An equation for excellence in clinical reasoning

Achieving diagnostic excellence is a universal goal. Excellence in diagnosis would help ensure effective and efficient diagnosis, minimize the risks of harm from diagnostic error, and reduce the social and economic costs of healthcare. But what do we mean by excellence? Understanding what comprises excellence and how to describe it would be essential first steps in organized efforts to promote excellence. Ideally, being able to describe excellence in an equation would be the clearest way to identify the critical elements that contribute to excellence, and the relationships between these elements.

The diagnostic process is complex, involving at its core the individual contributions to clinical reasoning involved in arriving at the diagnosis, and the many points of interaction of this process with the healthcare system and the context of care for each individual case. We propose that a valuable initial step in trying to describe diagnostic excellence is to separate out and focus on just the clinical reasoning process. In this paper we propose a new equation to describe excellence in clinical reasoning, based on original concepts described by Gary Klein from his studies of expertise in naturalistic decision making.

Klein proposed that excellence in decision-making could be described by a simple, general formula:

where P is performance excellence, realized by optimizing expertise and doing ‘all the right things’ to do things correctly and avoid errors (the down arrow, indicating error minimization), combined with insights later on (the up arrow) [1]. Insights refer to the subconscious emergence of ideas, decisions, and problem solutions that occur spontaneously at some later point in time.

We have proposed that Klein’s approach is relevant and instructive relative to diagnostic performance [2], and in this paper we describe a modified version of Klein’s equation to begin envisioning performance excellence in regard to clinical reasoning for diagnosis:

The equation says that optimal performance would result from doing ‘all the right things’ at the initial steps of developing the diagnosis, combined with any subsequent insights that emerge down the road. Any limitations encountered along the way would detract from excellence, and these include uncertainty, cognitive shortcomings, noise, and host of patient- and system-related contextual factors.

Expertise in clinical reasoning would reference all of the traditional elements of diagnostic excellence, including all of the knowledge and skills that an expert diagnostician would possess: An expert possesses a comprehensive knowledge base, including illness scripts for a very wide range of different conditions and their variants, coupled with a detailed knowledge of anatomy, physiology, biochemistry, etc. Expertise would also include the clinical skills essential for diagnosis: Being able to elicit the patient’s history, gain all the information possible from the physical examination, and extract all of the necessary information from medical records and test results. The expert would then be able to optimally synthesize this information in reference to their expert-level knowledge of the disease possibilities, to arrive at the most likely diagnostic possibilities. Finally, the expert would be skilled in using feedback and reflection to continually improve performance [3], and to adopt metacognitive processes to ensure that all the correct steps were taken, to look for inconsistencies and possible distortion by bias, and to ensure that the list of diagnostic possibilities is comprehensive and complete [4].

Insight doesn’t come into play very often in diagnosis, but when it does, it is a very powerful way to improve diagnosis and arrive at the correct answer [5]. Insights arise subconsciously, but many authors believe they can be facilitated through “System 2”, purposeful, analytical reasoning, including critical thinking and cognitive forcing strategies, such as using mnemonics or CubieDx [6]. Other approaches to facilitate insight include design thinking (a way of thinking that aims to determine the best solution to an unprecedented or unknown problem) [7], and abduction (setting up a new hypothesis that predicts a phenomenon when something unexpected happens) might also be helpful. The relationship between expertise and insight is likely to be synergistic, because individuals with expertise and skill in critical thinking are also likely to excel at experiencing insights later on. We therefore assume a multiplicative relationship between these two parameters, rather than an additive one.

The inherent limitations to diagnosis are important factors that predispose to diagnostic errors, and it is worthwhile considering how to limit their impact. Cognitive and affective bias, and shortcomings in critical thinking are common, but addressable. Second opinions, using decision support resources, self-reflection and active meta-cognition are all possible antidotes.

Uncertainty comes from the indeterminacy of future outcomes, lack of reliability, credibility, or adequacy of information, atypicality of the presenting symptom; the rarity of some diseases; and other contextual factors. After all is said and done, there is always some residual uncertainty involved in diagnosis, and a great deal of attention is being devoted to considering if and how to document this [8]. The least attractive option is to ignore it, leaving the patient and the next clinicians with the impression that the diagnosis is settled. Clearly documenting uncertainty, on the other hand, alerts everyone to keep an eye out for findings that indicate the need to reconsider the diagnosis. Being able to live with and manage the uncertainty inherent in diagnosis is a challenge for all clinicians, and their patients alike.

Noise is another source of error and refers to the randomness and variations in individual physicians’ judgments [9]. A physician’s training and experience is very idiosyncratic, and it is no wonder that two clinicians hearing the same story might arrive at very different diagnostic conclusions. Second and third opinions [10], and diagnosis via groups of clinicians [11] would seem to be appropriate answers to this problem, and hopefully will become more prevalent in clinical practice going forward.

Designating an equation that describes diagnostic excellence is novel, and provides a starting point for discussions on what comprises excellence and how to achieve it. Hopefully others will consider how to improve the equation going forward, and address its current shortcomings. Possibly the most important shortcoming of the equation is that too much is encompassed in the first term, ‘e’, excellence in the diagnostic process. Unpacking this and examining all of the factors inherent in this term would be valuable and worthwhile. A second problem is the apparent disconnection between an individual clinician’s diagnostic performance and the system in which he or she works. Both the context of care and a panoply of system factors are relevant to determining an individual clinician’s diagnostic performance. A third problem is that the equation applies to an individual clinician, whereas diagnostic excellence in the future is envisioned as been team-based. A more ideal equation would take this into account. Finally, the equation fails to account for the role that decision support and artificial intelligence might play in diagnosis. Going forward, these will be increasingly relevant to overall diagnostic performance.