Biomedical informatics is intrinsically entwined with the substance of biomedical science. It determines and analyzes the structure of biomedical information and knowledge, whereas biomedical science is constrained by that structure. Biomedical informatics melds the study of biomedical computer science with analyses of biomedical information and knowledge, thereby addressing specifically the interface between computer science and biomedical science.

Biomedical informatics is perhaps best viewed as a basic biomedical science, with a wide variety of potential areas of application (see diagram). The analogy with other basic sciences is that biomedical informatics uses the results of past experience to understand, structure, and encode objective and subjective biomedical findings and thus to make them suitable for processing. This approach supports the integration of the findings and their analyses. In turn, the selective distribution of newly created knowledge can aid patient care, health planning, and basic biomedical research.

Biomedical informatics is, by its nature, an experimental science. An experimental science is characterized by posing questions, designing experiments, performing analyses, and using the information gained to design new experiments. One goal is simply to search for new knowledge, i.e., basic research. A second goal is to use this knowledge for practical ends—applications research. There is a continuity between these two endeavors. In biomedical informatics, there is an especially tight coupling between the application areas, broad categories of which are indicated at the bottom of the diagram, and the identification of basic research tasks that characterize the scientific underpinnings of the field.

Work in biomedical informatics is inherently motivated by problems encountered in a set of applied domains in biomedicine. The first of these historically has been clinical care (including medicine, nursing, dentistry, and veterinary care), areas of activity that demand patient-oriented informatics applications. I refer to this area as clinical informatics. Closely tied to clinical informatics is public health informatics, in which similar methods are generalized for application to populations of patients rather than to single individuals. Thus clinical informatics and public health informatics share many of the same methods and techniques. Two other large areas of application overlap in some ways with clinical informatics and public health informatics. These include imaging informatics (and the set of issues developed around both radiology and other image-management and image-analysis domains such as pathology, dermatology, and molecular visualization). Finally, there is the burgeoning area of bioinformatics*, which at the molecular and cellular level is offering challenges that draw on many of the same informatics methods as well.

As is shown in the diagram, there is a spectrum as one moves from left to right across these application domains. In bioinformatics, workers deal with molecular and cellular processes in the application of informatics methods. At the next level, workers focus on tissues and organs, which tend to be the emphasis of imaging informatics work (also called structural informatics at some institutions). Progressing to clinical informatics, the focus is on individual patients, and finally to public health, where researchers address problems of populations and of society. Biomedical informatics has important contributions to make across that entire spectrum.

In general, biomedical informatics researchers derive their inspiration from one of the application areas, identifying fundamental methodologic issues that need to be addressed and testing them in system prototypes or, for more mature methods, in actual systems that are used in clinical or biomedical research settings. One important implication of this viewpoint is that the core discipline is identical, regardless of the area of application that a given individual is motivated to address. This argues for unified biomedical informatics educational programs, ones that bring together students with a wide variety of applications interests. Elective courses and internships in areas of specific interest are of course important complements to the basic core exposures that students should receive, but, given the need for teamwork and understanding in the field, it would be counterproductive and wasteful to separate trainees based on the application areas that may interest them.

Biomedical informatics draws from all of these activities—development of hardware, software, and computer-science theory. Biomedical computing generally has not had a large enough market to influence the course of major hardware developments; that is, computers have not been developed specifically for biomedical applications. Not since the early 1960s (when health-computing experts occasionally talked about and, in a few instances, developed special medical terminals) have people assumed that biomedical-computing applications would use hardware other than that designed for general use.

How, then, does biomedical informatics differ from biomedical computer science? Is the new discipline simply the study of computer science with a “biomedical flavor?” We believe biomedical informatics is more than simply the biomedical application of computer science. The issues that it addresses not only have broad relevance to health, medicine, and biology, but the underlying sciences on which biomedical informatics professionals draw are inherently interdisciplinary as well. Thus, for example, successful biomedical informatics research will often draw on and contribute to computer science, but it may also be closely related to the decision sciences (probability theory, decision analysis, or the psychology of human problem-solving), to cognitive science, to information sciences, or to the management sciences (see diagram below). Furthermore, a biomedical informatics researcher will be tightly linked to some underlying problem from the real world of health or biomedicine. As the diagram illustrates, for example, a biomedical informatics basic researcher or doctoral student will accordingly be motivated by one of the application areas, such as those shown at the bottom of the earlier diagram, but the dissertation worthy of a PhD in the field will usually be identified by a generalizable scientific result that also contributes to one of the component disciplines noted below and on which other scientists can build in the future.

Edward H. Shortliffe, MD, PhD
Department of Biomedical Informatics
Columbia University
http://www.dbmi.columbia.edu

*We use "bioinformatics" as the name of an applied informatics discipline, in contrast with "computational biology", in which the focus is on biology and computation provides the tools for solving biological problems. Computational biologists are fundamentally biologists; bioinformaticians are informaticians. We realize, however, that there is much confusion about the relationship of the two terms and that the differences proposed here, although rational and reasonably clear, are not yet widely accepted.