Several prospective studies have suggested that emergency physicians make approximately 80% of their decisions within 15 seconds of meeting a patient. The precision on these estimates could be challenged as no mind can be read out like a FORTRAN card. Whether or not the truth is actually 50% and 50 seconds, few experts doubt that a large component of diagnostic hypothesis generation occurs via type I, or fast processing, also known as “intuition”. Ironically, the vast majority of clinical didactic training focuses on type II, or analytical processing, which integrates knowledge and data analysis and knowledge, and unfolds more slowly, over minutes to hours. This raises a controversy in educational theory, inasmuch as no evidence has emerged to show that type II thinking can overcome erroneous intuition.
Experts in medical education generally agree that intuition is important to instant decision making in emergency care, but few can state what physicians use to formulate intuition. However, any emergency care practitioner can make prudent guesses about the makeup of the mental machinery of intuition. Intuition surely requires pattern recognition, and patterns comprise multiple data inputs such as past medical history, patient age and current data, such as chief complaint, vital signs, non-verbal communication, and trust in patient factors. The magnitude of the contribution of these components to the “sick/not sick” decision has not been measured. Moreover, the ability to meld all of these components into an impression that is stronger than its components—the definition of gestalt reasoning—has also not been assessed.
To help close this critical gap in knowledge, Dr. Steve Wipprecht, mentored by Dr. Jeffrey Kline, will undertake a simple observational study in which research associates will collect emergency physicians’ estimates of the probability of a true emergency on a visual analogue scale, ranging from 0 to 100%, and will ask the degree to which the physician used the components of intuition. This assessment will be made as soon as practicable after the physician first sees the patient. The research team will then follow patients prospectively to determine if they meet explicit criteria for a true emergency (e.g., acute organ injury, significant infection, vascular failure, new cancer or death). The primary analysis will test the diagnostic accuracy of intuition and the magnitude of importance of the pattern components described above as reported by the physician. In a hypothesis finding experiment, they will then determine the features of incorrect intuition, including false positives (the VAS was >505 and the patient had no evidence of any significant medical problem) or false negatives (VAS <10% and the patient had an explicit emergency). Through this hypothesis finding experiment, the researchers hope to find factors which lead emergency physicians astray in decision-making.
The intermediate goal of this work is to identify modifiable factors that can improve diagnostic intuition. For example, if the researchers find that extremes in trust are associated with false positive intuition (“this patient would not be here unless he/she were sick”) or possibly worse, false negative intuition ( “He/she is always here with this complaint”), this could allow a cognitive intervention to reduce framing and premature closure based upon trust, or implicit biases. This stage will only seek to identify the factors, and whether these factors can then be modified by heuristics which are shortcuts in thinking. One such method is cognitive forcing, which can take the form of machine learning-driven medical record alert to remind the physician to beware of pattern mismatch if a “frequent flyer” for the first time shows up with tachycardia. The long-range goal is to teach intuition without the need for the mythical 10,000 cases.
