Determinants of diagnostic hypothesis generation: Effects of information, base rates, and experience.

Abstract

Physicians generated diagnostic hypotheses for case histories for which 2 types of diagnoses were plausible, with one having a higher population base rate but less severe clinical consequences than the other. The number of clinical and background symptoms pointing towards the 2 diagnoses was factorially manipulated. The order and frequency with which physicians generated hypotheses varied with the amount of relevant clinical and background information and as a function of population incidence rates, with little evidence of base rate neglect. Availability of a hypothesis, made possible by diagnosis of a similar case before, also made doctors generate this diagnosis earlier and more frequently. Physicians' experience affected hypothesis generation solely by increasing the availability of similar cases. The results are consistent with the use of similaritybased hypothesis generation processes that operate on memory for prior cases. The process of hypothesis generation in problem solving has received inadequate attention. Increasing concern with problem-solving skills in areas as diverse as medicine or manufacturing has resulted in descriptive and prescriptive accounts of human problem solving that typically divide the solution process into the subprocesses of hypothesis generation, information acquisition, and hypothesis testing. However, psychological researchers from Bruner, Goodnow, and Austin (1956) through Wason and Johnson-Laird (1972), Newell and Simon (1972) to Klayman and Ha (1987), have

How certain is the uncertainty effect?

“…), as opposed to encountering symbolic descriptions thereof, can strongly influence people's reasoning and judgments (e.g., Gigerenzer et al 1988;Weber et al 1993;Koehler 1996;Hertwig et al 2004). The effect of our physical lottery implementation could be another manifestation of this finding.…”

Overestimation of base-rate differences in complex perceptual categories

“…Although much anecdotal evidence suggests that humans are accurate categorizers, and thus should be sensitive to base-rate information, many studies suggest that people underestimate or ignore base-rate information, leading to the claim that often humans show base-rate neglect (see, e.g., Balla, 1982;Balla, Elstein, & Gates, 1983;Casscells, Schoenberg, & Graboys, 1978;Kahneman & Tversky, 1973;Tversky & Kahneman, 1974;however, see Wallsten, 1981;Weber, Bockenholt, Hilton, & Wallace, 1993). These studies used story problems in which information about the category exemplars and base-rates was presented explicitly as percentages or probabilities.…”

Prevalence and alternative explanations influence cancer diagnosis: An experimental study with physicians.

“…In addition, we found that physicians gave cancer as their main working diagnosis and referred the patient urgently (i.e., within two weeks) much more often with repeated presentations that suggested a worsening of the condition, and more often for the colorectal cancer case, which included "alarm" symptoms, than for the lung cancer case, which included only vague symptoms. and the literature in the field (Dougherty & Hunter, 2003a;Thomas et al, 2008;Weber et al, 1993). It appears to contrast with the literature on base-rate neglect.…”

“…First, physicians perceiving a low probability of cancer in patients presenting with vague symptoms may fail to generate the cancer hypothesis or exclude it as an implausible explanation for the patients' symptoms in the hypothesis evaluation phase. Indeed, people tend to first generate the hypotheses with a high a priori probability (Dougherty & Hunter, 2003a;Weber, Böckenholt, Hilton, & Wallace, 1993). Physicians may also fail to diagnose cancer if there is an alternative, more common, explanation for the symptoms in an early stage of the diagnostic process.…”