Clinical information systems in the academic medical center.

For the past two years, as Medical Director of both Clinical Information Systems and the Critical Care Medical Unit at UMMC in Ann Arbor. it has been my responsibility to develop, introduce, and implement clinical information system strategies. Working half-time both in the department of medicine and in information and networking services has provided an opportunity to identify the clinical roots of the challenges to introducing new information systems in our environment. Several years ago, we began the development of a new strategic plan for clinical information systems with a fundamental belief that a clinical information environment should be:

* Patient-centered.

* Integrated.

* Directly used by physicians,

nurses, and allied health workers. At this point, the challenges to change in our environment began to appear. The first of these was the framework of the environment itself--the academic infrastructure.

Academic Infrastructure: Balancing on the Three-Legged Stool

The missions of an academic medical center--education, research, and patient care-have been represented as a three-legged stool, symbolizing the importance of maintaining the balance among these elements lest the institution lose its stability. The reality, however, is that university medical centers are not perfectly balanced in these three elements. This heterogeneous mixture leads to important practical considerations in the introduction of clinical information systems. If the strategic plan overemphasizes one element of academic operation in preference to the others, acceptance and interest become serious problems.

So it was at the UMMC when we first began to develop our plan for clinical information systems. The initial vision recognized the fact that a great deal of clinical computing was already in progress at the medical center. Excellent departmental systems in pathology, radiology, pharmacy, and other departments were in place. An identified problem was lack of integration of these systems for the clinical user, particularly for supporting the process of daily patient care. Additionally, the existing environment provided results reporting but little from the order communication standpoint. Hence, the first planned change in the clinical information environment was an integrated patient care information system designed for direct clinician order entry as well as results reporting. Additionally, considering the complexity of UMMC, we wanted to select an existing system that would meet our needs rather than add the variables of a total developmental project that might never deliver the required functionality.

After medical center leadership reviewed and approved the basic vision, it became our task to put together a network of task forces, obtaining detailed information from potential clinical users of this system. This information primarily dealt with the functional goals of the system rather than technological goals in order to maximally involve practicing clinicians. This integrated information system was labeled PCIS (Patient Care Information System). Discussions and functional goal development went well with most groups (nursing, house officers, and other allied health professionals), with the important exception of attending physicians.

Attending physicians were concerned that PCIS did not hold enough value for them. Specifically, they felt that the emphasis was too strong on supporting the logistics of daily patient care (which was a rather positive feature for house officers and nursing staff) and not nearly strong enough on research and supervision of their practices. This realization precipitated a complete and important modification of our clinical information system strategic plan. The need to balance the introduction of a new clinical information system strategy with the existing operating infrastructure of our academic medical center resulted in a major and beneficial modification of the entire strategic information plan. These changes included:

* A developmental dimension to the

original PCIS project.

* The Applied Clinical Informatics 

(ACI) project. The PCIS project was modified to include the development of a relational clinical database allowing ad hoc query in support of both clinician practice management and clinical research. This relational database will be developed in the next 24 months on an industry standard platform and will allow access from the new ACI computing environment. The drive to develop this environment with the vendor of PCIS also was important to our medical staff, as it signaled a willingness to work toward more and more open data access environments that will be important to avoid technological "dead ends" that could stem the continued growth of our clinical computing environment. This change in our original plan for the PCIS integrated system also achieved a better fit of our operational plan with the clinical data model that integrates practice, education, and research. The ACI project completed this match of model and operational plan for our clinicians.

In keeping with the major elements of the physicians' task force requirements mentioned above, the ACI project is subdivided into two portions:

* Clinical workstation project.

* Medical Information Highway. The clinical workstation project involves the development of a clinically intuitive graphic environment designed to present the existing clinical information systems in familiar "laboratory, clinic, and home' modules. This allows the clinicians easier access to practice data in an information "tool kit" for incremental growth of the existing clinical information offerings. To achieve this goal, two professional systems developers were hired and proceeded to interview a large cross-section of our physician and allied health staff. Particular attention was paid to frustrations with the existing clinical information environment and to specific requirements for evolution of the system. The primary requirements identified were:

* Ease of use.

* Ease of clinical data retrieval.

* Flexibility of individual user

interface.

* Incremental increase in clinical

data available on line. Selection of the method of development--rapid prototyping--of this new graphic environment is as important as the end product. This technique is especially important in our clinical computing environment because it signals a total positive break with the historical and mutually disregarded method of interaction between Information Networking Services (INS) and the clinical user community. The historical process of data modeling involves collection of a large amount of information from users anxious to begin using a new clinical information system. This is followed by long and laborious program coding with significant technical "tweaking" without further interaction of the user to get the system "just right." Finally, when the long-anticipated product is released, users are disappointed because the final product does not seem to be what they asked for. Furthermore, the information systems professional, after putting so much effort into technically perfecting this clinical system, is thoroughly frustrated with the response of the clinical user. The rapid prototyping model recognizes these central truths regarding the relationship of clinical users (especially academic clinical users), the information system professional, and the current state of the art:

* Excellent software development

tools now available allow rapid

deployment of working prototypes

to users for validation of concepts

and real interaction provoking

further insights.

* The more creative the interaction

that occurs during the process of

rapid prototyping, the greater the

satisfaction on the part of the user.

Additionally there is greater

satisfaction for information

professionals when their efforts

have been appreciated.

* The shorter the interval between

expression of a need or an idea and

the realization of it in a program,

the more both the clinical user and

the IS professional learn the

elements of what needs to be

created. In the ACI clinical workstation (ACI CWS) subproject, the familiar desktop metaphor is expanded to make a clinically intuitive information environment allowing easy navigation to all of our existing clinical information system elements, as well as reliable and consistent information presentation integrated by the specialized task at hand. That is, the ACI CWS environment "packages" the existing CIS environment to logically present information on practice when in the "clinic" environment and ad hoc research data query when in the "laboratory" environment etc. The basic desktop availability of E-mail, on-line medical references, and context-sensitive help all seek to make the computing environment parallel the real world environments clinicians face on a daily basis. The "home view" was developed in recognition of the need to present a "remote" view of the entire CIS environment, be it truly at home or in on-call areas, outreach clinic offices, etc.

The Medical Information Highway (MIH), the second portion of the ACI project, is directed at the development of a common information architecture that allows ease of new system integration and incremental expansion to clinical data access for all users. It has been said that standards in computing must be a good idea because there are so many of them to choose from. The situation at our medical center, as with many academic centers throughout the country, is that our existing clinical computing environment is already so diverse that to select and settle on a single information exchange standard requires the patience of Job and the wisdom of Solomon.

A major breakthrough occurred when INS decided to avoid further propagation of classic point-to-point interface technology and began work on a new technology of interface "engines." The interface engine is a middle layer of technology used to allow growth and modification of systems on both the host (server, where the data reside) and the client (user) sides of the interface relationship without requiring the never-ending "tandem" configuration changes heretofore required for point-to-point interfaces. The MIH portion of ACI is an ongoing effort involving partnerships among INS, clinical departments, clinical users, and industry leaders in connectivity technology. Because of the fault-intolerant aspects of clinical information, we have rejected strategies of so-called "screen scraping," where information is electronically "read" off foreign systems displays and fed into other databases, recognized only as a function of the data's location on screen in a particular row and column. We live in such a high-flux clinical computing environment that this sort of reliance on data location on a screen would rapidly result in deterioration of the value of the data transferred. Work is proceeding with an "SQL" relational database management product allowing data recognition on the basis of strict definitions and reliable data structures. MIH will be both a strategically important and technically challenging portion of our information strategy.

From Cottage Industries to Common Purpose

The second major challenge to introducing clinical information systems was the very characteristic that makes our academic environment so powerful--diversity. To a significant degree, the core vision of an integrated clinical information system environment that all can use requires identification of common methods of communication; common language; and, above all, a common purpose. Although the rallying point of "improving the quality of patient care" may be sufficient in the community hospital, motivations for a academic medical center are more complex. This challenge is epitomized by the reaction to the PCIS initiative by the directors of clinical department information systems.

The managers of department information systems at UMMC have a tradition of independent system development and technology selection. To a great degree, these department clinical information systems arose as cottage industries within the medical center, responding to requirements primarily from the perspective of the controlling clinical department. When the PCIS project was introduced, the independence of these groups was threatened by the concept of "medical center wide" requirements and the focus on "overall institutional needs to improve the quality, effectiveness, and efficiency of delivering patient care." Although careful negotiation with department CIS directors allayed the majority of the concerns, progress was ultimately made only when two factors came to bear:

* Leadership.

* Acceptance of risk. It is difficult to overemphasize the importance of strong visionary leadership in resolving the fundamental conflict between the "cottage industry" model of academic center clinical information systems and the common purpose of an institutionwide clinical information system strategy. From the CEO, dean, chairmen, CFO, CIO, etc., the "top down" message of collaboration and common purpose is essential to institute important change in the development of institutionwide clinical information systems. At UMMC, these leaders responded to the challenge of resolving conflicts and concerns so that our clinical information strategic plan could move forward. At the same time, "grass roots" clinician user community support for these changes was vital to completing the critical elements for change in our academic center's computing environment. The latter support was only won through many months of discussions, education, site visits, and demonstrations.

Communication: An Operating Model for Clinical Information Systems

To effectively communicate a new approach to clinical information systems, necessary for both the top down and bottom up understanding and acceptance of the common vision, it is imperative to articulate an operating model of clinical systems that makes sense to technologists and practitioners alike. The model we have hit upon, clinical information in three dimensions, has been most helpful to our users in the vital challenge of communication.

The literature is replete with "data models," all attempting to describe the complexity and enormity of the automation of medical practice. To improve clarity and simplicity of communication, we've developed a model based on the standard three-dimensional coordinate system, consisting of X, Y, and Z axes (figure 1, page 22). The computer screen versions of this diagram are shown in figures 2-4, page 24.

The Y-axis represents a "vertical" view of clinical information. The information is centered around an individual patient at a particular time, e.g., a particular episode of care. All data--laboratory, radiological, etc.--appear in relation to the patient. It is here that most results reporting in clinical information systems takes place, relaying results to physicians via hospital, clinic, or office terminals.

As the health care environment has changed, however, it has become obvious that results reporting alone is not enough. Therefore, the Y dimension of clinical information also includes direct physician order entry. There are several reasons to include order entry:

* Closing the quality-of-care loop.

* Improvement in the efficiency of

health care delivery.

* Improvement in the cost of health

care. The X axis represents the longitudinal patient record. This is also called the lifetime or "birth to death" computerized patient record. This dimension involves a comprehensive database of all patient encounters with any aspect of the health care system (e.g., office visits, emergency clinic visits, hospitalizations). As indicated in the CPR report, the ultimate scope of this X dimension will include any and all sites of health care delivery that an individual patient may use in a lifetime. The "Hospital without Walls" project best exemplifies this dimension of clinical information systems. Clearly this aspect of CIS must be tightly integrated into the Y dimension to allow for CPR support of the elusive goal of continuity of medical care. The contributions of the X dimension of CIS toward the goals of improvement in the effectiveness and efficiency of health care delivery are also obvious.

The Z dimension of CIS may be described as the ability to search clinical information for relationships among information not necessarily associated with an individual patient. Logically, this is the relational database dimension of the CIS and performs the all-important function of demonstrating connections between innumerable outcomes across populations of patients, providers, therapeutic interventions, etc. Although this dimension of CIS represents the ultimate power of a computerized patient record, it is not realizable until dimensions X and Y are in place. This simple fact has been lost to many medical centers. Those that hope for the richness of a complete clinical information environment from largely financial databases will be disappointed.

The Z axis' characteristics must include a flexible data access query method for ad hoc information search and retrieval. As has been postulated by numerous authors, if this capability were available in every health care system, each center could act as its own clinical research environment, greatly shortening the time and effort needed to conduct clinical trials. The Mayo Clinic experience using a structured paper medical record system gives a glimpse of what could be accomplished at every center after achievement of the Z dimension of CIS. This includes the ability to search and retrieve specific text elements within unstructured paragraphs of medical information.

Conclusion

We have attempted to present some of the key challenges to change in clinical information systems in the special environment of the academic medical center. Although we have spent little time discussing it, it is the third leg of the three-legged stool that holds the most promise in making these challenges to change less daunting in the future: namely, education.

Although there are several excellent exceptions, in 1993 there were few academic medical centers where applied medical informatics is a substantial and integrated portion of the curricula for medicine, nursing, and other allied health students. "Bench" research in medical informatics continues at a number of institutions, but the incorporation of clinical informatics (the applied dimension of the science) lags far behind. However, this situation is slowing changing. Locally, we are finishing plans to launch a program of Applied Medical Informatics for third-year medical students just prior to their beginning clinical work at the university hospital. This program will emphasize the importance to the student of mastering clinical computing as another of the information instruments that she or he will need in order to be a well-informed, effective physician. On the national scene, renewed efforts are under way by the American Medical Informatics Association (AMIA) and other groups to complete the discipline of medical informatics with applied curricula for clinicians and special postgraduate training opportunities. Once the next generation of clinicians fully understands the potential of applied medical informatics we will finally witness rapid and profound change in clinical information systems.

William F. Bria, MD, is Medical Director of Clinical Information Systems, Medical Director of the Critical Care Medical Unit, and an Assistant Professor of Medicine in the Department of Internal Medicine at the University of Michigan Medical Center, Ann Arbor.