Expert and novice clinicians judged the likelihood of disease alternatives and gave thinking-aloud protocols as they evaluated simulated cases of congenital heart disease. Specific combinations of cues in the patient data were manipulated to create alternate versions of a single case so that the use of critical cues could be identified. Analyses of variance of subjects' disease judgments revealed differences between expert and novice clinicians in their use of critical cues and cue combinations. Analyses of the thinking-aloud protocols revealed that clinicians with different degrees of expertise employed different interpretations of patient data cues as well as qualitatively distinct "lines of reasoning" in reaching clinical judgments.
Bibliographical noteFunding Information:
In experimental tasks of clinical judgment, individuals are typically presented with several sets of stimulus cues which must be interpreted and combined to make either numerical judgments (e.g., rating the severity of a disease) or categorical classifications (e.g., a malignant or benign tumor). Usually, the values or characteristics of the cues are predetermined based upon the nature of the investigation. If the judgments of each cue combination are considered as dependent variables, a multiple regression equation can be computed to identify the weight (relative importance) of cues used in making judgments. The regression equation is often presumed to represent the individual's judgment "policy." If the various cue combinations are orthogonal, an analysis of variance (ANOVA) technique can be used to indicate the cues or patterns of cues that were significant factors in making judgments. Both the multiple regression and ANOVA approach have been used for "capturing" policies in a broad variety of judgment tasks (Slovic, Fischhoff, & Lichtenstein, 1979; Slovic The research reported here has been funded by grants to the first author from the graduate school at the University of Minnesota, the Center for Research in Human Learning at the University of Minnesota, the National Institute for Child Health and Human Development (T36-HD-07151 and HD-01136), the National Science Foundation (NSF/BNS-77-22075), the University of Minnesota Consulting Group on Instructional Design, and the Dwan Family Fund in the University of Minnesota Medical School. We would express our appreciation to staff and students in the Department of Pediatrics at the University of Minnesota Medical School who have generously given their time and effort to this research. Requests for reprints should be sent to Dr. Paul E. Johnson, Graduate School of Management and the Center for Research in Human Learning, University of Minnesota, 205 Elliott Hall, 75 East River Road, Minneapolis, MN 55455.