Medical Informatics


University of Minnesota Medical School
Creating the clinician's workstation for better decision making

Reproduced with permission from NeXT Computer, Inc.
A Reference Guide to NeXT in Higher Education, Fall 1992
ยช 1992 NeXT Computer, Inc


Hospitals are data-rich environments. Doctors can turn to numerous sources to determine a patient's diagnosis and treatment. Unfortunately, this information is typically spread across many areas of the hospital and arrives in bits and pieces throughout a patient's stay. That makes it difficult to access data quickly or to review key data that might be buried in a flood of paper.

"Forced to make speedy decisions, doctors often find themselves relying heavily on the vagaries of memory, making decisions with less than all the available and relevant data," says Donald Connelly, M.D., associate professor of laboratory medicine and pathology.

Recently, though, Connelly and other medical information scientists in the Department of Laboratory Medicine and Pathology at the University of Minnesota Medical School have begun to develop a method for doctors to obtain more quickly the objective information they need, as well as feedback that encourages student physicians to follow practices that are recommended by the medical staff. With funding from the National Institutes of Health and the University of Minnesota Hospital and Clinic, a team of medical informatics researchers, physicians, nurses, and laboratorians are creating a NeXT computer-based clinicians' workstation to provide doctors and nurses with ready access to a variety of information-in this case, pertaining to transfusion medicine and clinical laboratory test results.

"Our intent is to provide our clinicians easy access to laboratory results in a form that makes the data easy to review and use," says Connelly. "We then hope to improve transfusion medicine practices at our hospital."

The workstation is connected to the hospital's laboratory information system and automatically receives and stores results as soon as they are completed in the laboratory. Physicians can review a patient's status at a glance and order platelet transfusions and other laboratory tests for patients with hematological disorders. The workstation then gives instant feedback on the appropriateness of the order, based on clinical guidelines developed by the local medical staff. These guidelines have been incorporated into the knowledge base of the workstation's expert system, Expert System for Platelet Request Evaluation, or ESPRE.

According to co-investigator Bruce Sielaff, Ph.D., "As with many expert systems, ESPRE was originally developed in LISP. The flexibility of the LISP environment greatly simplifies the incremental development characteristic of expert systems. On the other hand, to simplify integration of the expert system into the clinical workstation application, when the design stabilized, ESPRE was rewritten in Objective C. The object-oriented environment allowed the conversion to be relatively painless."

After analyzing the situation, ESPRE indicates which laboratory-based criteria for transfusion have been met. If no laboratory-based criteria seem to be satisfied, the expert system asks for additional clinical information that might satisfy other criteria for transfusion. Since guidelines cannot cover all possible clinical situations, physicians can also type in a reason why transfusion might be necessary.

Connelly says the development team has relied extensively on NeXT's Interface Builder in working with physicians, laboratorians, and nurses to develop the application. "Interface Builder has been very valuable in working with clinicians," he says. "We bring our workstation prototype to introductory meetings with our clinicians. Within 20 minutes we can cover the basic concepts underlying the project and be actively working on critiquing and improving the prototype. It's much easier for busy clinicians to consider and critique a model of the workstation than having to wade through a long list of technical specifications."

He continues, "We wouldn't be able to complete the workstation in time if we were working in any other development environment. We'd have to do everything on paper first; get people to sign off on specifications; and then, six months later, come back hoping we had created what was asked for, as well as hoping they asked for what they needed."

Connelly says, "The modern hospital environment is too complex and dynamic for conventional development approaches." He adds that because of the flexibility of the NeXT platform, the development team has incorporated a comments button so clinical users can easily send developers suggestions for improvement.

This summer, the workstation will be implemented in four patient-care units at the University of Minnesota Hospital. A crossover trial will be conducted to determine the effect of feedback on interns' knowledge of transfusion medicine principles and on platelet transfusion practice. The acceptability of the workstation to clinicians will also be assessed.

Connelly adds that the application is not limited to use in transfusion medicine, but can be adapted to other areas of medicine as well. "We have already expanded it to ordering and scheduling laboratory tests that are part of bone marrow transplantation research protocols," he says. "This is the beauty of working on the NeXT machine. We develop building blocks in a short amount of time and can then use them for other projects. Interface Builder has been a real lever for us. If we had been working on any other system, I'm sure we'd still be specifying functions and worrying how we were going to implement them. Instead, we are concentrating on how we can help our clinical colleagues do their very important tasks better and easier."

For more information, please contact:

Donald P. Connelly, M.D., Ph.D.
Associate Professor of Laboratory Medicine and Pathology
University of Minnesota Medical School
BOX 198 UMHC
Minneapolis, MN 55455
(612) 624-4689
doncon@tahiti.umhc.umn.edu