Uncertain decisions regarding stroke symptoms: Changing bias through consequences

The impact of stroke on the lives of individuals and the health-care system is considerable. Damage from stroke can be limited if the treatment is administered at the appropriate time, so early recognition is essential. Some common interventions (e.g., FAST) designed to help potential stroke victims discriminate stroke symptoms often result in false negatives. Strokes can present with a wide variety of symptoms, making it difficult to discriminate stroke symptoms from non-stroke symptoms. Because the probability that a given set of symptoms are stroke symptoms is typically unknown to the victim, the problem is a decision under conditions of uncertainty. Signal detection methodology allows us to consider the ability of an individual or group to discriminate between stroke symptoms and non-stroke symptoms, as well as measure the motivation or bias toward a particular decision. We examined the effects of levels of feedback on performance of a random sample of participants from Amazon Mechanical Turk. We found that feedback designed to generate liberal bias toward stroke detection yielded fewer misses than FAST while maintaining a false alarm rate below 50%. Given that strokes are difficult to discriminate, this suggests that interventions should be focused on incentivizing help-seeking behaviors in conditions of uncertainty for those most at risk.