Colloquium: What is Biomedical Informatics and Why is it Hard?

Marksbury Building Theater


Todd R. Johnson, Phd
Professor and Director, Division of Biomedical Informatics, Dept. of Biostatistics, College of Public Health
The University of Kentucky

Abstract
Over the past 30 years, computer power has drastically increased and many activities now benefit from information technology. Despite some early successes in biomedicine, most notably in genomics and proteomics, many areas of biomedicine have not benefited.  This is particularly prevalent at the clinical level, where the failure rate of information technology is over 50%, and studies have shown that even successful IT has had little effect on either cost or quality of care. Although many factors influence the success of IT, in this talk we argue that the nature of biomedical information and information processes makes them inherently difficult to automate with computers. We lay out this argument by proposing a framework for understanding the nature of information, information processing, and the role of computers and computation in information systems. This framework pinpoints several sources of difficulty with the use of computers in biomedicine and suggests that to effectively use computers in biomedicine, we must fundamentally change the way we design, view, and interact with our information systems.  We conclude by briefly describing a two-pronged approach to designing information systems for biomedicine. The first, which we call Collaborative Semantic Information Processing, involves designing information systems that explicitly support the complementary abilities of humans and machines for processing meaning vs. symbols. The second approach involves a shift from primarily symbolic, logical modeling of biomedical concepts to a probabilistic approach that explicitly represents the uncertainty and vagueness implicit in most biomedical concepts and terms.