Increasingly, medical choices involve deciding whether to look for evidence of undetected, asymptomatic conditions, or increased risk of future conditions (i.e. screening). Those who screen at sufficiently high risk face decisions about interventions to prevent or postpone the onset of possible, but not certain, future symptomatic conditions. Other preventive decisions include whether or not to accept population-based intervention, such as vaccination. Using decision trees, we model the normative structures and associated uncertainties that underlie five medical decision situations, each of which involves assessing the probabilistic hypothesis that a person has, or will in the future have, a given symptomatic condition. The probability estimate that results from assessment becomes an input into predicting treatment benefit, with the probability of benefit decreasing as that of the symptomatic condition decreases. The five situations identified in this paper involve assessing: (1) a symptomatic patient; (2) an asymptomatic individual for an undetected condition; (3) an individual for risk of a future condition; (4) an individual for multiple risks simultaneously (shotgun assessment); and (5) an individual for a population-based intervention. Analysis of these situations facilitates examination of intuitive probabilistic reasoning. Drawing on evidence in related literature, we discuss some implications of decision-makers imposing the wrong structure or probabilistic reasoning when making medical choices. In particular, we discuss (1) overestimation of expected benefit due to systematic underestimation of uncertainty in a given decision; (2) overconfidence in probabilistic test results; and (3) failure to understand the implications of cumulative probabilities when 'shot-gun' testing.