focused approach to testing rather than a shotgun approach. Other gains have come from common-sense rethinking of mundane operations, such as cheap electronic delivery of laboratory results rather than parcel-posting them en masse to providers at 20 or 30 cents per item. In one early report, access to the Regenstrief repository saved about $26 per ED visit in one of 2 hospitals studied.
      • Overhage J.M.
      • Dexter P.R.
      • Perkins S.M.
      • et al.
      A randomized, controlled trial of clinical information shared from another institution.
      Finnell's research involves a special data infrastructure for ED visits; observing that patients with more than a single ED encounter often visit multiple EDs in the region,
      • Finnell J.T.
      • Overhage J.M.
      • Dexter P.R.
      • et al.
      Community clinical data exchange for emergency medicine patients.
      he and his colleagues have built a community-wide data exchange that removes artificial barriers between institutions and between specialties. “Our residents,” he says, “when they leave here and go to practice in other regions of the country, [ask] ‘That thing we had in Indiana: how do we do that here?'… It started in the ED, and as patients got admitted to the hospital, the hospitalists are saying, ‘How do you guys know that? We don't have access to that data set; how do we get access?'”
      Finnell has also joined colleagues from multiple institutions in exploring ways of marshaling clinical decision support technologies to disseminate American College of Emergency Physicians practice guidelines.
      • Melnick E.R.
      • Nielson J.A.
      • Finnell J.T.
      • et al.
      Delphi consensus on the feasibility of translating the ACEP clinical policies into computerized clinical decision support.
      Another area in which Regenstrief personnel are doing pioneering work is the extension of RMRS data into the field: paramedics and emergency medical technicians affiliated with Indianapolis-based Wishard Health Services, equipped with ruggedized laptops, are the nation's first emergency medical services group to have access to patients' EHRs.
      A common thread among these efforts is that they move emergency care and the associated information services outward into the community, taking these services where people are instead of awaiting their arrival at particular sites. Despite the widely held assumption that centralized data resources inevitably foster top-down control and replacement of human judgment by machine logic, the effect of all this connectivity has been to support a distributed and flexible model of practice. How it might serve a population whose ties to particular physicians, hospitals, insurers, employers, and locations have been steadily weakening during recent years is a separate discussion.

      The Form of a Parallel Universe

      The VA's Dr. Olszyk offers 2 distinct metaphors for the current state of medical informatics and possible future arrangements. One involves transportation; the other, construction. At present, with several protocols under development but no uniformly accepted EHR standards, it is as if “every city had its own air-traffic-control language,” requiring geographically specific signals for planes to avoid crashes and execute landings. To cut down on errors and waste, medical informatics needs a common language so that it resembles the aeronautical system, in which “we pass you off from Cincinnati tower to Sioux Falls tower to Los Angeles tower [and] it all goes seamlessly, and the risk of plane crash from miscommunication is vanishingly low.”
      Like many commentators on information systems (whether or not they are, like Dr. Stults, actual architects), Dr. Olszyk uses architectural images to describe the virtual structures he would like to see implemented. He quotes a well-known aphorism from Winston Churchill
      • Churchill W.
      speech to House of Commons (meeting in the House of Lords), October 28, 1943.
      : “We shape our buildings, and afterwards our buildings shape us.” “That's what it is with the medical record,” Dr. Olszyk observes: “we're having all these people design this architecture, and we move in and we don't know what to make of it. Is it a cathedral? a factory? a house? And the people who design it aren't necessarily living in it.” A strong implicit argument, presumably, for the people who will live in it to be the ones guiding its design.

      References

        • Shapiro J.S.
        • Genes N.
        • Kuperman G.
        • et al.
        Health information exchange, biosurveillance efforts, and emergency department crowding during the spring 2009 H1N1 outbreak in New York City.
        Ann Emerg Med. 2010; 55 (doi:10.1016/j.annemergmed.2009.11.026. Available at: http://www.annemergmed.com/article/S0196-0644%2809%2901800-9/fulltext. Accessed June 4, 2010): 274-279
        • Millard W.B.
        Emergency medicine in the VA: the battleship is turning.
        Ann Emerg Med. 2008; 51: 632-635
        • American Council for Technology Industry Advisory Council
        VistA Modernization Report: Legacy to Leadership.
        American Council for Technology, Fairfax, VA2010 (Accessed June 6, 2010)
        • Blankenhorn F.
        Is the VA going proprietary?.
        ZDNet Healthcare. November 12, 2007 (Accessed June 6, 2010)
        • Overhage J.M.
        • Dexter P.R.
        • Perkins S.M.
        • et al.
        A randomized, controlled trial of clinical information shared from another institution.
        Ann Emerg Med. 2002; 39 (doi:10.1067/mem.2002.120794. Available at: http://www.annemergmed.com/article/S0196-0644%2802%2945529-1/fulltext. Accessed June 6, 2010): 14-23
        • Finnell J.T.
        • Overhage J.M.
        • Dexter P.R.
        • et al.
        Community clinical data exchange for emergency medicine patients.
        AMIA Annu Symp Proc. 2003; 2003: 235-238
        • Melnick E.R.
        • Nielson J.A.
        • Finnell J.T.
        • et al.
        Delphi consensus on the feasibility of translating the ACEP clinical policies into computerized clinical decision support.
        Ann Emerg Med. April 5, 2010; (doi:10.1016/j.annemergmed.2010.03.006. Available at: http://www.annemergmed.com/article/S0196-0644%2810%2900223-4/fulltext. Accessed June 6, 2010)
        • Churchill W.
        speech to House of Commons (meeting in the House of Lords), October 28, 1943.
        (Accessed June 6, 2010)